When the World Wide Web was made publicly available over 30 years ago, business leaders weren't just working out how to develop an “internet strategy”. Instead, it heralded entirely new fields, such as e-commerce and online marketing, that simply didn’t exist before.
I see the same thing happening with AI. We should be looking beyond implementing AI for the sake of AI and exploring instead how we use the technology to do things you couldn’t do before. This is where the real potential lies, not just speeding the same things up, but experimenting to find new ways to deliver additional value. It requires a different type of thinking which some organisations are struggling with.
It's my belief that the more AI can do, the more human judgement matters. In this blog, I’ll set out what human-led AI means in practice, and why it drives our strategy at UP3.
Human-led AI puts people at the centre
As companies and individuals hand more decisions and output to AI, it’s vital that people stay in the loop as the experts.
AI is brilliant at pulling huge amounts of information together and producing something new. But people bring experience, taste and a sense of nuance that AI can’t match.
That makes AI a powerful assistant. It can take on work you never had time for or stretch you into areas just beyond your expertise. That’s where the productivity gains come from.
You might no longer build a document or design from scratch, but it’s crucial that a person still reviews the output and checks it’s accurate.
AI also unlocks the creativity inside all of us. People who understand a challenge inside out can now bring once impossible ideas to life, without needing to write thousands of lines of code.
But new ideas still need checking. Think of a technical architect reviewing the code behind a ‘vibe-coded’ app, one built by describing it to AI rather than writing it by hand, before it goes into full production. Someone with deep expertise can spot errors or misconfigurations that an AI might miss. And because AI has written the code, the expert’s time is freed up, meaning they can continually focus on the problems they can help fix.
That validation step is critical for any AI programme, and that is why it is now a core part of our updated AI Unlocked methodology. It’s the point where people check AI is doing what’s expected, decide what to adjust, and choose a different path if needed.
Setting the rules AI plays by
Human-led AI starts with checking outputs, but it can’t stop there.
The average person reads around 240 words a minute. A large language model can process millions of words in a second, and automated decisions can happen in milliseconds. At scale, no one can review every decision at that speed.
People must lead AI in a different way, resetting the conditions it works within, starting with data. Good data has always been fundamental to processes that work. What’s changed is that AI amplifies whatever it’s given. If the data is wrong or misaligned, AI will confidently make the wrong decisions, faster and in more places.
AI’s ability to act at speed on flawed data is a risk that many organisations are extremely wary of. According to ServiceNow’s Enterprise AI Maturity Index 2026, 41% of employees rank data silos among their organisation’s biggest AI issues. And only 16% have widely or fully replaced fragmented legacy systems with an integrated platform.
Governance is another rising concern, with leaders increasingly wary of AI agents going rogue and for example deleting entire production databases. Organisations need visibility, control and accountability over what their agents are doing.
That’s the problem ServiceNow’s AI Control Tower is built to solve. It gives visibility to your AI Systems, Models and Agents across your organisation, when they were last used and what value they are providing.
Data and governance can sound like technical plumbing. But both come down to human decisions – what good data looks like, and where the line between right and wrong sits.
Without those foundations, no AI strategy can move forward for long.
Human-led AI comes to life in UP3’s AI Centre of Excellence
Keeping people ahead of AI takes constant work which is why we established our AI Centre of Excellence (CoE) at UP3.
Made up of cross-functional representatives, the CoE brings real customer challenges to the table, from testing a new use case to proving the ROI of a new concept. It also shapes how we use AI to improve our own services for customers and our internal processes.
For example, we recently looked at how we triage incoming requests - deciding how urgent each one is, what it needs and who should handle it. Our starting point was an AI agent. But when we assessed the ROI and the quality of the output, a custom AI skill in ServiceNow, designed for that one task, delivered the same consistent results. It uses a fraction of the "assists", ServiceNow's measure of AI usage, so it costs far less to run.
The team stays at the forefront of AI developments and the latest ServiceNow releases, so we can pass that knowledge straight to our customers. It’s where we build the expertise that helps our customers make the judgement calls on what good looks like.
The golden question: what comes next?
AI is moving so fast that it’s getting harder to predict what comes next. Many of the customers I speak to are still bedding in AI capability that first became available two or three years ago. But the important point is that they have started on the journey.
What I can say with confidence is that data quality will be a major focus for many organisations over the next 12 months. And closing that data quality gap requires people to define what great looks like.
The Internet’s greatest value came from the people who saw what it made possible. AI will be no different and that’s what we mean by human-led AI. The technology sets the pace, but people need to decide where it goes and what good looks like when it gets there.
