Unlocking AI for NHS: Now Assist in IT Service Management

Artificial Intelligence (AI) has already started changing the way we consume digital services. Healthcare is an industry where the benefits of AI will be exponential, and the UK National Health Service (NHS) can use AI to make significant productivity gains with ServiceNow. This post will walk through some example readiness steps that can be taken to use Now Assist, ServiceNow’s Generative AI assistant.

First, let’s level set. AI is an encompassing term for tasks performed by computer systems that would typically need human intelligence. They include learning, reasoning, problem-solving, pattern-recognition, and decision-making. AI operates based on algorithms, data, and machine learning techniques.

AI has been built into the ServiceNow Platform for years, from Natural Language Understanding (NLU) in the Virtual Agent, to sentiment analysis, intelligent case routing, and recommended actions for agents. You’ve probably used some of these features, perhaps without knowing.

So what’s new? Generative AI is a category of AI that can create new content or data by learning patterns from a large data set. Recently, GenAI has been added to the Now Platform to enhance productivity and efficiency through conversational and proactive experiences. Moreover, ServiceNow has developed its own domain-specific data set, known as Large Language Models (LLMs), to ensure higher levels of accuracy, control, and transparency, when compared with more general-use data sets.

The focus here will be on Now Assist for IT Service Management (ITSM), as this is the typical starting point for NHS trusts, but there are many more advanced and far reaching use cases. If you’re new to AI you can jump to the bottom of the page and check the terminology table, although this isn’t a technical deep dive post.

Top 3 Now Assist for ITSM Benefits

1. Faster incident handling and resolution:

More times than not, IT staff in the NHS are dealing with a high volume of cases. Coupled with managing complex IT environments, it can be difficult to effectively prioritise workload. Incident or case summarisation provides clear and concise summaries to help IT teams get up to speed quicker, understand request context and previous actions, and seamlessly switch between tasks.
By understanding the impact of an incident in seconds, agents can focus on high-priority issues related to patient or hospital operations first; reducing resolution times and disruption to clinical systems. The manual effort involved in reading through case history, writing case updates or resolution notes, and drafting knowledge articles is removed, freeing up valuable time for IT personnel to work directly on resolving issues.

2. Effective communication, handover, and escalation:

Leadership teams and non-technical stakeholders need to be kept up to date on critical issues that may impact patient safety or hospital operations. More complex cases may require input from multiple teams or escalation to senior or specialist staff. AI-generated summaries provide clear, non-technical overviews, to quickly simplify the impact and status of an incident.
Applying standardised and automated summaries in a consistent format ensures all necessary information is captured systematically, providing a smooth transition between shifts or teams. Escalations can be actioned faster, and management teams have better visibility to aid decision making.

3. Continuous improvement and knowledge sharing:

Due to constant firefighting, IT teams often struggle to find time to capture documentation, build knowledge bases, and improve existing processes. The generation of content such as summaries and resolution notes can contribute to the development of a comprehensive knowledge base. Over time, this content can be used to automatically create or enrich knowledge articles, better understand common issues, and train new staff.
With automation of content generation, human error and inconsistencies are reduced, and reporting aligns to a common data model. For example, longer term, patters across incident summaries can be identified to help uncover underlying problems that may help improve service availability and reduce call outs. With low-value tasks like manual documentation taken care of, productivity in other areas can increase.

Data, Privacy, and Security Considerations for Now Assist

We’re not using AI, where do we start?

People are the biggest asset in the NHS, you can point to data, but without people the data and AI can’t be applied (yet at least!). As with any transformation or new project, it’s absolutely vital that people are brought on the journey. As a starting point, you need to consider:

1. Governance

Check out Building an Enterprise AI Governance Plan for more information, but principles of governance could include fairness, transparency, data security and privacy, and human centredness. This phase will include extensive employee involvement, and an effective comms and upskill plan.
For ServiceNow specifically, your IT team can make use of resources like the Jumpstart your Generative AI accelerator through the Impact (customer success) team.

2. Risk

3. Improvement

As your organisation, requirements, and AI maturity, all evolve you’ll need your processes to remain optimal and relevant. Make sure you regularly review performance, impact, and compliance alongside IT and user feedback.
For ServiceNow specifically, your Impact and account teams can help evidence realised value, and you can share experiences and lessons learned with the NHS Community.

Finally, make sure to involve your ServiceNow team at all stages. They want you to be successful and can help you tap into a variety of resources. ServiceNow are top of the Foundational Model Transparency Index which is testament to the transparent and accountable principles ServiceNow apply to their AI development.

What happens to my data when I use Now Assist?

For NHS customers, your instance should be running in the UK. You can check this in Now Support by selecting the instance from the instances dashboard. ServiceNow recently announced a £1.15bn investment into its UK business, which includes expanding AI capabilities and NVIDIA hardware into both UK data centre sites.

At the time of writing (Dec ’24), when you use Now Assist, the prompt is sent to our regional data centre hub to be processed. For EMEA, the Now LLM Service regional hub is based in Germany. Do note that prompt data is never stored or co-mingled with other customers data; customer prompts (inputs) and responses (outputs) are not available to other customers.

There is no data residency during processing, as once a response is sent back to your instance (UK), data is immediately deleted from the regional data centre (Germany). Your data is only ever stored (written to database) in your own instance (UK). All data, including prompts and responses, stays within the regional data boundaries specified. Once the Now LLM Service is available in the UK data centres, prompts will be processed and returned on-site.

Additionally, processing data is not used for model training. Training of the Now LLM Service is an entirely separate and anonymised process, for which you can easily opt out at any time from the Now Assist admin panel. More information on LLM training, accuracy, context, risk, and limitations can be found in the ServiceNow LLM Model Cards.

