Classification and Regression Using Supervised Learning 3 (Logistic Regression classifier)

Logistic regression is a technique that is used to explain the relationship between input variables and output variables. The input variables are assumed to be independent and the output variable is referred to as the dependent variable. The dependent variable can take only a fixed set of values. These values correspond to the classes of the classification problem.

Our goal is to identify the relationship between the independent variables and the dependent variables by estimating the probabilities using a logistic function. This logistic function is a sigmoid curve that's used to build the function with various parameters. It is very closely related to generalized linear model analysis, where we try to fit a line to a bunch of points to minimize the error. Instead of using linear regression, we use logistic regression. Logistic regression by itself is actually not a classification technique, but we use it in this way so as to facilitate classification. It is used very commonly in machine learning because of its simplicity. Let's see how to build a classifier using logistic regression as shown in the program below:

import numpy as np
from sklearn import linear_model
import matplotlib.pyplot as plt

# Define sample input data
X = np.array([[3.1, 7.2], [4, 6.7], [2.9, 8], [5.1, 4.5], [6, 5], [5.6, 5],[3.3, 0.4], [3.9, 0.9], [2.8, 1], [0.5, 3.4], [1, 4], [0.6, 4.9]])
y = np.array([0, 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3])

# Create the logistic regression classifier
classifier = linear_model.LogisticRegression(solver='liblinear', C=1)

# Train the classifier
classifier.fit(X, y)

# Visualize the performance of the classifier
visualize_classifier(classifier, X, y)

Our program begins with the necessary imports and then we define sample input data with two-dimensional vectors and corresponding labels.


X = np.array([[3.1, 7.2], [4, 6.7], [2.9, 8], [5.1, 4.5], [6, 5], [5.6, 5],[3.3, 0.4], [3.9, 0.9], [2.8, 1], [0.5, 3.4], [1, 4], [0.6, 4.9]])
y = np.array([0, 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3])



Once we have the labeled data we will train the classifier using this labeled data. First we'll create the logistic regression classifier object:


classifier = linear_model.LogisticRegression(solver='liblinear', C=1)
   
Then we train the classifier using the data that we defined earlier:   

classifier.fit(X, y)

Finally we visualize the performance of the classifier by looking at the boundaries of the classes:

visualize_classifier(classifier, X, y)

The visualize_classifier(classifier, X, y) function can either be defined in the same program or in another program. As we'll be using this multiple times in future posts,it's better to define it in a separate file and import the function. This function is given in the utilities.py file as shown below:

import numpy as np
import matplotlib.pyplot as plt

def visualize_classifier(classifier, X, y):
    # Define the minimum and maximum values for X and Y
    # that will be used in the mesh grid
    min_x, max_x = X[:, 0].min() - 1.0, X[:, 0].max() + 1.0
    min_y, max_y = X[:, 1].min() - 1.0, X[:, 1].max() + 1.0

    # Define the step size to use in plotting the mesh grid
    mesh_step_size = 0.01
    # Define the mesh grid of X and Y values
    x_vals, y_vals = np.meshgrid(np.arange(min_x, max_x, mesh_step_size),
    np.arange(min_y, max_y, mesh_step_size))
   
    # Run the classifier on the mesh grid
    output = classifier.predict(np.c_[x_vals.ravel(), y_vals.ravel()])
    # Reshape the output array
    output = output.reshape(x_vals.shape)
    # Create a plot
    plt.figure()
    # Choose a color scheme for the plot
    plt.pcolormesh(x_vals, y_vals, output, cmap=plt.cm.gray)
    # Overlay the training points on the plot
    plt.scatter(X[:, 0], X[:, 1], c=y, s=75, edgecolors='black',    linewidth=1, cmap=plt.cm.Paired)
   
    # Specify the boundaries of the plot
    plt.xlim(x_vals.min(), x_vals.max())
    plt.ylim(y_vals.min(), y_vals.max())
    # Specify the ticks on the X and Y axes
    plt.xticks((np.arange(int(X[:, 0].min() - 1), int(X[:, 0].max() + 1),    1.0)))
    plt.yticks((np.arange(int(X[:, 1].min() - 1), int(X[:, 1].max() + 1),    1.0)))
    plt.show()

We create the function definition by taking the classifier object, input data, and labels as input
parameters:

def visualize_classifier(classifier, X, y):

We also defined the minimum and maximum values of X and Y directions that will be used in our mesh grid. This grid is basically a set of values that is used to evaluate the function, so that we can visualize the boundaries of the classes.

min_x, max_x = X[:, 0].min() - 1.0, X[:, 0].max() + 1.0
min_y, max_y = X[:, 1].min() - 1.0, X[:, 1].max() + 1.0

Define the step size for the grid and create it using the minimum and maximum values:

mesh_step_size = 0.01
x_vals, y_vals = np.meshgrid(np.arange(min_x, max_x, mesh_step_size),
np.arange(min_y, max_y, mesh_step_size))

Next we run the classifier on all the points on the grid and then reshape the output array:

output = classifier.predict(np.c_[x_vals.ravel(), y_vals.ravel()])
output = output.reshape(x_vals.shape)

Once we have the output we create the figure, pick a color scheme, and overlay all the points:
plt.figure()

plt.pcolormesh(x_vals, y_vals, output, cmap=plt.cm.gray)

plt.scatter(X[:, 0], X[:, 1], c=y, s=75, edgecolors='black',linewidth=1, cmap=plt.cm.Paired)

Finally we specify the boundaries of the plots using the minimum and maximum values, add the tick
marks, and display the figure:

plt.xlim(x_vals.min(), x_vals.max())
plt.ylim(y_vals.min(), y_vals.max())

plt.xticks((np.arange(int(X[:, 0].min() - 1), int(X[:, 0].max() + 1),1.0)))
plt.yticks((np.arange(int(X[:, 1].min() - 1), int(X[:, 1].max() + 1),1.0)))
plt.show()

To use this in our main program let's import this function by adding the following statement in our program:

from utilities import visualize_classifier

Thus our program now is:

import numpy as np
from sklearn import linear_model
import matplotlib.pyplot as plt
from utilities import visualize_classifier

# Define sample input data
X = np.array([[3.1, 7.2], [4, 6.7], [2.9, 8], [5.1, 4.5], [6, 5], [5.6, 5],[3.3, 0.4], [3.9, 0.9], [2.8, 1], [0.5, 3.4], [1, 4], [0.6, 4.9]])
y = np.array([0, 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3])

# Create the logistic regression classifier
classifier = linear_model.LogisticRegression(solver='liblinear', C=1)

# Train the classifier
classifier.fit(X, y)

# Visualize the performance of the classifier
visualize_classifier(classifier, X, y)

When we run the program, we get the following output:

If we change the value of C to 100 in the following line, we will see that the boundaries have become more accurate:

classifier = linear_model.LogisticRegression(solver='liblinear', C=100)

The reason is that C imposes a certain penalty on misclassification, so the algorithm customizes more to the training data. We should be careful with this parameter, because if we increase it a lot, it will overfit to the training data and it won't generalize well. If we run the code with C set to 100, we get the following output:

Read More

Classification and Regression Using Supervised Learning 2 (Label encoding)

While performing classification we often deal with labels which can be in the form of words, numbers, or something else. The machine learning functions in sklearn expect them to be numbers. So if they are already numbers, then we can use them directly to start training. But this is not usually the case as in real world, labels are in the form of words, because words are human readable. We label our training data with words so that the mapping can be tracked. To convert word labels into numbers, we need to use a label encoder. Label encoding refers to the process of transforming the word labels into numerical form. This enables the algorithms to operate on our data. The following program will define some sample labels and create the label encoder object:

import numpy as np
from sklearn import preprocessing

# Sample input labels
input_labels = ['red', 'black', 'red', 'green', 'black', 'yellow', 'white']

# Create label encoder and fit the labels
encoder = preprocessing.LabelEncoder()
encoder.fit(input_labels)

# Print the mapping
print("\nLabel mapping:")
for i, item in enumerate(encoder.classes_):
    print(item, '-->', i)

The output of the program is shown below which print the mapping between words and numbers:

Label mapping:
black --> 0
green --> 1
red --> 2
white --> 3
yellow --> 4
------------------
(program exited with code: 0)

Press any key to continue . . .

