Watching the adoption of AI over the past 12 months, particularly generative AI, has been interesting for many reasons – but particularly so because of the frenzied nature in which vendors (and enterprises) seem unwilling to slow down and take pause for thought. Despite numerous warnings from experts and industry leaders, there appears to be a consensus in the market that AI is going to dominate the economy for the next few decades and those that are first to utilize it effectively will yield the biggest returns.
This is a simplistic take, of course there are other reasons companies and governments are unwilling to slow down (geopolitical reasons being a big one); but it’s undeniable that in a lot of cases the general attitude has been ‘act now, we will worry about the unintended consequences later’.
Yes, platforms and builders of models will tell you that there are sufficient guardrails in place and that work is being done to avoid nefarious ethical scenarios, but the reality is that we don’t know what the long-term adoption of AI tools will mean for society, nor do we really understand how these generative models really work. The neural networks they are based on defy a certain level of explainability and people seem to be just fine with accepting the net positive efficiency gains, if the outputs are broadly in line with what we would expect.
How we control the adoption of generative AI tools now, will play a role in how we course correct for unwanted outcomes down the line. And ServiceNow sees this as an opportunity – a competitive advantage it could have in the market – as people wise up to the need for stringent governance.
Governance in enterprise technology has always been important. From monitoring outages, to data lineage and identity management, governance has played a key role in establishing a trustworthy IT estate. However, ServiceNow believes that this will become even more critical as generative AI adoption advances.
New governance tools
By way of background, ServiceNow itself offers swathe of generative AI capabilities, largely underpinned its Now Assist tool – a conversational interface that can summarize information, carry out tasks, help build low-code apps and automate workflows for business areas that include IT, HR and customer service. It recently also announced the introduction of Agentic AI into the Now platform, as part of its latest Xanadu platform update.
The reason ServiceNow may be able to play a critical role in generative AI governance, is that the Now platform is based on a ‘one data model, one platform’ approach. The vendor operates as an engagement layer that sits atop systems of record, acting as a workflow platform that unifies enterprise systems of record, allowing buyers to scale ServiceNow as part of a ‘one company strategy’.
Rather than having to cobble together multiple integrations, driven by various department leads, ServiceNow argues that its customers can rather think about service as a unified approach – with data and work flowing through one platform across the organization.
With this in mind, ServiceNow this week announced new tools for AI Governance for Now Assist, with the vendor stating that it is laying a foundation to help customers build trust with their AI deployments. Speaking with Adam Spearing, ServiceNow’s Head of AI Innovation EMEA, he said:
I think this is one of the most under-underestimated areas of AI. If you think about AI, it’s a phenomenal opportunity for transformation. but actually, when you break it down to what’s practical and deliverable, you start thinking about how you govern AI.
It goes way beyond our traditional IT governance. IT governance used to be integration, UAT testing, data integrity. When you start talking about AI governance, it becomes much broader, because the technology is so pervasive, the full breadth of AI governance starts going into things like: how do we train our organization? How do we skill them to be AI competent? What we don’t want them doing is putting our quarterly report out on ChatGPT the night before we’re about to release it.
The three core components to ServiceNow’s AI Governance release this week include:
- Now Assist Guardian – built‑in monitoring and AI guardrails for customers to better control the use of GenAI on the Now Platform. This includes the management and mitigation of offensive content, security vulnerabilities, and the exposure of sensitive information.
- Now Assist Data Kit – ServiceNow highlights that managing and consuming data for AI use cases is often cumbersome. With Data Kit, users can create and manage datasets for AI skills and applications, develop ground truth datasets to benchmark for accuracy and help predict AI outcomes, and evaluate the effectiveness of experiences built with Now Assist Skill Kit.
- Now Assist Analytics – aimed at offering visibility into the adoption, usage, and performance of Now Assist across the enterprise; ServiceNow hopes that with these insights, customers can make more informed, data‑driven decisions to advance GenAI adoption, evaluate ROI on GenAI investments, and enable better business outcomes.
Commenting on the significance of why governance is so important in this context, Spearing added:
I think with AI, when we start looking at language models, it’s not just about what went wrong – it becomes more systemic. Once a language model starts to go off in a certain direction, you’ve really got to catch that quickly before it snowballs into a bigger problem overall.
I think this world of AI governance is a superset of our traditional governance, if you like. It’s the traditional governance, the data integrity, but then it’s the algorithmic bias, it’s the language model programming, it’s all of that.
It’s more important, because most of these language models, where we’re seeing the value in the ROI, it’s how we are changing that customer and consumer experience. And you can’t afford to get that wrong, because you can’t get that wrong too often.
I think this is going to be one of the defining pivots between people who are super successful and people who dabble with AI – how they run that governance across the company.
The long-term view
With the advancements in agentic AI, where more autonomy is given to AI models to carry out tasks on behalf of humans, Spearing believes that this is where governance becomes absolutely critical. If enterprises want to take advantage of the efficiency gains provided by autonomous agents, then they need to be able to understand and control their working. Governance is key:
When we start talking about agents, governance becomes even more important, because ultimately the only way you can reduce the human in the loop is if you have absolute trust in your governance.
Then the question is: how do you make your governance, your guard rails, sufficiently rigorous without being burdensome and slowing you down? Then you manage by exception – you trust the technology to tell you ‘this algorithm has got bias in it’. Governance, I think, is one of the most important things for the long term – we have to get it right.
The responsibility of IT is the integrity of the data. Now it’s becoming the integrity of the algorithms and workflows, as well as the data. I think you will always want a higher degree of governance and oversight, more so than ever before.
My take
The speed at which generative AI is not only developing but also being adopted in the enterprise means that governance will eventually become a strong focus for buyers. Mistakes will be made and I have no doubt that there will be a retrospective stock take on how generative could be better managed. It’s by no means the only vendor to push governance as a competitive advantage, but ServiceNow does have somewhat of a unique proposition to excel at it because of its ‘one data model’. It could be a savvy move for ServiceNow if it can get its customers to showcase how governance in unison with generative AI adoption leads to better outcomes.
Spearing’s hope, though, is that buyers pay close attention to ServiceNow’s roadmap, as more updates will be coming thick and fast over the coming months:
We need our customers to look at what we’re doing every quarter. This is coming out really quick. We need them to go, right, ‘I want to use the analytics, I want to use the data kit, I want to use the Guardian’ – because in 90 days, we’re going to have another conversation about a suite of other things.
If they build that in now, when IT go to finance and say ‘you’ve got to pay the bill for ServiceNow’, they can say ‘here’s the value it’s delivering’. Every customer I meet, I say to them, you have got to be looking at our roadmap with us every quarter, every 90 days, picking off the things you’re going to deploy – not just every other release. This is a true SaaS model of deployment.
link