Machine learning in company? With Now Platform, it’s easy

Sii
BlogerSii
Machine learning in your company? With Now Platform, it’s easy!

Artificial intelligence, machine learning and natural language processing are the buzzwords that keep people talking, not only from the IT sector. Since its release in late 2022, ChatGPT has become the AI model for a wide audience. It can generate entertainment content, solve school homework, or support one at work (Samsung engineers can tell you more about it for sure).

In this article, we will focus on the work-related use cases of AI. We will take a look at Now Intelligence – the AI solutions prepared by ServiceNow on their Now Platform.

Predictive Intelligence

The first module we will examine is called Predictive Intelligence. It’s based on machine learning (ML) and natural language processing (NLP). This module allows for a quick setup of one of four frameworks that learn from the historical data available on the platform and may help you in assigning cases to groups or warn you about a potential outage.

Predictive Intelligence is available in the following licensing packages:

ITSM Professional (solutions for IT processes management),
CSM Professional (solutions for customer service management),
HRSD Professional/Enterprise (solutions for HR),
Now Platform App Engine Professional/Enterprise (solutions for app creation),
Software Asset Management (solutions for software management).

We said setting it up is quick, but why? Well, the why is the low-code/no-code philosophy followed by ServiceNow. And what does that mean? That means developers with limited – or none at all – coding experience may develop apps and features without the need to manually code.

Instead, developers use graphical interfaces, such as drag & drop, to build their solutions faster and cheaper. A fine example of such an interface is Flow Designer used to design workflows which can be used by citizen developers with ease but allow for coding for professional developers as well. You can learn more about low-code/no-code here. And fret not – you don’t need a data science background either to implement Predictive Intelligence.

Frameworks in Predictive Intelligence

The first step should be deciding which framework out of the four available will solve our business needs best:

Classification

This framework is used to make field value predictions when writing to the database. In other words, it can set assignment group, assigned agent, priority, urgency, category and more when the record is created. For example, it can predict the assignment group based on the email’s subject and body.

Classification makes solving cases faster and more efficient since the case is assigned to the group without needing a human dispatcher.

Similarity

This framework is used to find records that are similar in some way, for example based on description. This framework can be used to speed up resolution time of cases by suggesting agent cases that were resolved in the past in order to reuse the solution.

Clustering

This framework is used to find patterns and group records based on them. It can be utilised in various areas, for example:

Incident Management (detect outages based on a group of similar incidents),
Change Management (find candidates for standard changes),
Knowledge Management (gaps in the knowledge coverage),
Virtual Agent (possible new conversation topics).

Regression

This framework is used to predict numerical values based on historical data. A good use case is providing the end-user with an estimated resolution time for their request.

Support in choosing a framework

If you’re still unsure which framework would be best suited for you, we may take a quick look at a couple of ServiceNow’s own implementations of machine learning in their wide array of products. Artificial intelligence can be found in areas such as:

Vulnerability Response – vulnerabilities and remediation tasks may be assigned to the most fitting assignment groups with confidence scores presented in percentages.
Customer Service Management – tasks can be created, assigned and resolved faster thanks to automatic assignment; AI may recommend field values and knowledge articles most likely to be helpful.
IT Service Management – machine learning helps to analyse raw data, and keep the database clean from duplicate and/or inaccurate data.
HR Service Delivery – easier detection of gaps in knowledge articles, case categorization, and recommending knowledge articles.
Flow Designer – Predictive Intelligence may be called to make a prediction during the run of a flow. For example, for a flow that is triggered when an incident is created, the system may request a prediction for the assignment group.

Those are only selected examples of the machine learning implementation in ServiceNow products – if you want to learn the whole scope, be sure the check out their docs.

Creating definitions

Once we’ve chosen the most fitting framework, we may create a Solution definition. In this process we should decide:

what kind of predictions we need,
what data set we should use for AI training,
what should be omitted in the AI training.