The class sklearn.preprocessing.LabelEncoder() Encode labels with value between 0 and n_classes-1. It can be used to normalize labels. It can also be used to transform non-numerical labels (as long as they are hashable and comparable) to numerical labels.The fit() Fit label encoder.

Now let's encode a set of randomly ordered labels to see how it performs. See the following program:

import numpy as np
from sklearn import preprocessing

# Sample input labels
input_labels = ['red', 'black', 'red', 'green', 'black', 'yellow', 'white']

# Create label encoder and fit the labels
encoder = preprocessing.LabelEncoder()
encoder.fit(input_labels)

# Print the mapping
print("\nLabel mapping:")
for i, item in enumerate(encoder.classes_):
print(item, '-->', i)

# Encode a set of labels using the encoder
test_labels = ['green', 'red', 'black']
encoded_values = encoder.transform(test_labels)
print("\nLabels =", test_labels)
print("Encoded values =", list(encoded_values))

The output of the program is shown below which prints encoded values:

Label mapping:
black --> 0
green --> 1
red --> 2
white --> 3
yellow --> 4

Labels = ['green', 'red', 'black']
Encoded values = [1, 2, 0]

------------------
(program exited with code: 0)

Press any key to continue . . .

Our next program decodes a random set of numbers:

import numpy as np
from sklearn import preprocessing

# Sample input labels
input_labels = ['red', 'black', 'red', 'green', 'black', 'yellow', 'white']

# Create label encoder and fit the labels
encoder = preprocessing.LabelEncoder()
encoder.fit(input_labels)

# Print the mapping
print("\nLabel mapping:")
for i, item in enumerate(encoder.classes_):
print(item, '-->', i)

# Encode a set of labels using the encoder
test_labels = ['green', 'red', 'black']
encoded_values = encoder.transform(test_labels)
print("\nLabels =", test_labels)
print("Encoded values =", list(encoded_values))

# Decode a set of values using the encoder
encoded_values = [3, 0, 4, 1]
decoded_list = encoder.inverse_transform(encoded_values)
print("\nEncoded values =", encoded_values)
print("Decoded labels =", list(decoded_list))

The output of the program is shown below which prints encoded values:

Label mapping:
black --> 0
green --> 1
red --> 2
white --> 3
yellow --> 4

Labels = ['green', 'red', 'black']
Encoded values = [1, 2, 0]

Encoded values = [3, 0, 4, 1]
Decoded labels = ['white', 'black', 'yellow', 'green']

------------------
(program exited with code: 0)

Press any key to continue . . .

The inverse_transform() function transform labels back to original encoding. From the output we can check the mapping to see that the encoding and decoding steps are correct.

Read More

Classification and Regression Using Supervised Learning 1 (classification)

The process of classification is one such technique where we classify data into a given number of classes. During classification, we arrange data into a fixed number of categories so that it can be used most effectively and efficiently.

In machine learning, classification solves the problem of identifying the category to which a new data point belongs. We build the classification model based on the training dataset containing data points and the corresponding labels.

For example, let's say that we want to check whether the given image contains a person's face or not. We would build a training dataset containing classes corresponding to these two classes: face and no-face. We then train the model based on the training samples we have. This trained model is then used for inference.

A good classification system makes it easy to find and retrieve data. This is used extensively in face recognition, spam identification, recommendation engines, and so on. The algorithms for data classification will come up with the right criteria to separate the given data into the given number of classes.

We are required to provide a sufficiently large number of samples so that it can generalize those criteria. If there is an insufficient number of samples, then the algorithm will overfit to the training data. This means that it won't perform well on unknown data because it fine-tuned the model too much to fit into the patterns observed in training data. This is actually a very common problem that occurs in the world of machine learning. It's good to consider this factor when you build various machine learning models.

Preprocessing data

Machine learning algorithms expect data to be formatted in a certain way before they start the training process. In real world we deal with a lot of raw data and in order to prepare the data for ingestion by machine learning algorithms, we have to preprocess it and convert it into the right format. There are several different preprocessing techniques such as -

• Binarization
• Mean removal
• Scaling
• Normalization

Let's start with:

1. Binarization

This process is used when we want to convert our numerical values into boolean values. The following program shows how to do it:

import numpy as np
from sklearn import preprocessing

input_data = np.array([[5.1, -2.9, 3.3],[-1.2, 7.8, -6.1],[3.9, 0.4, 2.1],[7.3, -9.9, -4.5]])

# Binarize data
data_binarized = preprocessing.Binarizer(threshold=2.1).transform(input_data)
print("\nBinarized data:\n", data_binarized)

We start with importing the packages and then define some sample data which is stored in the variable input_data.

input_data = np.array([[5.1, -2.9, 3.3],[-1.2, 7.8, -6.1],[3.9, 0.4, 2.1],[7.3, -9.9, -4.5]])

Next we use an inbuilt method to binarize input data using 2.1 as the threshold value.

data_binarized = preprocessing.Binarizer(threshold=2.1).transform(input_data)

Once we run the code, we will see the following output:

Binarized data:

[[ 1. 0. 1.]
[ 0. 1. 0.]
[ 1. 0. 0.]
[ 1. 0. 0.]]

From the output we can see that all the values above 2.1 become 1. The remaining values become 0. Hence our objective is fulfilled.

The next preprocessing technique is:

2. Mean removal

This preprocessing technique is useful to remove the mean from our feature vector, so that each feature is centered on zero. We do this in order to remove bias from the features in our feature vector. See the following program:

import numpy as np
from sklearn import preprocessing

input_data = np.array([[5.1, -2.9, 3.3],[-1.2, 7.8, -6.1],[3.9, 0.4, 2.1],[7.3, -9.9, -4.5]])

# Print mean and standard deviation
print("\nBEFORE:")
print("Mean =", input_data.mean(axis=0))
print("Std deviation =", input_data.std(axis=0))

# Remove mean
data_scaled = preprocessing.scale(input_data)
print("\nAFTER:")
print("Mean =", data_scaled.mean(axis=0))
print("Std deviation =", data_scaled.std(axis=0))

We start with importing the packages and then define some sample data which is stored in the variable input_data.

input_data = np.array([[5.1, -2.9, 3.3],[-1.2, 7.8, -6.1],[3.9, 0.4, 2.1],[7.3, -9.9, -4.5]])

Next we display the mean and standard deviation of the input data:

print("\nBEFORE:")
print("Mean =", input_data.mean(axis=0))
print("Std deviation =", input_data.std(axis=0))

Finally we remove the mean and print the mean and standard deviation:

data_scaled = preprocessing.scale(input_data)
print("\nAFTER:")
print("Mean =", data_scaled.mean(axis=0))
print("Std deviation =", data_scaled.std(axis=0))

Once we run the code, we will see the following output:

BEFORE:
Mean = [ 3.775 -1.15 -1.3 ]
Std deviation = [ 3.12039661 6.36651396 4.0620192 ]
AFTER:
Mean = [ 1.11022302e-16 0.00000000e+00 2.77555756e-17]
Std deviation = [ 1. 1. 1.]

