HR doesn't need more AI tools. It needs a new foundation. For three years we set out to answer the hardest questions about the workforce, and we learned that the answers were never going to come from the systems everyone keeps mining. They live somewhere no software had ever reached. Here is what we found, and why we think it begins a new era for HR.
The questions that decide the company have no answer
HR has never had more data. Headcount, attrition, engagement, time to hire, all of it live on a dashboard. And yet when the executive team asks the questions that actually shape the company, the dashboard goes quiet.
What HR can measure
Headcount
4,210
Attrition
11%
Engagement
7.4
Time to hire
38d
Plenty of data. None of it strategic.
What the executives actually ask
“If AI absorbs 30% of the work in the next three years, which roles do we re-skill, and which do we let go?”
“Who are our top 50 people, and what's the probability we lose them in the next 12 months?”
“If we take 15% out of the cost base, where do we cut, and what breaks?”
These are not edge cases. They decide who gets re-skilled, who gets retained, where the business cuts and where it invests. The dashboard cannot answer one of them. And it is not because HR lacks data. HR is drowning in it.
We went all in on analytics. Then we hit a ceiling.
For three years we worked on a single problem: answer the most important questions about the workforce. We started where everyone starts, with analytics. We integrated the HR systems, unified them into one model, and built an AI analyst that answers any question in plain language in seconds. On the analytics, we believe we built the best there is.
And we still hit a wall. Not a wall of analytics, a wall of data. The best model in the world, asked why attrition is rising or which roles to re-skill, can only reason over what it has been given. And what it had been given was never enough.
Forty years of HR data, all from the same 20%
Step back and the pattern is obvious. Payroll in the 1980s. Employee master data in the 1990s. Engagement and performance in the 2010s. People analytics in the 2020s. Four decades of progress, and every layer drew from the same place: your systems. We got better and better at analyzing the same sliver of the truth.
1980s
Payroll
Pay, tax, time & attendance
1990s
Employee master data
Org, role, compensation, tenure
2010s
Engagement & performance
Surveys, reviews, talent ratings
2020s
People analytics
Smarter methods, same 20%
Four layers, four decades, one source: your systems.
20%
In your systems
HRIS, surveys, dashboards. The what. Mined dry.
80%
In your people's heads
Context, judgement, the real why. Never captured.
That sliver is about 20% of what matters, and it is all the what. The 80% that explains it, the context, the judgement, the real why behind every number, never entered a system. It lives in people's heads. Researchers named it tacit knowledge sixty years ago: we know more than we can tell. Every restructuring debate, every culture call, every "where does AI go first" decision turns on that 80%. No dashboard ever held it.
This was our ceiling: the best analytics in the world, running on a fifth of the picture.
The 80% was never a secret. It was just unreachable.
Plenty of people have tried to get at it. There were three ways in, and each gives something up:
- Consulting. Deep, but €500K and up, months of work, and only a sample of the org is ever heard. The insight is stale on arrival and walks out the door when the engagement ends.
- Surveys. Broad, but a tick-box. Around 100 words per person, and never the why.
- Workshops. Insiders interviewing insiders. Candour drops, and the loudest voice in the room wins.
Slow, shallow, or skewed. The 80% stayed locked, not for lack of trying, but because no one had a way to reach everyone, in depth, at once.
Then three things changed
In the last two years, three breakthroughs landed at once, and together they changed what is possible.
Natural voice
~1.5s
Real-time speech that hears emotion and handles interruptions. People talk instead of filling in a form.
Memory at scale
8K → 1M
Tokens of context. Hold a full conversation in mind and cross-reference hundreds of them at once.
Reasoning
66% → 83%
Reasoning in conversation, in a single year. It follows a guide and probes, like a skilled interviewer.
For the first time, software can hold a real conversation, with everyone, at the same time. The one thing that made the 80% unreachable, scale, was gone.
So we built the second engine
The logic was simple once we saw it. The valuable data lives in people's heads. The technology can finally reach it. So the job was to get the 80% out of people's heads and into the analytics. That is the engine we spent the last stretch building: conversation intelligence. AI runs hundreds of real conversations across the workforce in parallel, probes and cross-references like a skilled interviewer, and turns what it hears into structured data the analytics can reason over. Not a survey. Not a chatbot. A new source of HR data, the first in forty years.
This is People Intelligence
Put the two engines together and you get what neither could deliver alone. Analytics reads the what from your systems. Conversation intelligence gets the why from your people. Fused into one answer, that is People Intelligence: the practice of answering a workforce question by combining the data in your systems with the knowledge in your people's heads.
1980s
Payroll
Pay, tax, time & attendance
1990s
Employee master data
Org, role, compensation, tenure
2010s
Engagement & performance
Surveys, reviews, talent ratings
2020s
People analytics
Smarter methods, same 20%
Now
People Intelligence
The 80%, from people's heads
This is the new layer. For forty years, every advance in HR was a smarter way to analyze the same 20%. People Intelligence adds the 80%. It is not a better dashboard. It is the foundation the dashboard was always missing.
The new era
Individual AI is everywhere. Every employee has a copilot in their browser. But organizational AI, the kind that understands how your specific company actually works, who holds the knowledge, why people stay or leave, where AI should go first, does not exist yet. It cannot, as long as the 80% stays locked in people's heads.
That is the era we are building toward. HR doesn't need more AI tools. It needs People Intelligence.