It’s important to think those questions through, because the higher the quality of data in, the better the quality of data out. Of course, the data set must be sufficiently big – ServiceNow recommends having at least 30,000 records. One more thing to remember in the setup process is the training frequency – this allows us to ensure the predictions are kept up to date.

Training

Once the solution definition is ready, we may schedule a training. The data set defined in the previous step is sent to the ServiceNow training servers where our solution will be prepared based on the current availability. The training server resides within the same data centre as the instance, so it can be used even with data sovereignty requirements.

The communication between the instance and the training servers is done over HTTPS. Rest assured, only the necessary data is sent and every set is separated from other clients’ sets.

Once it’s done, the data sent to the training server is deleted and all that’s left to do is extensive testing and tuning. Every framework has its own characteristic features, for example, Classification may be tuned based on three factors:

Precision,
Coverage,
Recall.

Once the solution produces satisfactory results, it may be promoted to the production environment where it’ll support agents in their day-to-day responsibilities.

With this knowledge about Predictive Intelligence, we should note that it might not always be a good solution due to various reasons. One of the most obvious is the requirement of having a sufficiently big data set of consistent and accurate data to train the solution. Secondly, machine learning, as every feature, requires a delegated supervisor – its predictions should be reviewed regularly to ensure they’re accurate enough. Additionally, with the passage of time, the data set used for scheduled training may change, so the learning conditions should change too.

Nevertheless, potential value gains – cost reduction, resource optimization, productivity increase, improved customer experience – stemming from Predictive Intelligence are worth exploring the possibility of introducing machine learning on the instance. Maybe it’s even already available in your license, but not yet used? Be sure to check it!

AI Search and NLP

In our discussion about AI on the Now Platform, we mustn’t forget about AI Search and Virtual Agent (a chatbot). Since the use of AI in both these features is similar, we’ll bundle them.

The biggest asset of the search engine on the platform is the use of natural language processing. Thanks to NLP, the user may make queries using normal speech as if they were talking to an agent.

AI may quickly find an accurate answer based on the knowledge bases, FAQs and other sources of information. Additionally, the contextual search takes into account factors such as the requester’s language or location. It’s also interesting to note that the Recommended for You widget available on Employee Center utilizes similar methods – recommendations are based on similar users (for example from the same office or in the same position).

AI Lighthouse/Now Assist

At the end, we should also mention the AI Lighthouse project – a collaboration between ServiceNow, NVIDIA, and Accenture. The goal of the program is the development of generative AI. Such an AI can generate new content based on a prompt provided by the user. For example, a developer may prepare a description of the program in natural language and AI will generate a code fulfilling the description. This would make writing code faster and more efficient.

In fact, generative AI – called Now Assist – has been introduced to the Now Platform in the latest Vancouver release. It can be used in a limited capacity in CSM, HRSD, ITSM and Creator products. For now, available features involve preparing summaries of cases, chats, resolution notes and search results utilizing a large language model (LLM). Additionally, the AI may aid in creating flows in Flow Designer and writing scripts by generating code.

Summary

As we can see, ServiceNow has already implemented plenty of solutions based on AI. One of the biggest assets of their features is definitely the ease with which AI-based solutions may be developed and implemented on the platform.

Despite their successes, ServiceNow continues to invest in AI development in order to offer better & better solutions to their customers. I believe the collaboration with NVIDIA and Accenture will yield good results. And if you ask me whether I’m afraid of AI pushing the developers out of business, I say – if that happens, we’ll have much more serious issues on our plates 😊

Sources

***

If you’re interested in ServiceNow, also take a look at other articles by our experts.

4.5/5 ( votes: 2)

Rating:

4.5/5 ( votes: 2)

Author

Paweł Kulik

A philologist by education, a ServiceNow specialist at Sii by choice. In his day-to-day work with business people, he uses the communication skills he acquired from his studies in the humanities. Privately a Magic player, Sci-Fi fan and cat's owner

All articles written by the author