From the output we can see that the mean value is very close to 0 and standard deviation is 1.

The next preprocessing technique is:

Scaling

Usually in our feature vector, the value of each feature can vary between many random values. So it
becomes important to scale those features so that it is a level playing field for the machine learning algorithm to train on. We don't want any feature to be artificially large or small just because of the nature of the measurements. See the following program:

import numpy as np
from sklearn import preprocessing

input_data = np.array([[5.1, -2.9, 3.3],[-1.2, 7.8, -6.1],[3.9, 0.4, 2.1],[7.3, -9.9, -4.5]])

# Min max scaling
data_scaler_minmax = preprocessing.MinMaxScaler(feature_range=(0, 1))
data_scaled_minmax = data_scaler_minmax.fit_transform(input_data)
print("\nMin max scaled data:\n", data_scaled_minmax)

Once we run the code, we will see the following output:

Min max scaled data:
[[ 0.74117647 0.39548023 1. ]
[ 0. 1. 0. ]
[ 0.6 0.5819209 0.87234043]
[ 1. 0. 0.17021277]]

From the output we can see that each row is scaled so that the maximum value is 1 and all the other values are relative to this value.

The next preprocessing technique is:

Normalization

Normalization is used to modify the values in the feature vector so that we can measure them on a common scale. In machine learning, we use many different forms of normalization. Some of the most common forms of normalization aim to modify the values so that they sum up to 1.

L1 normalization, which refers to Least Absolute Deviations, works by making sure that the sum of absolute values is 1 in each row. L2 normalization, which refers to least squares, works by making sure that the sum of squares is 1.

In general, L1 normalization technique is considered more robust than L2 normalization technique because it is resistant to outliers in the data. A lot of times, data tends to contain outliers and we cannot do anything about it. We want to use techniques that can safely and effectively ignore them during the calculations. If we are solving a problem where outliers are important, then maybe L2 normalization becomes a better choice. See the following program:

import numpy as np
from sklearn import preprocessing

input_data = np.array([[5.1, -2.9, 3.3],[-1.2, 7.8, -6.1],[3.9, 0.4, 2.1],[7.3, -9.9, -4.5]])

# Normalize data
data_normalized_l1 = preprocessing.normalize(input_data, norm='l1')
data_normalized_l2 = preprocessing.normalize(input_data, norm='l2')
print("\nL1 normalized data:\n", data_normalized_l1)
print("\nL2 normalized data:\n", data_normalized_l2)

Once we run the code, we will see the following output:

L1 normalized data:
[[ 0.45132743 -0.25663717 0.2920354 ]
[-0.0794702 0.51655629 -0.40397351]
[ 0.609375 0.0625 0.328125 ]
[ 0.33640553 -0.4562212 -0.20737327]]

L2 normalized data:
[[ 0.75765788 -0.43082507 0.49024922]
[-0.12030718 0.78199664 -0.61156148]
[ 0.87690281 0.08993875 0.47217844]
[ 0.55734935 -0.75585734 -0.34357152]]

From the output we can see L1 and L2 normalized data.

Read More

CRM Artificial Intelligence Trends across Salesforce

The amount of data companies have on customers and the number of channels customers are using to interact with businesses have grown significantly in the past decade. Artificial intelligence may hold great promise in optimizing customer and client interactions.

The five largest Customer Relationship Management (CRM) vendors by market share in 2015 were Salesforce, Oracle, SAP, Adobe Systems, and Microsoft. These five companies make up almost half of the entire CRM market. All of them have been investing in their internal development of machine learning and AI, while also buying AI startups.

We will explore the AI applications of each of these five CRM leaders, helping business readers to understand:

1. AI capabilities are currently available for each of the five CRM giants
2. Tangible results have been yielded by today’s AI applications in CRM
3. Upcoming AI applications are being developed and may be available soon

This post will focus on is Salesforce. They are by far the dominant player in this sector with more market share than their top four competitors combined. They have been aggressively investing in AI to maintain their position and are the benchmark by which most other companies are being judged by possible clients.

Salesforce wants corporate clients to be convinced that AI is this the future of CRM. An IDC White Paper, sponsored by Salesforce, projects that the use of AI in CRM will boost global business revenue by $1.1 trillion from 2017-2021. They claim that it could result in 800,000 net new jobs.

Based on corporate data and their surveys, IDC believes this use of AI will grow dramatically in the next few years. A survey conducted by IDC found that 28% of all respondents say their organizations have already started using AI and an additional 41% plan to adopt AI in the next two years. IDC concluded that worldwide spending on cognitive/AI systems was only $8 billion in 2016 but believes that this spending could reach $46 billion by 2020.

It is probably in Salesforce’s interest to heavily emphasize the prevalence of AI in CRM, as they are the largest firm in the space, and the one most likely to invest heavily in CRM AI applications. For this reason we should take this report with a grain of salt. Whether or not you believe IDC’s projections, CRM services is a highly competitive and growing business sector.

Over the past few years, Salesforce has been aggressively developing AI services in-house and acquiring many AI companies. They have also signed major deals with other companies focused on AI.

According to CB Insights, Salesforce acquired four AI startups (Tempo, MinHash, PredictionIO, and MetaMind) during 2015 and 2016 to add their technology to their applications.

1. Tempo – An AI powered smart calendar app that shows you the information you need before a meeting, which Salesforce has since shut down.
2. MinHash – An AI that looks for marketing trends, which Salesforce has since shut down.
3. PredictionIO – An open source machine learning company, which Salesforce acquired for their technology and donated the open source addition to the Apache Software Foundation  
4. MetaMind – A deep learning company that specialized in technology, which Salesforce has since shut down.

This focus on AI allowed Salesforce to introduce their AI tool, Einstein, in late 2016. Their goal is to simplify the use of AI for their clients, and make AI capabilities accessible to clients without a robust technical AI skill set. This emphasis on “accessibility” is a common value proposition for nearly all B2B AI applications, in CRM or otherwise – and time will tell if Salesforce can indeed make the tools accessible to less technical users.

Salesforces offers a broad range of uses for Einstein, including account insights, lead prioritization, automated data entry, ad personalization, insights into social media conversations, product recommendations, image classification, and more.

Black Diamond, a Utah-based outdoor equipment company, claims that it initially used to manually upload on-site product recommendations for its customers, which ate up their time. These manual recommendations were not personalized for the customer, meaning that it most likely did not recommend related or relevant products that the customer was most likely to buy.

Black Diamond also claims in its case study that after integrating its system with Einstein’s recommendation engine, which uses customers’ historic data, purchase behavior, and affinities to offer relevant recommendations through machine learning, it was able to increase conversions by 9.6 percent and drive a 15.5 percent increase in revenue per visitor. However, it is not clear in what time period (in days or months) they supposedly achieved these results.

They also claim Einstein Activity Capture was able to eliminate an hour a day of manual data entry for sales representatives at Silverline.

While the market price and demand for the CRM AI technology are currently high, Salesforce CEO Marc Benioff no longer sees major acquisitions as likely in the short term. Instead, he is working on improving their offering via partnerships. Their most high profile AI partnership is with IBM.