How is my data secured on the Now Platform?

The ServiceNow Security for the UK Public Sector white paper provides more specific information in relation to the overall security of the ServiceNow platform. But in addition to the steps outlined above, the platform has achieved the following industry and compliance standards:

ISO 27001, 27017, 27018, 27701, SOC 1/2, and Cyber Essentials Plus certification
GDPR compliance (includes Data Protection Act 2018)
Data Security Protection Toolkit (DSPT) completion, migration to Cyber Assessment Framework (CAF) for suppliers will be completed inline with NHS England’s roadmap

ServiceNow provides tools, such as the Sensitive Data Handler, to redact Personally Identifiable Information (PII) in prompts. This helps to mitigate against any PII being included in cases that are summarised by Now Assist. Security and offensiveness guardrails are also applied with Now Assist Guardian, to protect against inappropriate or unsafe answers, and security threats like prompt injection.

Getting Started with Now Assist for ITSM

What are the prerequisites for Now Assist?

There are very few prerequisites for Now Assist from a technology perspective, they include:

A ServiceNow instance running a supported version
A ‘Pro Plus’ Now Assist license, e.g. ITSM Pro Plus – note that you can start out with a small subset of Now Assist licenses to prove value before rolling out to the wider teams

What should I do next?

In terms of additional resources, start with the free Now Assist Essentials course, which covers:

Generative AI terminology
How Now Assist works in the platform
How to Install Now Assist
How to use the Generative AI Controller in Flow Designer and Virtual Agent
How data is used with Generative AI

You’ll also find plenty of information in the Now Assist Documentation and Now Assist Community, such as the Activate Now Assist for ITSM in 15 Minutes article. To get started with Now Assist and enable the capabilities, the best resource to follow is the Now Assist for ITSM Implementation Guide in Now Create.

How do I install Now Assist?

To demonstrate how easy it is to start using Now Assist, I’ll show the activation process below. If you are implementing Now Assist, do refer to the Now Create guide linked above.

1. Now Assist is installed from Application Manager. Search your licensed applications and plugins for ‘Now Assist for ITSM’. Locate the Now Assist for ITSM store application from the search results. Open the application, and click Install.

2. From the navigation pane, browse to Now Assist Admin. The Now Assist Admin Console is the single place to activate plug-ins, configure skills, and analyse GenAI performance for Now Assist products across all workflows (including ITSM).

3. Plugins are managed from within the Now Assist Admin console, under Settings. Here you can install the Now Assist for ITSM plugin.

4. You can see your license information from the Account section under the Settings tab too. This is where you’d opt out from anonymised data training of the Now LLM Service.

5. The Now Assist panel is an optional conversational interface for agents to interact with Now Assist skills and features, it provides recommendations to help resolve customer issues faster.

The Now Assist panel can also be turned on from the Settings tab of the Now Assist Admin console.

6. It’s as easy as that to install the Now Assist application and plugin for ITSM! The third and final part is to activate the skills (features) you want to use. A Now Assist skill is essentially a GenAI powered use case. You can see some examples in the screenshots below.

Review the Now Assist for ITSM Implementation Guide in Now Create to understand the individual configuration options for each available skill.

This great demo from the ServiceNow Community provides a further, extended, example of Now Assist skills:

Summary and Terminology

To summarise, Now Assist opens the door to productivity gains for both employees and IT agents alike. We’ve shown how quickly Now Assist for ITSM can be enabled on an existing ServiceNow instance. The technology is smooth, easy to use, and constantly evolving. Once licensed, customers will get access to new skills and feature improvements as they are added through store or family releases.

The steps and considerations discussed are the same for enabling Now Assist plugins for different ServiceNow applications. You’ll find below a table of terminology that is also consistent across applications. We haven’t discussed all of these, but it’s a good starting point for fully understanding the ServiceNow content in the additional links provided.

The technology is only part of the picture. Hopefully the resources, transparency, and information outlined in this post can help NHS organisations put together a plan that incorporates people and process to deliver meaningful value and experiences from the ServiceNow platform.

TERMINOLOGY	DESCRIPTION
Now Assist	ServiceNow’s generative AI assistant designed to enhance user productivity and efficiency through conversation and proactive experiences.
Now LLM Service (Large Language Model)	Collection of models that ServiceNow hosts and manages, including both ServiceNow fine-tuned Now LLMs, and third-party open source models that are not fine-tuned.
Skill	A Generative AI powered use case.
Assists	The unit of measure of Now Assist. Outcome based, not token based. Skills may consume different levels of assists based on complexity and/or value. Assists are pooled at the account level.
Model Cards	Fact sheet for a machine learning model. Explains what the model is designed to do, how well it performs, the data it was trained on, and any limitations or potential issues of the model.
Pre-trained or foundational models	Related concepts, though there are differences. Trained on a large dataset, used as a starting point for fine-tuning or more specific models. We start with open-source models in our lifecycle.
Prompt	The textual input supplied to the LLM to process and generate content. Inputs can be in the form of a question, statement, or instruction.
Fine Tuning	Further training of an LLM using generated responses, prompts, and edits to responses. Helps the model become better at a specific task, such as summarisation or content generation.
Alignment	Process of ensuring the LLM behaves in ways that are consistent with human values, intentions, and ethical considerations. Involves tuning the model so that outputs are ethical, safe, and aligned with user expectations.
Inference	The process of using an LLM to understand and generate responses based upon a prompt. Inference is performed in our regional GPU compute hubs.

Source link