Integrating AWS Artificial Intelligence (AI) Services to ServiceNow Enterprise Apps and Workflows

[Anand Trivedi]

Artificial intelligence (AI) helps enterprises to enhance customer experience, achieve operational excellence, automate workflows, reduce time-consuming manual efforts and improve predictive analytics. Many organizations, business functions, and developers think of embedding AI in existing enterprise apps and workflows. However, the expertise required to manage a huge scale of contextual data, training Machine Learning(ML) model and orchestrating infrastructure for AI limit the eventual growth. 

Amazon Web Services (AWS) offers a vast and growing array of AI services, including hundreds of free or paid ML algorithms and model packages across a broad range of categories such as Audio, Computer Vision, Image, Natural Language Processing, Speech Recognition, Structured, Text, and Video. Hence, organizations are turning to AWS to accelerate innovation and bring down IT costs. Integration between ServiceNow® and AWS can enable Machine Learning enabled workflow solutions, thus completing the missing piece in the puzzle for a complete enterprise automated application. 

In this post, you will learn :

• how to create ServiceNow based Enterprise Apps or workflows using optimized ML algorithms offered in Amazon SageMaker, or
• how to bring-your-own ML algorithms and models to use from Amazon ML to Servicenow.

Below are some sample use cases where ServiceNow enterprise workflows or Enterprise Apps can leverage Amazon AI services:

Area of work	Use cases for Embedding Artificial Intelligence to improve productivity
IT Workflow	Automate IT Ticket categorization using the ticket description text and other information analytics Predict ITSM Ticket completion time using historical records predict ITSM Ticket time for the competition. Automatically assigns new incidents to the right teams based on past records or availability Adaptive security and response against known threats Enforce regulatory compliances on all IT Ops
Employee Workflow	Import documents, interactions, and inquiries as employee records or requests using NLP and speech recognition services Build intelligent chatbots to resolve common issues Automate employee onboarding events
Customer Workflow	Shorten customer service response times with machine learning-powered case categorization, prioritization, and assignment.

 

Addressing challenges of Integrating AI with ServiceNow Apps

Amazon provides Machine Learning as a Service (MLaaS) that takes away all the headache of managing storage devices, managing machines for training and then deploying and testing newly trained models.

Let’s understand the challenges associated with machine learning and how the Amazon cloud brings a solution to them.

Challenge	How Amazon Solves your Problem
Data labeling on ServiceNow data requires huge and time-consuming manual effort	Amazon Provides tools and manpower (paid) for faster data labeling with less effort at some additional costs.
Preparing/deploying ML algorithms requires data scientists or similar skillsets	Amazon provides free and paid ML algorithms that are pre-trained and currently being used by AWS, hence you have access to the world’s best ML algorithms.
Training algorithms outside ServiceNow requires very heavy GPU and TPU machines	AWS offers bundled machines with an ML algorithm to solve this problem at a very low cost.
Preparing graphs and metrics for training requires in-depth expertise on Machine Learning	Using  Amazon, you can quickly build a performance analysis framework.
Deploying a model and observing its scalability is a big-time issue and headache, as models are heavy	The whole exercise of automatic deployments and continuous monitoring is undertaken by amazon to monitor the performance of the entire ML framework.

So as you see most of your headache is taken away by Amazon Machine Learning as a Service. Lets now throw some light on how we can integrate Amazon AI Services to Servicenow.

 

Common ways to Integrate Amazon AI Services on ServiceNow:

The ServiceNow integration with the AWS Cloud gives users a rich, self-service portal where they can request and track all of their AWS cloud resources using role-based permissions. ServiceNow also provides visibility into cloud costs, based on user, service, line of business, or cloud provider. Remember that here we are taking individual data out of ServiceNow instances, shipping it to storage and model training services on AWS, building a model and putting it back to ServiceNow instance.

There are four different ways ServiceNow can access AWS AI services, includes:

• Lambda Functions:  Lambda functions can easily access Amazon Sage Maker Models. Calling Lambda Functions from ServiceNow is as easy as calling a Normal RestV2 function from service now API. Here is a quick guide on Amazon AWS and ServiceNow integration using Lamda

• Amazon Connect: Leverage Amazon connector provided by VoiceFoundry, explore how to integrate AWS AI on ServiceNow in AWS Quickstarts.

• Amazon Cloud as MID Server: you can set up the MID Servers on your service-now network or in one of your cloud networks. Data is encrypted to the MID Server and between the MID Server and the API endpoint.
• Rest API: Traditional rest API is of course supported, but instead of using your own server, it’s better to use Lambda functions for exposing rest API, which can reduce the headache of another server management.

Considering you already know about the ServiceNow platform, I will briefly talk about the AWS AI services and how to build ML Models to fuel AI in ServiceNow. 

Steps for creating a custom machine learning model on AWS and deploying on ServiceNow

Amazon Machine Learning is available at console.aws.com/machinelearning.the. The following steps will be required for implementing your Machine Learning use case. A step-wise machine learning implementation is as below:

1. Dataset Collection:  As a first step identify relevant data set and transfer the dataset from ServiceNow instance to AWS. You may also import data sources on AWS, remove extra noise and prepare the data for the training ML model. 

2. Data Labelling: Before we train the model using the data, we need to label all the data, so that it can be trained for different categories of data. You also need to split the data into training and validation set. Usually its 70% training and 30% Test Data. Push the best-trained model as per benchmark to production. The entire exercise of data labeling can be done on AWS at an additional cost.

3. Algorithm development/Selection- Amazon offers pre-trained ML algorithms so you don’t need to create them from scratch. some pre-trained AI services that are most commonly used:

1. • Amazon Rekognition: Identifies objects inside an image
   • Amazon Comprehend: Finds entities inside your text
   • Amazon Polly: Converts text to speech
   • Amazon Transcribe: Automate speech recognition
   • Amazon Extract: Performed OCR

4. Training: This is the most critical function of building ML inference engines. You need to select the proper cloud resources before you start training ML inference engine. While training the ML algorithm, you need to compare various models for performance. AWS provides an inbuilt projection of performance graphs, that helps you in visualizing training errors and tells us how well your dataset has been able to minimize them.

5. Model Deployment and Scalability: Once the inference engine is trained and built on Amazon, it is available to integrate with ServiceNow using Amazon Sage Maker or any of the above-mentioned methods. You even don't need to worry when too many requests increase, AWS will automatically handle all of these things. 

6. Storage Management: You can always store your data in Amazon Buckets. All you have to do is create a bucket and create proper roles for accessing them. 

Real-World Example of Integrating AWS AI on ServiceNow:

Smart Contract Management for ServiceNow using Machine Learning

Contracts are very sensitive documents and organizations would like to store them on secure platforms like ServiceNow. The organizations like to get recommendations for contract terms during the negotiations phase. Moreover, the users would also like to extract terms, clauses, and obligations from the contract documents auto-magically! 

Aavenir has developed an enterprise Contract Management Software Solution for ServiceNow, natively on ServiceNow, that uses the Machine Learning from Amazon while storing and executing all contract management functions on ServiceNow. 

To see Live Demo of the AI-enabled contract management solution on ServiceNow, Request a Demo