In March, Salesforce announced it was teaming up with IBM to integrate Watson’s data and tools into their CRM system, giving their clients access to Watson’s existing information sources and its ability to analyze their data. For example, integrating IBM weather predictions into their market strategies. Just one small example of how this partnership works is through the use weather forecasting, which IBM has heavily invested in become a world leader in. With real time access to Watson’s weather data, companies using Salesforce can directly warn their customers who might be negatively impacted by changing weather conditions.

For example, Watson might determine it is about to hail in Houston. Car insurance companies can then use Salesforce to quickly alert all their customers in the area about to be hit to take precautions.

Salesforce is also working to add new capabilities to Einstein. Two functions currently in beta are Einstein Intent and Einstein Sentiment. They use natural language processing to classify whether the text of a message is emotionally positive or negative and determine what the intent of the message is, respectively.

For example, a company could put to use Einstein Sentiment to classify the tone of their inbound customer emails and accordingly identify positive brand evangelists and escalate dissatisfied customer responses into service cases. Einstein Intent, on the other hand, can be used to augment Einstein Sentiment. It could classify each negative response and identify the source of customer dissatisfaction, like lost shipments and returned orders.

Another function in pilot is Einstein Object Detection. It is designed not only to detect and classify objects but also to identify different objects in an image and their location. Like the below video demonstrates, a company could use a picture of a shelf to quickly know how much inventory is on display or customer could take a picture of something they want and the program could tell them where to buy it.

Salesforce’s documentation page for Einstein Vision states that this application can be used within CRM to help customers find products, to help customer service agents identify a product related to a customer complaint, and more. We were unable to find any current use-cases of this technology with Salesforce’s customers.

Read More

CRM Artificial Intelligence Trends across Oracle, SAP

Oracle 

Oracle announced the launch of their Intelligent Cloud Applications the exact same day Salesforce announced Einstein. At the time, Oracle’s project leader told they were “trying to avoid the hype and build apps that people can buy, use and make money with.”  Instead of creating a single, all-encompassing new “AI brand,” Oracle has been focused on specific AI and machine learning applications in their cloud services — ready-to-use apps that can be quickly and easily tailored to a specific uses.

Earlier this year, Oracle introduced several new AI powered functions to their customer experience cloud (CX Cloud Suite). According to the company, these include:

1. AI-powered personalized marketing/experience – personalizing the content each customer receives.
2. Predictive recommendations – using a customer’s data to recommend products they would be most interested it.
3. Optimizing the selling process for representatives – opportunity analysis of clients to create guidance to help close deals.
4. Chatbots

Since then Oracle claims to be adding more AI powered functions to its cloud.

Last year, Oracle bought Crosswise, a data company that offers machine-learning based cross-device data, which could be used in cross-device advertising, personalization and analytics. Oracle appears to have bought this technology to augment its data cloud platform.

Oracle claims their big advantage in AI is their unique access to enterprise data. The company claims to have “5 billion global consumer and business IDs, with more than 7.5 trillion data points gathered monthly.” These data points include historical and dynamic customer data such as click-stream and social activities, and inputs such as weather, lookalike audiences, etc. Such customer information can be mined to find customer behavior patterns. The long term vision is being able to have AI that can provide more seamless and natural direct customer service across platforms.

Oracle President of Development Thomas Kurian said at the recent Oracle Openworld conference that there vision of the future is, “No longer is it just web and mobile screens, but you could speak to the application. You can interact with it with messaging. You can take pictures, and we can identify images, compare them with other things, and automate transactions.”

Our research seems to indicate the these claims about the use of AI in Oracle’s CRM are aspirational, there don’t seem to be any presently developed products for machine vision at the company. While this may merely be posturing by Kurian, we suspect that Oracle will indeed be developing AI initiative actively in its CRM offerings. How it will compete with Salesforce in this regard – only time will tell.

SAP

German software giant SAP has the goal – according to Global VP Volker G. Hildebrand – “to build machine learning technology into all our software, across every line of business and industry we serve.” This includes SAP Hybris, their main cloud CRM service.

In July the company relaunched SAP Leonardo, their integration platform to enable companies to more easily integrate AI and machine learning into their business. SAP Leonardo appears to integrate the various SAP product offerings for machine learning, IoT, big data, analytics, etc.

Similar to Salesforce (and to the claims made by Oracle), SAP is also developing machine vision applications for it’s Hybris CRM product.

The company also announced machine learning facial recognition capacity for SAP Hybris, a platform that “sells omni-channel customer engagement and commerce solutions.” SAP claims the system will allow stores to automatically determine the age and gender of customers entering their locations, and make personalized product recommendations based on the stock information.

Among the products SAP is currently developing is Charly the Chatbot. This chatbot is a customer-facing, “conversational commerce app.” It is a prototype designed to function as a digital assistant for individual consumers to help them find products, make recommendations, request refunds, etc.This prototype, according to SAP, is an “experiment on microservice-based business models.”

How SAP plans to use this in the B2B CRM context is not very clear. At present, Charly seems to be limited to Facebook Messenger.

SAP is also experimenting with using their technology on Pepper Instore Assistance, a robot that is designed to act as a personal, in-store customer service rep. Both Charly the Chatbot and Pepper Instore Assistance are currently prototypes. We aren’t certain of when these initiatives will become regular offerings for SAP, and it’s unclear exactly how they’ll integrate with Hybris.

Read More

Best Examples Of Using Artificial Intelligence For Retail Experiences

Artificial intelligence has the potential to completely transform the traditional retail experience and take it to the next level with personalization, automation and increased efficiency. And it’s already happening! Here are 20 of the best examples of AI to improve the retail experience.

1. Lowes Uses Robots To Locate Items

Navigating a hardware store can be difficult, but Lowes created the LoweBot to help customers find their way around the store and get the items they need. LoweBots roam the store and ask customers simple questions to find out what they’re looking for. The robots can provide directions and maps to products and share specialty knowledge with customers. LoweBots also monitor inventory so the store knows what items need to be restocked.

2. Walgreens Uses AI To Track Flu Spread

The flu can be uncomfortable, inconvenient and even deadly if not treated properly. With the right information, people can take action to keep their families healthy. Walgreens uses data from the number of anti-viral prescriptions it fills at more than 8,000 locations to track the spread of the flu. The online, interactive map not only helps customers know how bad the flu is in their area, but also helps Walgreens stock more inventory of flu-related products in infected regions. AI is empowering customers and the store.


3. Sephora Makes It Easy to Find Makeup

Step into a Sephora store to find your perfect makeup shade without ever putting anything on your face. Color IQ scans a customer’s face and provides personalized recommendations for foundation and concealer shades, while Lip IQ does the same to help find the perfect shade of lipstick. It’s a huge help to customers who know the stress (and cost) of finding the perfect shade by trial and error.

4. North Face Helps Customers Find The Perfect Coat

Don’t know what coat to buy? North Face can help with that. The company uses IBM Watson’s cognitive computing technology to ask questions about where they’ll wear the coat and what they’ll be doing. Using that information, North Face can make personalized recommendations to help customers find the perfect coat for their activities.

5. Neiman Marcus Uses AI For Visual Search

Luxury department store Neiman Marcus uses AI to make it easier for customers to find items. The Snap. Find. Shop. app allows users to take pictures of items they see while out and about and then searches Neiman Marcus inventory to find the same or a similar item. Instead of using vague search terms to find an item, the photos can usually find a very similar match.

