Operating Model Compression: A 2025 AI Year in Review

I thought frameworks would be enough. I was wrong. Here is a review of my 16 posts that mattered most in 2025

Hi there,

It’s the end of the year and I wanted to look back at what I’ve been writing about since February. I left my corporate job, relaunched this newsletter, and spent most of 2025 trying to figure out what AI actually means for how businesses run.

What follows is a tour through the sixteen posts that mattered most. They build on each other, even though I didn’t plan it that way at the time.

Starting Over

In February I wrote Back in Business to explain why I was relaunching this newsletter. I’d run an ESG research publication during COVID but the AI story had become too big to ignore.

A few weeks later I published One Month On From Leaving A Corporate Job. Twelve years in financial services and then suddenly not. But I was also spending hours each day inside AI tools, building things I couldn’t have built before, and that changed how I saw everything. The skills I thought were becoming obsolete turned out to be exactly what the moment needed: change management and understanding how big organisations actually work.

Getting Into The Weeds

By April I was deep into AI 2027 and Vibe Coding. A group of researchers had published a detailed scenario for how AI development might unfold through 2027, and I wanted to stress test it against what I was actually seeing. I’d been using agentic coding tools to build side projects. These are AI interfaces that can write, test, and iterate on code based on natural language instructions. They were genuinely good. Not gimmicks. Real ways to build things even if you’re not a developer.

That same month I wrote Problem Solving With Time vs Tokens. You can now solve problems by paying for human time or paying for AI tokens. Once you see business through that lens you start noticing opportunities everywhere. But you also start seeing which human skills still matter and which ones are being priced out.

May brought The Whiteboard Test. I was getting frustrated with companies doing AI pilots that never went anywhere. So I wrote a simple diagnostic. Can you draw your business on a whiteboard? Can you point to where AI should fit? Clarity precedes automation.

The Big Idea

Then came the post I’m probably most known for now: Why Every Business Has 5 Years to Achieve 90% Operating Model Compression.

Operating model compression is the idea that AI-first companies will radically shrink the people, process, and technology required to deliver value. Not 10% efficiency gains. 90% or more, starting with back-office functions and customer service.

I know it sounds aggressive. But the math is the math. If you increase productivity 10x you only need 10% of your original resource footprint. This won’t happen everywhere at once, and regulated industries will move slower, but the direction is clear.

A week later I published AI, APIs, and the End of the Firm as We Know It. If agents are talking to agents and APIs are connecting everything, what’s left of the traditional organisation?

The Triple Boundary Framework gave people a practical tool for thinking about where AI changes their business: the boundaries of capability, learning, and coherence. The post landed well because it answered questions people were actually asking.

Making It Real

August was about getting practical. What GPT-5 Really Means For AI Transformation argued that most analysts were missing the point. Technical benchmarks don’t drive adoption. Non-expert perceptions do. How your average employee sees these tools matters more than what the leaderboards show.

Your Data Is Ready for AI pushed back on the “we need to fix our data first” excuse I kept hearing. Most organisations already have better data than they realise. The bottleneck is usually somewhere else: culture, capability, or willingness to actually change.

The Adaptive Operating System extended operating model compression into something more practical. Rather than ripping everything out and starting again, I outlined how to evolve towards a system that learns and adjusts. It’s a path, not a cliff.

The Human Bit

I spent years doing project and change management before all this. That experience kept coming back.

The Human Side of AI was about why bringing your team along isn’t optional. I’ve watched too many technology projects fail because nobody thought about how people would actually experience the change. AI makes this worse, not better. The fear is real.

When AI Does Your Job Faster tackled the elephant in the room. A huge chunk of employees secretly use AI at work. Many others are in denial about what’s coming. Both responses are destroying transformation efforts.

Why Your Smartest People Are Your Biggest Problem came from watching high performers become blockers. AI threatens their identity, their value, their entire professional framework. Smart people need smart change strategies.

The Monday Morning AI Test offered a simple starting point. Tomorrow morning, ask five employees what wastes their time. That’s where your AI should start. Not another pilot. A real solution to a real problem.

Where It All Lands

I finished the year with Your IT Policy Is About to Become a Talent Problem.

The gap isn’t between people who use AI and people who don’t. Most knowledge workers have typed something into a chat interface by now. That’s table stakes.

The gap is between people who can install tools on their machines and people who can’t. And most corporate workers can’t.

The tools doing the heavy lifting require installation. They require permissions most workers don’t have. A browser-based chat window is safe. It’s also limited. A tool that can read your files, search the web, execute code, and chain together complex tasks is powerful. It also needs permissions that most IT policies don’t grant.

This is where all the year’s thinking came together for me. The experiential gap I kept writing about isn’t just a mindset problem. It’s a permissions problem. The organisations that figure out how to give their people access to the tools that matter, with sensible guardrails, will pull ahead. The ones that default to prohibition will wonder why their people seem slower than everyone else’s.

What I Learned

I thought the frameworks would be enough. Write about operating model compression, give people the Triple Boundary Framework, and they’d have what they needed to move.

They don’t. The gap isn’t conceptual. It’s imaginative.

The way to win at AI transformation is to rethink processes end-to-end without reference to the existing people, processes, platforms, and operating model constraints. But if you haven’t used the tools yourself, your imagination won’t be rich enough to figure this out. This isn’t digital transformation where you optimise what exists. It’s a total remix of value delivery.

The organisations that are stuck have a common pattern: decisions about AI happen at the top, far from where the actual work gets done. Executives commission pilots. Consultants write strategies. IT blocks the tools that would make a difference. And the people who know where the friction actually is never get their hands on anything that matters.

The organisations making progress look different. They put tools in the hands of people who understand the problems. They create permission to experiment. They build feedback loops so that what gets learned at the edges flows back to the centre. And critically, the leaders have spent enough time in the tools themselves to imagine what’s actually possible.

What’s Coming in 2026

Leaders need to get hands on with AI tools. Not to do the work themselves, but to expand their imagination for what’s possible.

Most leaders I talk to are still thinking in terms of efficiency gains. Ten percent here, twenty percent there. They’re asking “how do we do what we do, but faster?” That’s the wrong question. The right question is “if we were starting from scratch with these tools, how would we deliver value to customers?”

You can’t answer that question from a slide deck. You have to feel what’s possible. And you can only feel it by using the tools yourself.

2026 will be the year this becomes harder to ignore. The organisations that move fastest will be the ones where leadership has spent enough time in these tools to reimagine their business without the constraints of how it currently runs.

I’ll be writing about how to do that. And if you’re wrestling with AI transformation in your own organisation, I do one-hour AI change leadership sessions where we work through your specific blockers together. The goal isn’t more frameworks. It’s helping you see what’s actually possible so you can lead the change.

See you in 2026.

Best regards,

Brennan