There’s been a shift in how people talk about AI. The early excitement — the novelty, the fun experiments, the “make me a picture of a giraffe in a spacesuit” moment — is starting to fade. The puppy-love phase is ending, and a more realistic relationship is taking shape.
That’s a good thing. Because the next phase is about trust, not tricks.
Generative AI tools like Microsoft 365 CoPilot impressed us quickly. They helped with emails, meeting notes, summaries, and all the small things that make work easier. But when CIOs start exploring AI for billing, caregiver scheduling, case documentation, or anything with compliance risk, the bar gets much higher.
And that’s when the cracks begin to show.
If someone asks CoPilot a simple question — “When is our next company holiday?” — and it gives the wrong answer, trust evaporates. The tool is dismissed. Doubt sets in. And the entire AI initiative suffers.
The truth is that CoPilot usually isn’t getting these answers wrong because the model is broken. It’s getting them wrong because it’s pulling from content that shouldn’t be there at all.
CoPilot Only Knows What You Feed It
One human services organization we support discovered this firsthand. They rely heavily on SharePoint pages for procedures, forms, FAQs, and operational guidance. Over the years, different versions piled up. Some kept current. Some forgotten. Some created by one team and replaced by another.
And SharePoint’s analytics didn’t help much. You can see activity for individual pages, but there’s no simple way to view all pages at once and understand which ones employees actually use.
Meanwhile, CoPilot was doing exactly what it was designed to do.
It was looking at everything.
That meant mixing outdated content with current content. Drafts with official guidance. Old department versions with newer company-wide standards.
So when someone turned to CoPilot for help, the answers reflected the chaos underneath. And once that happens, people stop relying on it — not because they don’t believe in AI, but because they no longer trust the content behind it.
Seeing What Your Staff Actually Uses
To help them get clarity, we automated a way to surface the information SharePoint doesn’t show out of the box: which pages employees use, which ones have been untouched for months or years, which ones are competing versions of the same idea, and which ones are quietly polluting search and AI outputs.
Once they could see the environment clearly, the path forward became obvious. They archived old versions, consolidated duplicates, and tightened their internal rules around what gets published and maintained.
CoPilot didn’t suddenly become smarter — it simply had better material to work with. Its answers improved noticeably because the noise had been removed.
This Is a Governance Problem, Not an AI Problem
The real story here isn’t about AI at all. It’s about governance — the everyday kind that ensures information stays accurate, up-to-date, and intentional.
This sits at the foundation of Tommy’s Hierarchy of Productivity. When governance is weak, everything above it wobbles. When governance is strong, tools like CoPilot finally have a healthy environment to operate in.
Organizations are now realizing this. The hype cycle is calming down, and the practical work is becoming clearer. Clean content. Fewer versions. Clear ownership. Sensible lifecycle rules. Nothing flashy — but everything essential.
CoPilot wasn’t inconsistent because it was flawed. It was inconsistent because the content behind it was.
Trust Comes From Clarity
People trust AI when it reflects their reality. They rely on it when it removes friction instead of creating confusion. And none of that is possible if the underlying content is outdated, fragmented, or unmanaged.
So if your team is beginning to depend on AI for real work — not just playful experiments or convenience tasks — strengthening your governance practices is one of the highest-leverage steps you can take. It isn’t glamorous, but it’s foundational. AI can only be trusted when the environment around it is trustworthy.
A Helpful Next Step
If you’re thinking about how governance fits into your broader productivity journey, our free Productivity Toolkit for Care Providers includes maturity guides and practical frameworks for building the kind of foundation AI depends on. It’s available if you want a simple place to start — and if not, I hope this post gave you something useful to think about.