6. Taco Bell Helps Customers Order Tacos On The Go

When you want tacos, there isn’t time to wait. Taco Bell was the first restaurant to allow customers to order food directly through AI. The Tacobot works with Slack and makes it easy for customers to text or say their order. The bot even allows for customized and large group orders. In true Taco Bell form, the bot responds with quippy remarks after each order.

7. Macy’s Adds AI To In-Store Experience

Have you ever walked into a department store and had no idea where to find what you’re looking for? Macy’s On Call app is tailored with answers for each individual store. Customers open the app when they’re in the store and can chat with an AI bot to get directions to a specific item or check if something is in stock. The bot can even detect if a customer is getting frustrated and alert a human employee to go help the customer.

8. Walmart Deploys Robots To Scan Shelves

Walmart is one of the largest retail stores in the world, and it plans on using robots to help patrol those vast aisles. Walmart is testing shelf-scanning robots in dozens of its stores. The robots scan shelves for missing items, things that need to be restocked or price tags that need to be changed. The robots free human employees to spend more time with customers and ensure that customers aren’t faced with empty shelves.

9. ThredUp Uses AI To Remember Customer Preferences

Online consignment store ThredUp recently released Goody Boxes, which include a number of secondhand clothing items tailored to match each customer’s style. Customers keep and pay for the items they like and return the items they don’t like. An AI algorithm remembers each customer’s preferences so that future boxes can better fit their style. The non-subscription boxes are easier for customers than searching for individual pieces.

10. Amazon Eliminates Cashiers With AI

No list on retail AI would be complete with Amazon and its revolutionary Amazon Go store. Customers can simply walk into the store, take what they want from the shelves and walk out without going through a cashier. Sensors and cameras throughout the store track what customers purchase, and their Amazon account is charged when they leave. AI helps create a quick and seamless shopping experience so customers aren’t stuck waiting in line.

11. Uniqlo Can Read Minds With AI

Clothing store Uniqlo is pioneering the use of science and AI to create a unique in-store experience. Select stores have AI-powered UMood kiosks that show customers a variety of products and measures their reaction to the color and style through neurotransmitters. Based on each person’s reactions, the kiosk then recommends products. Customers don’t even have to push a button; their brain signals are enough for the system to know how they feel about each item.

12. West Elm Connects Style And Products

Customers used to bring in physical style boards of furniture items they liked, but now furniture retailer West Elm does it through AI. The Pinterest Style Finder scans a customer’s Pinterest boards to understand their personal style and create a list of recommended home décor and furniture items to match. It’s an easy way for customers to get a beautifully designed home that reflects their style.

13. Sam’s Club Makes Warehouse Shopping Simple

The warehouse superstore recently opened a smaller, AI-powered version of its store called Sam’s Club Now. The store uses AI technology so that customers can shop without having to go through a traditional checkout line. The corresponding app can even map the most efficient route through the store to get everything on a customer’s shopping list.

14. Olay Uses AI To Personalize Skincare

With the help of AI, Olay customers can get personalized skincare treatment without having to see a dermatologist. With Olay’s Skin Advisor, customers take a selfie of their plain face, and the app uses AI to tell the true age of their skin. The app evaluates skin health and makes recommendations for problem areas with personalize skin care regimen recommendations.

15. Kroger App Customizes Product Recommendations

Grocery chain Kroger is testing out the idea of smart shelves. When a customer walks down an aisle and has their Kroger app open, sensors identify the shopper and highlight products they might be interested in. The app might highlight gluten-free products for a gluten-free shopper or kid-friendly snacks for a parent. The app can also provide personal pricing and alert shoppers if an item on their shopping list is on sale.

16. H&M Uses AI To Keep Popular Items Stocked

Popular clothing store H&M relies on staying on top of trends to be successful. The store uses AI to analyze store receipts and returns to evaluate purchases at each store. The algorithm helps the store know what items to promote and stock more of in certain locations. The data could find that floral skirts sell well at urban stores and change the inventory to match what customers want.

17. Zara Streamlines Order Pickup With Robots

Buying online and picking up in store is a popular option for customers. Fashion retailer Zara recently started using robots to help customers when they come to pick up their orders. When they get into the store, customers enter a pickup code that starts the robot moving in the warehouse. Once the order has been found, the robot delivers it via drop box. It’s a quick and efficient way to get orders to customers.

18. Starbucks Bot Makes It Easy To Order Coffee

The coffee giant makes it even easier to get your morning pick-me-up with its AI-enabled voice ordering. Customers can chat with the My Starbucks Barista app to place their order with voice or text. When the customer gets to their local Starbucks, the order will be waiting and they can skip the line.

19. American Eagle Creates Fitting Rooms Of The Future

There’s nothing worse than getting to the fitting room, only to discover you grabbed the wrong size or want to try another color. Instead of getting dressed and leaving the room or flagging down an employee, American Eagle is creating interactive dressing rooms. Customers simply scan the items they want and can see what’s in stock. Store employees are notified if the customer needs an item delivered to the fitting room. The technology can even make product recommendations based on what the customer has tried on.

20. Rebecca Minkoff Designs AI-Powered Smart Store

Clothing designer Rebecca Minkoff has three stores around the country and is one of the first brands to create “connected stores.” The stores use AI to run touchscreen smart mirrors, which allow customers to browse through clothing items and inspiration. Shoppers can then try them on in an interactive fitting room with custom lighting options. The fitting room mirrors use RFID technology to automatically know what customers are trying on and tell them what other colors and sizes are available.

Read More

Examples of how AI integrated in the retail industry

Innovation is a double-edged sword, and as with any innovation results are a mixed bag. While many AI applications have yielded increased ROI—this case study of AI in retail marketing segmentation is one example—others have been tried and failed to meet expectations, shining a light on barriers that still need to be overcome before such innovations become industry drivers.

Below are a few brief use cases across five retail domains or phases. Each provides a fraction of a glimpse as to how AI technologies are being used today and which are being created and piloted as potential retail industry standards in eCommerce and brick-and-mortar operations.

1. Sales and CRM Applications

a. Pepper Robot

In 2010, Japan’s SoftBank telecom operations partnered with French robotic manufacturer Aldebaran to develop Pepper, a humanoid robot that can interact with customers and “perceive human emotions.” Pepper is already popular in Japan, where it’s used as a customer service greeter and representative in 140 SoftBank mobile stores. According to Softbanks Robotics America, a pilot of the Pepper in California’s b88ta stores in both Palo Alto and Santa Monica yielded a 70% increase in foot traffic in Palo Alto, and 50% of Neo-pen sales in Santa Monica were attributed to Pepper.

Additionally, the AI creation spent time at hip apparel store the Ave, where the retailer experienced a 98% increase in customer interactions, a 20% increase in foot traffic and a 300% increase in revenue. Nestle announced in January 2016 that it planned to acquired Pepper robots to put in 1,000 of its Nescafes in Japan.


While not the only retail robot in use, in-store robots like Pepper appear to at least initially boost store interest and sales. Whether this is a novel effect that will wear off once retail robots become “the norm” remains to be determined.

b. Conversica

Conversica “sales assistant” software is designed to automate and enhance sales operations processes by identifying and conversing with internet leads. The sales lead and management company claims the authentic-sounding messages result in an average engagement rate of 35%.

In one case study, Star2Star Communications implemented its Conversica-powered sales rep “Rachel” in 2016 and saw a 30% email response rate within hours. The customizable sales assistant software is also used to cross-sell or re-engage existing leads. New England-based Boch Automotive also employed Conversica software, which it attributed to a an average 60-sale increase per month at one Toyota dealership. 

