How We're Surviving AI
Collab365 watched AI dismantle fourteen years of work in eighteen months. Here is the unvarnished account of what we built next.
It was November 2022. We were at a team away-day in Buxton.
A member of the team opened a laptop and showed us ChatGPT. It went around the table. Questions flew at it from every direction. It answered better than any of us managed. Someone fed it a half-finished blog post and watched it finish the thing in about fourty-five seconds.
For about twenty minutes, it felt like Christmas. Then the thought arrived.
If AI is making us this massively productive, what does this mean for our customers?
Collab365 was a Microsoft 365 training business. Fourteen years. 18,000+ professionals trained. Clients including Microsoft, Nike, the NHS, and Deloitte. Four people. Telford. Profitable.
And in that moment, sitting in the living room of an Airbnb in the Peak District, we realised the business we had spent fourteen years building might not survive what was coming.
We did not act fast enough. Nobody does, when something is working.
In the eighteen months that followed, we watched six distinct forces dismantle the foundations of our business simultaneously. Search traffic collapsed. The community stopped asking humans for help. Even the software we were teaching began competing against us.
We had to adapt, or the AI Revolution was going to claim its next victim.
Forget The Industry Reports. They Cannot Save You.
We know. You have heard a version of this story a hundred times. AI is coming. Jobs will go. Industries will be disrupted. Every conference, every LinkedIn post, every vendor pitch starts with the same scaffolding of alarm.
This is not that.
This is a specific account of what happened to one specific business. The mistakes we made. The things we built. The exact decisions we had to make under real commercial pressure. Not as a warning. As a map.
Because here is what we learned: when you are inside the collapse, there is no industry report that helps you. There is no benchmark. The only thing that is any use is someone who has already been through it telling you exactly what they did.
The Day Two Directors Walked Away So The Company Could Survive.
Coursera and Udemy collectively lost over $8 billion in market value. Pluralsight, once worth $3.5 billion, nearly ceased to exist when lenders had to inject $275 million just to keep the lights on. If companies with hundreds of millions in funding could not survive on the old model, Collab365, a four-person operation in Telford, had no chance playing it.
So we stopped playing it.
What came next was the darkest part of the crisis. It still cuts.
These weren't just directors. They were close friends who had built this company with me. But the impact on sales was so violent that the business could no longer support all four of us.
So two of those friends made the ultimate sacrifice: they walked away.
They voluntarily stepped out of the business to give the remaining two of us the runway we desperately needed. I will be eternally grateful for what they gave up to give Collab365 a fighting chance.
Their exit threw us instantly back into start-up mode.
When you are post-50, going back into "Start-Up" mode to build from scratch is brutal. But I am a massive Rocky fan. It felt like walking back into a freezing gym after years of success, lacing up the gloves, and getting back into training so we could run up those steps in Philadelphia one last time.
And if I am being completely honest? I was excited.
Deep down, I enjoy the struggle. I enjoy that raw, "fight or die" adrenaline that only hits when your back is firmly against the wall.
We threw out fifteen years of assumptions. We knew that just bolting a chatbot onto our legacy content library wasn't going to save the business.
So we had to architect something entirely different. A new kind of automation that forces AI to do the one thing it hates: stick to the pristine facts.
Today, as a two-person operation, that exact system allows us to execute at a speed that would have been impossible for our larger team just two years ago.
What follows is that story. Eight chapters. Each one covers a decision we had to make in real time, under pressure, with no roadmap. You will find the actual architecture, the actual mistakes, and the exact reasoning behind every choice.
Read it in order if you want the full picture. Or jump straight to whatever chapter title already made you wince.
The Series
Adapt or Die
We saw what was coming in November 2022 and still moved too slowly. Here are the six forces quietly dismantling every knowledge business - including yours.
The Decision
There are two ways to deploy AI in your business. One is fast and constrained. The other is hard and compounding. Most teams choose wrong - by default.
The AI Wrapper Trap
Our first instinct was to bolt a chatbot onto 14 years of training content. Here is exactly what happened - and why every business reaches for the same dangerous solution first.
The 4-Month Prototype
We thought automating content with AI would be fast. Instead, our manual proof-of-concept took four man-months of brutal work. Here is why.
The AI Migration Nightmare
The hardest part of AI has nothing to do with the model. Here is the brutal truth about your data - and why getting it clean is the most valuable work your team can do.
The Intelligence Engine
We threw the LMS model in the bin. Here is what we built from scratch, what it does, and why a 45-minute session now beats a four-hour course every time.
The Zero-Server Ecosystem
Why we threw away Microsoft Azure, built an orchestration layer to avoid OpenAI lock-in, and now run the entire engine with zero servers.
Where Do You Stand?
24 questions across eight Survival Rules. One honest score. Know exactly where your AI readiness gaps are - and what to fix first.
The 2026 Strategy Roadmap
See the precise timeline mapping our journey from a traditional LMS to an autonomous Intelligence Ecosystem.
I have already made the mistakes you are about to make.
Every decision in this series cost real time and real money to figure out. The data migration. The AI architecture. The choice of infrastructure. Hundreds of hours of building, breaking, rebuilding, and learning what actually works when you do this for real.
The result is a system that finds real problems, researches them intelligently, and produces targeted answers at speed. An AI-powered knowledge engine. Smart search. Automated workflows. All of it running on infrastructure fast enough to handle it without enterprise-level bills.
If your organisation is trying to move beyond the chatbot experiment and build something that actually works, I am happy to talk. This is paid consulting work, but we will both know quickly whether it is the right fit.
Book a call with Mark →No obligation. We'll work out in the first conversation whether this makes sense.