If you’re a CTO or IT Director right now, there’s probably a question hanging in the air from the Board, the exec team, or both:
“So… when does the AI magic start happening?”
And I know what you’re probably looking at when that question comes up.
Your SharePoint estate. Your permissions model. Years of accumulated digital debt. And a quiet sense of dread.
This isn’t because Microsoft Copilot doesn’t work. It’s because most organisations are being sold a marketing professional’s version of Copilot that doesn’t match reality.
Most vendors talk about Copilot as if it’s plug-and-play. Turn it on, train people for an hour, and suddenly meetings summarise themselves, emails write themselves, and productivity shoots up.
In practice, what we see far more often is this: Copilot becomes a very expensive meeting summariser.
Not because the technology is broken. But because it’s been dropped into an environment that simply isn’t ready for it.
Copilot is a high-performance engine. It assumes clean data, sensible permissions, consistent structure, and users who understand how to work with it. Trying to run it on a messy tenant is like putting a Formula
1 car on a potholed B-road and wondering why it doesn’t perform.
One of the most common issues we see is what we half-jokingly call the digital dumping ground.
SharePoint sites full of documents from 2014. Multiple versions of the same policy. Folders called “Final_FINAL_v3”. No clear ownership. No agreed source of truth.
Copilot doesn’t fix that. It just finds it faster.
If the underlying data is outdated, duplicated, or poorly organised, Copilot will confidently surface the wrong thing. And once users see that happen a few times, trust evaporates. They stop using it, or they double-check everything manually, which rather defeats the point.
This is where a lot of pilots stall.
Then there’s permissions.
Every experienced IT leader has had that shudder-inducing moment when they realise how much access has been granted over the years “just to get things moving”. Shared drives, inherited permissions, exceptions piled on exceptions.
Now introduce an AI that can search across everything a user can access.
I’ve spoken to CTOs who’ve had the cold realisation that Copilot could surface something like a “Salary_Review_2025” document simply because no one ever cleaned up the permissions. The tool didn’t do anything wrong. It followed the rules it was given.
This is where AI changes the risk profile. In a pre-AI world, poor permissions were a latent issue. In an AI-enabled world, they become immediately visible.
Another uncomfortable truth: training isn’t a one-hour webinar.
Giving people Copilot and hoping they’ll “figure it out” doesn’t work. Equally, teaching them which buttons to press isn’t enough.
What actually changes with Copilot is how work gets done. Drafting becomes reviewing. Searching becomes interrogating. Speed matters less than judgement. People need to know when to trust an output, when to challenge it, and when to ignore it entirely.
Without that cultural shift, adoption stays patchy. Some people use it heavily. Others avoid it. A few quietly use their own tools outside IT’s visibility. None of that delivers meaningful operational impact.
There’s also an infrastructure truth that rarely makes the sales deck.
AI assumes a healthy tenant. Reliable connectivity. Consistent platform configuration. Identity and security that reflect how people actually work, not how an org chart looked five years ago.
We see organisations trying to layer Copilot on top of:
You can’t build a penthouse on a crumbling foundation. AI just exposes the cracks faster.
As a Microsoft partner, we see the full range of Copilot rollouts. The ones that underwhelm, and the ones that genuinely change how people work.
The difference is almost never the licence. It’s the foundations.
The organisations that succeed tend to have:
None of that is glamorous. All of it matters.
You’ll see plenty of messaging that says: “Transform your business with AI.”
The more honest version is this: Fix your data foundations so AI actually has something worth finding.
Copilot is powerful. We’re genuinely excited about what it can do. But our job isn’t to help you switch it on quickly. It’s to help you make sure it actually works - safely, sustainably, and at scale.
If you’re feeling pressure to “do something with Copilot”, but you’re not convinced your foundations are ready, you’re not alone.
We don’t want you to waste six months on a pilot that never scales. That’s why we built our AI Readiness Assessment. It focuses on the unglamorous reality: permissions, data structure, governance, performance, and user readiness.
No sales fluff. Just a clear view of what needs fixing so Copilot can deliver real value, not just better meeting notes.
If that sounds useful, let’s have a conversation before you flip the switch.