2. Customer Recommendations

a. IBM Watson Cognitive Computing
It’s no longer a secret that IBM’s Watson is providing a slew of order management and customer engagement capabilities to eCommerce retailers. In 2016, 1-800-Flowers.com launched Gifts When You Need (GWYN), which the company calls an AI gift concierge.

Through information provided by consumers about a gift recipient, the software tailors gift recommendations by comparing specifics provided to gifts purchased for similar recipients. The GWYN experience (which we included in our more complete article on chat bot use cases) attempts to replicate the role of a concierge at a store through a personal and detailed conversation with users. 1-800-Flowers Chris McCann spoke to Digiday, noting that within two months, 70% of online orders were completed through GWYN. 

North Face has also adopted IBM Watson’s cognitive computing technology to help consumers determine what jacket is best for them, based on variables like location and gender preference. For example, hiking in Iceland in October and commuting in Toronto in January will yield different results.

Published 2015 pilot results, based on data collected from 55,000 users, resulted in a 60% click-through rate and 75% total sales conversions. It’s important to note that we weren’t able to tell if these results represented more or less than North Face’s normal results, and whether these results are sustainable or are merely driven by initial novelty in the user interface.

Above is an example of the North Face’s conversational interface, which prompts users with a series of questions related to their purchase. It’s safe to say that similar systems like the one above could be built with simple if-then rules, and no machine learning whatsoever.


The advantage of using machine learning in such a recommendation Q-and-A interface, is that North Face can potentially run tens of thousands of consumers through this conversational engine. At a certain volume of customer interactions, the system might be expected to glean important insights and patterns on suggestions that “work” (high take-rate to purchase, or high cart value), and those that don’t – allowing the company potentially gain higher and higher conversions over time.

3. Manufacturing

a. Brilliant Manufacturing

General Electric’s (GE) Brilliant Manufacturing software, in part inspired by GE’s relationships with client manufacturing companies over the past two decades, was designed to make the entire manufacturing process—from design to distribution and services—more efficient and hence save big costs over time. The software includes a suite of analytics and operational intelligence tools appropriate for a range of manufacturers.

The WIP Manager software, for example, provides industrial and discrete manufacturers with plant-floor and plant-wide collaborative visibility of all work in process. An operational supervisor sitting behind a computer can now identify in real-time a floor-based problem that arises in workflow, rather than spending time making time-consuming walkthroughs of entire manufacturing facilities.


Toray Plastics is one example of a company that is using GE’s Plant Applications product, which allows management to collect granular-level data throughout production and reduce defective products and wasted productivity.

b. Gakushu Learning Software

Fanuc’s Gakushu Learning Software  (“Gakushu” means “learning” in Japanese) embedded in manufacturing robots speeds up “smart” operations on a specific task, originally designed for spot-welding and assembly lines. In 2016, Fanuc partnered with Nvidia, with the goal of accelerating deep learning in robots via Nvidia’s GPUs. Gakushu-endowed robots learn a manufacturing task through use of a sensor that collects and stores data.

The robots’ ability to adjust to real-time environmental conditions and adjust motion can (according to Fanuc) result in up to 15 percent cycle-time improvements in spot welding.  Once the robot’s learning process is complete (about 18 cycles later), the sensor is removed and the trained robots are then able to complete a task autonomously.

The robots’ learning is paired with vibration control in the form of an accelerometer that learns the robot’s motion and provides increased movement stability. Tesla has employed about 600 Fanuc robots at its factory in Fremont, and according to Bloomberg Businessweek put in a significant order for more robots back in September 2016 in an effort to speed manufacturing efforts for the next slated delivery of its Model 3 in July 2017.

4. Logistics and Delivery

a. Domino’s Robotic Unit (DRU)

In addition to Domino’s claims that its prototype delivery robot can keep food and drinks at the appropriate temperature, the DRU’s sensors help it navigate a best travel path for delivery. DRU integrates robotics technology previously used for military combat training. In March 2016, DRU pilots were rolled out in Australia, New Zealand, Belgium, France, the Netherlands, Japan and Germany. Domino’s doesn’t provide dates as to when the DRU might be rolled out on a commercial scale, but it seems plausible that robotic delivery of food—and other goods—could be a burgeoning reality within the next decade.

b. Amazon Drones

In July 2016, Amazon announced its partnership with the UK government in making small parcel delivery via drones a reality.  The company is working with aviation agencies around the world to figure out how to implement its technology within the regulations set forth by said agencies. Amazon’s “Prime Air” is described as a future delivery system for safely transporting and delivering up to 5-pound packages in less than 30 minutes.

The first 13-minute unmanned flight by Amazon took place on December 2016, as seen in the video below. At present, determining proper safety and reliability of operations and systems is Amazon’s top priority. Amazon notes that it’s working with regulators in “various countries”, though there haven’t been any updates on forecasted dates for commercial use. Similar to Domino’s DRU concept, it seems possible that autonomous delivery of goods and food by air could be rolled out at scale within the next decade. 

5. Payment Services

a. Amazon Go

Amazon’s touted brick-and-mortar locations, known as Amazon Go, employ check-out-free technology that allow customers to shop and leave  Customers use the Amazon Go app to check in, but thereafter the entire shopping experience is designed to be automated. Sensors track which objects customers pick up and put in their basket, and customers’ Amazon accounts are automatically charged after exiting the store.

The intended launch hasn’t been without its barriers, and at the end of March 2017 sources close to the retail giant announced that Amazon was delaying the opening of its convenience stores while it worked out “technology glitches” in the automated shopping and purchasing process. 

b. PayPal

Since 2013, PayPal has leveraged fraud detection algorithms to protect customer’s digital transactions. Over the last few years, thousands of purchase patterns or “features” have been learned by the security detection system, which can now (in an example provided by MIT Tech Review) decipher between friends who are buying concert tickets simultaneously and a thief making the same kind of purchases with a list of stolen accounts.

A referenced study by LexisNexis found PayPal’s deep learning approach to transaction security reduced fraud rate to 0.32% of revenue, which is 1% less than the average rate seen my most eCommerce merchants.

c. Payment Fraud

Fraud and payment security are a massive area of AI investment, and there are plenty of fraud / security companies worth looking at. Sift Science is one of many companies applying machine learning to detecting user and payment fraud – both of which are relevant for retail applications. This will become more so as US eCommerce continues to swell as a percent of retail sales (and that growth is happening consistently, according to the US Census Bureau).

6. AI in Retail – Concluding Remarks

A lot of retail-focused AI vendor companies, says that “big box” retailers (Best Buy, Target, Walmart, etc) are extremely slow to adopt cutting-edge technologies. Because it’s mostly large companies that have the budgets and data volume required to make the most of many of today’s best AI technologies, we outright surmise that an “AI revolution” in the retail space is unlikely. It may be another three to five years before most large retailers have substantial, business-critical AI applications in manufacturing, supply chain logistics, or customer service. 

Applications that have the highest likelihood of broader retail adoption are those that have a direct, hard-line return on investment. “Improving customer engagement”—even with case studies and examples—is a softer benefit than “reducing lost packages by 6-10%.”  Our retail executive guests using AI expect that the relatively stodgy committees in these large companies is likely to be extremely critical, safe, and bottom-line focused.

As with many areas of AI innovation that are led by bigger industry players, the future will likely be dictated by the retail AI use-cases that are proven effective by leading industry players. It’s safe to say that every at-scale retailer in the world is looking to Amazon for hints on “next steps,” and we can expect that the broad swath of relatively smaller retailers will be looking at Amazon, Walmart, Best Buy, and others for their own ideas on strategy.

