At the start of May, ServiceNow announced what it called “Autonomous Workforce” at its Knowledge 2026 conference – the promise that AI agents will manage HR processes with growing autonomy. Days earlier, they had published a 64-page playbook called How to Make AI Work for People.
I read it. It’s well made. The framework is solid, the use cases are real, the AstraZeneca and Siemens numbers are impressive. But there’s one line somewhere in the middle that doesn’t quite fit the narrative – and for that reason, it strikes me as the most honest thing in the entire document.
“Only 24% of organisations have formal HR data cleaning processes.”
Think about what that actually means. ServiceNow is building autonomous HR agents. It’s selling the future of work. And the data that autonomy needs to consume is – in 76% of cases – managed without any formal quality process.
I’m not criticising ServiceNow. The problem they identify is real. I’m using that number as a way in to something that’s been on my mind for a while: the gap between what the industry promises about people intelligence and what actually happens when you try to implement it.
The problem isn’t the tool
The dominant narrative around AI in HR follows a pattern we’ve seen before: the technology is here, all that’s left is adoption. Playbooks are written in that logic – here are the five steps, here are the use cases, here is the ROI.
ServiceNow’s playbook defines five phases: modern HR technology, optimised processes, AI implementation, capacity reallocation, and operating model transformation. And there’s an observation buried in the final pages: most organisations stop at step 3. They implement the technology. They never reach step 4, where freed-up capacity gets reallocated, or step 5, where the operating model actually changes.
The value is in steps 4 and 5. The investment is in steps 1, 2 and 3.
This isn’t a criticism of organisations. It’s an observation about what technology adoption actually requires that rarely makes it into playbooks: changes to decision structures, routines, accountabilities. That’s harder to package into a framework.
Who is this playbook actually for?
ServiceNow is a US company valued above $100 billion. Its reference clients for people intelligence are AstraZeneca, Siemens, Adobe. Platform licences start where the annual budgets of many European companies end.
That’s not a problem with ServiceNow. It’s an observation about context.
Market playbooks – and this one is better than most – are written for organisations with dedicated data teams, enterprise-grade system integrations, and the capacity to run an 18-month implementation. They represent the standard for a specific segment of the market.
Most European companies, and nearly all SMEs, aren’t in that segment. Which doesn’t mean the problem is different – it means the path has to be different.
People intelligence isn’t a concept that belongs to AstraZeneca. It’s the capacity to make decisions about people based on contextualised, up-to-date data, rather than intuition, precedent, or internal politics. That’s as relevant for a company of 200 people as for one of 200,000. What changes is the scale, not the principle.
What does “people intelligence” actually mean?
I’ve been hearing this term more and more over the past two years. In very different contexts, meaning very different things.
Sometimes it’s a synonym for people analytics – dashboards of turnover, headcount, absenteeism. Sometimes it’s talent intelligence – understanding skills, potential, career trajectories. Sometimes it’s just a new name for what used to be called human resources management.
My reading, built over ten years working on people systems: people intelligence is useful when it informs concrete decisions that would otherwise have been made differently – or not made at all.
It’s not a dashboard. It’s a shift in the quality of questions an organisation can ask about its people.
ServiceNow documents a good example of this: their internal HR agent now supports 1,042 employees, versus 412 before the implementation. That number isn’t about HR efficiency – it’s about what becomes possible when capacity is reallocated. But there’s a question the playbook doesn’t directly answer: which decisions could the HR team now make that they couldn’t before? That would be the interesting number.
Where this leaves us
If you’re a large organisation with budget and team, ServiceNow’s playbook is a solid reference. The principles are sound, the use cases are real, and the five-step framework is honest about where the value actually sits.
If you’re not – and most of us aren’t – the problem is the same but the path starts somewhere else. It starts with the quality of the data you already have, not the systems you don’t yet have. It involves simpler questions: which decisions about people do we make regularly that could be better with more context? What do we know about each person that’s scattered and inaccessible when we need it most?
The underlying question is the same for everyone: does the data we have about our people actually inform the decisions we make about them?
The honest answer, in most cases, is still no. And that’s true whether you’re running ServiceNow or a spreadsheet.