The night I built my own CRM
One evening. That's all it took. I built a CRM. I connected every tool I use, day to day, into one workflow. I automated the whole thing in Python and HTML. Code I had never touched before that day. I built an HTML dashboard that refreshed the data in real time.
I'm not a developer. I'm a transformation consultant. Thirty years in this industry, and I built that in an evening. I’ve never seen tech. change this simple and fast. That's not a party trick. That's what one person, with no engineering background, can now do, alone, with almost no support.
Hold onto that thought. We're coming back to it.
Reframe the problem
In thirty years of transformation work, I've used the same six levers of change over and over again:
Customers and Markets.
Products and Services.
Process.
Technology.
Structures and Facilities.
People and Rewards.
Pull one, and you need to rebalance the others too. Get the right combination, and the organisation moves.
You'd expect me to tell you AI transformation is about people, not technology. It's the safest sentence in change management today. But I don't think it's quite true. Not this time. In thirty years, I've never seen one lever move entirely on its own the way Technology is moving right now. It's not waiting for the other five. You've felt it yourselves this year, I'd guess:
“The tech isn't struggling to keep up with the organisation anymore.
Organisation are struggling to keep up with the tech.”
So no, this round, it really is about the technology. But that doesn’t remove the people question, it changes it.
It used to be: get people to adopt the technology. That's not the question anymore - people are adopting it faster than anyone asked them to.
The job question is about judgement -deciding where this power should go, before it decides for you. The bottleneck didn't disappear. It moved upstream.
The risk - doing nothing is no longer an option
Here's why that matters more than it sounds. You've got people inside your organisation right now, moving faster than anyone is directing them, on a technology that still hallucinates, still trips over data sensitivity, still gets things confidently wrong. Power without direction, on a tool that isn't fully trustworthy yet. That's a risk-based storm, and it's already inside your building. Which is why doing nothing isn't the cautious option. It's not an option at all. Every week you wait to work out where this power should go, someone in your organisation has already pointed it somewhere without you.
The noise
Where do you start? Ask the AI experts, and you'll get an answer. You'll just get a different one depending on who you ask.
Go all in, one camp says.
Test and learn, says another.
AI is a tool, nothing more, says one.
Others will tell you it's the biggest inflection point of our lifetime.
Some say the bottleneck is people - skills, governance, judgement.
Others say no, it's not about people at all anymore.
Focus on agentic workflows, not the biggest model, say some.
The hype is years ahead of the reality, real adoption is a decade away, not a quarter, say others.
Augment, don't replace, says nearly everyone, until the next voice tells you that's exactly the trap, and you need to let humans be human and machines be machines, and design the roles apart from the start.
Notice something. They contradict each other. And a good number of them have something to sell you. So before you borrow anyone's conviction, mine included, ask what their interest is.
What a transformation expert says
Strip the noise away, and there are really only two ways into transformation:
You can start from a blank sheet, building as if none of your existing operating model exists. That's how the digital challenger banks won - no legacy to work around.
Or you can hotspot. Take what you already do, overlay it, and find where AI genuinely earns its place. Not everywhere. Deliberately.
Right now, in the test-and-learn phase most organisations are in, blank sheet isn't the right call for most of you reading this. Hotspot is.
Here's the filter I use to find the hotspots: don't automate what you can already code. If it's a decision tree (yes or no, this rule or that rule) that's not AI, that's programming. You don’t need a language model for it, you needed an engineer, ten years ago.
AI earns its place somewhere else: large volumes of information, multiple sources to pull together, complex rules to apply, judgement to support, not replace.
The evidence
Let me show you what that actually looks like. I've been working with a bank. We used AI to pull together the full history of every conversation a customer has had with them, every call, every note, apply the relevant rules, run the analytics on their income and earnings, and hand the adviser a recommendation.
The AI did the synthesis. The crunching. The application of the rules.
The human did the conversation. The empathy. The judgement call on what to actually say to someone, and how.
The AI then recorded it and executed the actions.
That's the hotspot method and the augment-don't-replace principle, doing their jobs, at the same time, on a real customer.
What to actually do
So here's what I'd tell you to do, starting Monday.
Test and learn properly. Not a pilot that quietly disappears. Share what you find, even the failures, so the next team doesn't waste six weeks rediscovering it.
Build a genuinely safe space to experiment - time, tools, and permission. People are going to try this with or without you. Better they do it somewhere you can see.
Pick a bounded problem, deliberately. Not a lab with no edges. A specific use case, chosen on purpose.
Design the human checkpoint in before you go live - not after something's gone wrong. Decide now where a person sees the output and makes the call.
Make the feedback loop a deliverable, not an afterthought. Name who owns folding what you learn back into how the organisation actually runs. If nobody owns it, it won't happen.
Don't expect a straight line. Nothing happens for weeks. Then a breakthrough. Then nothing again. That breaks the three-year roadmap and the stage-gate plan completely, so stop planning like it won't.
I first wrote about this in October 2024 here and offered a very similar checklist 8 steps for the adoption of AI in business transformation.
Close
Let's go back to where we started. An evening, building my own CRM. Code I'd never touched. Six levers, and for the first time in my career, one of them is moving almost entirely on its own. This time, it really is about the technology. But not the way you'd expect.
The lever moved. The question moved too.
It's not "will people adopt this." They already have, with or without your permission.
The question is who's deciding where that power goes. That's not a technology question. It never was. It was always going to be about people (just not the people question anyone was expecting.) Which is why doing nothing isn't the safe option. It's the one you don't actually have. And here's what I'll leave you with. AI's graveyard isn't full of failed technology. It's full of pilots that got put into practice, that impressed the room, got the applause, got the case study and never made it into how the organisation actually runs.
Don't build another headstone. Augment the operating model instead.
Paul Cook is The Transformation Guide. If this raised more questions than it answered, get in touch or book a call.