Read More

Will AI change the game for retailers?

Over the last few years, artificial intelligence technology has made some interesting advancements across multiple industries. While it may not be so obvious to the end consumers, artificial intelligence has been applied in the retail sector as well. Even though not every retailer has been using it due to high costs, inaccessibility, and proprietary systems, the largest players in the retailing have been pretty active about it. It's no wonder given the benefits AI can bring to the actual businesses.


How Artificial Intelligence (AI) Can Help?

So how exactly can artificial intelligence help retail store owners? First off, let's say a few words about how AI and machine learning work. Namely, artificial intelligence technology takes a big data set about something, runs it through AI algorithms such as neural networks and then produces a model which can provide answers like a real human. The answers that are given are based on whatever AI was able to learn about the matter from the dataset.

As it's probably obvious at this point, the dataset that AI learns from in the retail industry example is the actual sales data linked to the customer data. When this information is run through the machine learning algorithms, an AI model is produced which discovers actionable information about a business, customers, and inventory which are not normally obvious or known to the business owner. With this in hands, a retailer can do a variety of things to benefit his own business.

For example, retail AI can learn about customers, their preferences, and their behavior to get to know them. And, in recent years, it can get to know them so well that it knows what they need and when they need it. And it knows this before even they do. When AI is capable of producing such information on the fly, numerous benefits can be obtained immediately for the business cash flow and the general experience. First one is to get the customer to buy more. If AI knows what they need before they do and issues them a coupon for a given product, this will trigger them to buy an additional item which they didn't come for while generating extra revenue for the business.

Another benefit would be improving your customer's experience because if you offer them exactly what they need at a better price, you start building a personal selling relationship with your customers.

Product Layout

Another area where retail AI can help greatly is how items are physically laid out in the store. We've all heard stories about how big retail chains order items to get us to buy more. And we probably all think there is some magic formula by which we can display our inventory and get our customers to buy more. However, there is no such formula. Even the same store with same items but on different location can have completely different triggers for its customer base. So there is no one-size-fits-all solution here.

However, what is available is the artificial intelligence technology which can help with this.

When provided with the sales data, machine learning can discover patterns in your customer's buying preferences and learn what they buy together. Based on this, AI can provide you with suggestions on what items to put next to each other in your store to get your customers to purchase more. And since those suggestions are based on real data provided by real customers that are buying in the particular store, they are certainly guaranteed to deliver results and trigger your customers.

There's a lot more AI can do for retail businesses. The examples listed above are just the tip of the iceberg. Whatever part of your business, you can provide data for; AI can extract really valuable insights which are to be treated like gold. It can be used to improve your inventory turnover, optimize your stock, predict future revenue, and a lot more.

There is one catch. Namely, with the things as they are, Artificial Intelligence systems are mainly proprietary and available only to big retail chains having big technology budgets. It will be interesting to watch this space over the next few years to see solutions coming out, which will be crafted more for small retailers than the big ones. Some companies have already started this development so it won't be long for the independent retailer to have access to the same tools as the national brands.

Artificial intelligence in retail is being applied in new ways across the entire product and service cycle—from assembly to post-sale customer service interactions, but retail players need answers to important questions:

Which AI applications are playing a role in automation or augmentation of the retail process? How are retail companies using these technologies to stay ahead of their competitors today, and what innovations are being pioneered as potential retail game-changers over the next decade?

In the next post we'll cover a variety of examples in which AI is being integrated in the retail industry, broken down into the following sub-categories:

• Sales and CRM Applications
• Customer Recommendations
• Manufacturing
• Logistics and Delivery
• Payments and Payment Services

Read More

AI in retail

With consumers expecting premium levels of service at the press of a button, it’s vital that businesses in the retail industry are at the cutting edge of technology. Expectations are higher, and competition is fiercer than ever, and as such, innovation and industry disruptions are more necessity than a luxury.

For decades, consumers have relied on friends and family, product reviews and tastemakers when making purchasing decisions. A loved one could recommend a particular brand of tools that’s worked well over the years. A consumer watchdog publication could inform and educate on which car models offer the most reliability. A fashion magazine could highlight the latest trends that speak to any style. But while each of these different influencers remains relevant to today’s buyers, shoppers seeking out buying advice are increasingly being guided by artificial intelligence.

Through sophisticated AI, retailers are diving deeper into personalization by building solutions that suggest the best products for a user to purchase bolstered by data-driven insights.

Thanks to powerful AI-driven supply chain management, retailers can easily track what’s in store, what’s being shipped and what’s in the warehouse; ensuring customers can get what they want when they want it. But to create a more personalized shopping experience, retailers are also putting together better product collections, embracing trends like “showrooming” and crafting entirely new ways of shopping.

Here’s a look at how AI-driven personalization is transforming brick-and-mortar retailing:

1. b8ta

Retailers long maligned the trend of “showrooming” — that is, trying out a product in-store only to make an eventual purchase online. AI-driven supply chain management has allowed omnichannel retailing to alleviate some of these fears, but new retailers like b8ta have embraced this trend even further by building new stores around the showrooming concept. Offering retail-as-a-service, b8ta is an open-concept store that offers companies a flexible way of selling through brick-and-mortar locations. Companies can showcase products in b8ta stores from online brands that desire a physical presence.

For consumers who wish to purchase something online but also want to see it in person, b8ta changes the game. And for online retailers with a wide range of SKUs or a limited desire to expand into physical retail, b8ta offers the best of both worlds by showcasing products for limited amounts of time. Combined with AI-gathered data for personalized product targeting, a manufacturer could take advantage of b8ta by offering a small sample of their most popular products that customers wish they could try out in real life.

2. Amazon 4-Star

If you’re buying a kitchen gadget, how do you know it’s the best kitchen gadget? Well, if you’re Amazon, you know it’s the best because it’s got a wealth of customer purchasing data behind it. This includes star ratings, which lets users rank products after they’ve been purchased.

That’s the core concept behind Amazon’s newest retail store in New York City: Amazon 4-star. Carrying a curated selection of products that have all received large amounts of four-star ratings, Amazon uses its sophisticated product recommendation engines to bring its bestselling, most popular items into physical stores.

By offering a hand-picked selection of products that are beloved, trending or hidden gems, the service allows customers to shop from a collection of highly personalized recommendations in a brick-and-mortar setting.

Considering 35 percent of Amazon’s revenue comes from its AI-enhanced product recommendations, it’s a profitable shortcut to give customers what they already want.

Selling only the top-rated products might also be the right approach for adjusting an existing retail strategy. In early 2018, home furnishing retailer Crate & Barrel shuttered all physical locations of its children’s furniture chain The Land of Nod and began offering a smaller, curated collection of the same products under its in-store label, Crate & Kids. For Crate & Barrel, it became clear that offering a more personalized selection of products to its customers was more valuable than propping up an underperforming retailer that featured wider selections.

3. AlgoFace

Buying new makeup can be a long and messy process. Dropping by Sephora or the makeup counter at Macy’s means waiting for an associate to help you apply lipstick or eyeshadow to find the perfect color — from a tube that’s already been used by somebody else.

AlgoFace is making this process simpler (and far more sanitary) through its virtual-makeup SDK, which is available for makeup retailers to build into their apps. Shoppers can virtually apply an endless array of makeup shades to a live video of their face. Their AI-driven augmented reality interface makes it look like users are actually, physically wearing the makeup they’re thinking about buying.

The result isn’t just a highly personalized experience that lets users try out makeup combinations with no mess: It’s an incredible way to cut down on costs by saving on makeup samples. As for customer experience, this means being able to try out different looks in a mobile app or at a physical location.

How AI can transform the retail customer experience

It then comes as no surprise that retail businesses are projected to spend $7.3 billion on AI by 2022 — with the majority of that money dedicated to transforming the customer experience.

Through the strategic and nuanced implementation of AI, retailers will offer customers a more personalized, intuitive, and convenient experience, leading to enhanced levels of brand trust.

Such a dramatic technological shift in the industry is evidenced by leading retailers — such as Amazon, Dominos Pizza, and Walmart — making AI a significant facet of their business model.

1.Improved Delivery Logistics

Since Amazon is a leader in almost every aspect of retail and eCommerce, it’s no surprise that it’s breaking some serious ground with AI.

Amazon’s thirst for AI innovation is no more evident than when in July 2016, it partnered with the UK government in developing drones to make small parcel deliveries. Currently, the eCommerce giant is working with regulatory agencies towards implementing the technology.

Known as ‘Prime Air,’ Amazon’s drone delivery service is a future delivery system dedicated to the safe transportation and delivery of packages up to 5 pounds in less than 30 minutes.

In December 2016, the first 13-minute unmanned flight of a drone was launched. At this point, Amazon steadfastly aims to determine optimal safety and reliability parameters ASAP. While there are no projections for commercial use, there’s plenty of collaboration with regulators in a multitude of countries.

If all goes according to plan, it seems entirely plausible that within the next decade, instantaneous AI air deliveries to customers will be a reality.

2. Personalization

Another way the retail industry is leveraging AI is its use of chatbots — which offer customers the convenience of their own personal shopper at no additional fee.

For instance, Walmart’s “Hoo the Owl” asks questions to future parents registered online. The chatbot finds out the baby’s due date, gender, and nursery theme then customizes suggestions depending on the answers. Walmart intends to enhance the levels of personalization by adding more questions about organic and sustainable products.

The shopping experience is brought directly to Walmart’s customers with its savvy implementation of AI technologies. Instead of customers aimlessly searching for ideal products, Walmart’s chatbots ensures that the best choices fall directly into their laps.

Cosmetic industry juggernaut Sephora applies the same principles with chatbots, offering each customer ideas on new makeup looks. Sephora’s chatbot is an in-store tool too, providing product recommendations, reviews, and ratings during a visit to any one of their stores.

3. Improved Customer Communication

Beyond personalization, AI also lets customers contact businesses no matter the place or time. Without AI, there is a litany of factors preventing customers from getting in touch with reps. In most cases, the phone lines are too busy, or it’s no longer operating hours.

While chatbots can’t adequately respond to every customer query and complaint, they filter through the simple issues, allowing the human agents to focus their energies on more challenging circumstances.

What’s more, is these bots are highly adaptable, handling increasingly more difficult complaints. Then, once a situation exceeds a chatbot’s capabilities, it directs the issue to a live rep.

Immediate access to these chatbots increases both customer satisfaction and productivity while decreasing the operational costs of customer service.

Domino’s Dom the Pizza Bot provides immediate responses to customers with hilarious and engaging content. Dom also gives customers the choice of fully automated ordering experience, or they can talk to a live agent.

It’s this kind of savvy thinking that led to Dominos knocking Pizza Hut off its throne of ‘largest pizza company in the world’ at the beginning of 2018, after generating an astonishing $12.3 billion in global revenue.

It behooves retail companies to see that the future is now. The quicker they start utilizing AI, the more they’ll remain at the head of the pack.

Nearly all the market segments, industries, and business domains are actively interested in AI and are looking to harness the technology for business benefits. AI offers companies the power to reduce costs and make the shopping experience much more delightful and efficient for the end consumers. Naturally, how can the retail segment stay away from the promising technology?

The Retail-AI Syndicate

According to Global Market Insights, investments in AI by retail segment will exceed USD 8 billion by 2024. As more applications for machine learning, predictive analytics and deep learning technologies are experimented with success, digital disruption in retail segment is bound to happen at a much rapid pace.

Globally, AI experts believe that the technology offers a wide range of applications for the retail segment. In the coming year, retail will see greater infusion of AI-based solutions in day-to-day operations. AI will definitely change the customer service cycle in retail and both retailers as well as the consumers are bound to benefit from the AI-Retail syndicate.

The AI Impact

Artificial Intelligence can serve as a boon for retail companies that gather and possess customer data. AI can derive meaningful conclusions from massive amount of data and help companies to create personalized shopping experiences via highly-structured web shops, intelligent in-store bots and online chatbots. Let us see how AI will transform the retail experience for the shoppers in 2019.

• Digital Racks for Apparel & Fashion Products

With the help of AI technology, apparel retail brands can create virtual racks and trial rooms with gesture walls and touch-free monitors to find the right style, without having to shuffle through a pile. Customers can instantly see how a dress would look on them and can browse through recommendations based on their preferences and style quotient.

All this will not only enhance the shopping experience but will help customers to choose from millions of options, that is currently not possible due to space constraints inside a physical store. With AI, a store can be converted into a bustling repository of design and draping ideas that users can choose from. Also, stores will be able to gather deeper insights into consumer behaviour, optimising their product portfolio for much better retail experience delivery.

• Virtual Trial Rooms for Swift Decision Making

While shopping for a new dress or an apparel, trying out different options can get quite frustrating and time consuming. With the help of virtual trial rooms equipped with digital mirrors, customers can actually try dresses without having to change again and again.

Using a gesture and touch-based interface, a shopper can mix and match various outfits, accessories, shoes, etc. to finalise the perfect look, in no time. Not only for apparel brands, virtual mirrors can make it easier for cosmetic companies and beauty brands to portray how a lipstick shade or foundation would appear on real skin without forcing customers to apply the product.

• Robotic/digital assistance will be a reality

Using AI, smart analytics and natural language processing technology, retail stores can give customers the power to get instant support, inside the stores. By placing robots and touch panels, stores can help customers locate an item, get answers to their queries and find out how a product can make their life easier.

With the help of customer service bots powered by AI, stores can reduce their manpower costs and provide 24x7 assistance to the customers. Not only will it improve customer service levels but will definitely attract more buyers to the store in 2019.

• Behavioural Analytics Powered by AI-backed Surveillance

Using modern surveillance equipment powered by AI and computer vision technologies, retailers can capture and study customer behaviour inside stores. This will help them understand engagement levels with current store layout and optimise operations for higher engagement and revenues.

Video analytics can also improve in-store security and reduce the chances of theft. With the help of AI, surveillance footage can be monitored in real time and alerts can be sent to administrators, store owners for immediate action.

• Better Customer Support via Chatbots

AI-powered chatbots are enabling retail brands to engage customers, efficiently. With the help of chatbots, brands can handle thousands of queries simultaneously, without having to employ a large workforce.

Chatbots can be configured to answer questions, provide shopping suggestions and provide prompt support. Using a mixture of AI chatbots and humans, brands will be able to effectively handle customers and provide users with a better resolution of their problem. When customers will get prompt and personalized attention, they will engage deeply with the brand, paving way for enhanced customer loyalty.

All these scenarios are not just a fantasy but will be experienced by retail shoppers in the coming years. Artificial Intelligence will rule the retail domain and help brands become more customer-centric and efficient in retail operations.

Read More