The Adaptive Operating System

A Practical Path to Operating Model Compression

Hi there,

Today, I’m writing more about a concept I’ve previously explored: operating model compression and its evolution towards an “adaptive” operating system. The benefit of AI technology is leverage, and business owners have a once-in-a-generation opportunity to rethink many of their processes from scratch.

The Performance Gap

The economics tell a story that should keep traditional executives up at night: while OpenAI, Apple, and Meta are generating over $2 million in revenue per employee, the average Fortune 500 company limps along at $642,000, and though industries naturally differ in their capital intensity, this chasm has nothing to do with how many hours people work and everything to do with how that work flows through the organization.

The gap keeps widening because new AI companies are pulling away from traditional tech leaders by automating significant portions of production and review work that previously required armies of coordinators, reviewers, and managers to keep the machine running.

I call this Operating Model Compression: radically reducing the complexity of your entire operating model to meet the moment that AI technology presents us with. It means connecting your systems so they can communicate with each other via API, capturing data from every interaction to improve tomorrow's decisions automatically, and automating routine decisions to software, allowing humans to focus on problems that require judgment and creativity.

This isn't just about cutting jobs, it's about getting people out of endless email chains and status meetings and into the kind of complex, creative work that initially excited them about their careers.

Consider an insurance claims adjuster who currently spends most of their day copying information between three systems that should have been integrated years ago, and now imagine that same person investigating complex fraud patterns that save the company millions: same person, completely different value, and probably a lot happier at the end of the day.

Most companies struggle with this shift because, as research consistently shows, fewer than one in three digital transformations deliver lasting change, and the reason is almost always the same: companies try to digitise broken processes instead of fixing them first, as if adding technology to chaos somehow creates order.

What Actually Works

The companies that get this right don't chase every new technology trend; instead, they focus relentlessly on three fundamentals that sound simple but prove surprisingly difficult to execute.

First, they connect their systems, which sounds embarrassingly basic until you realise that most large companies are running nearly 900 different applications, with less than a third talking to each other. This creates an endless cycle of copy-paste and email coordination that would be comedic if it weren't so expensive.

Second, they turn every customer interaction into data that informs tomorrow's decisions better - not vanity metrics for quarterly presentations or reports that sit unread in someone's inbox, but living data that actively improves operations every single day without requiring any additional effort.

Third, they let software handle routine decisions while maintaining clear boundaries about what machines can decide and what requires human judgment, because most organisations dabble with AI as if it were a science project rather than fundamentally rethinking how decisions flow through their business.

Companies That Made It Work

Netflix has essentially become an algorithm with a content library attached, where the vast majority of viewing comes not from users browsing through endless titles but from recommendations so precisely tuned that they've turned binge-watching from a guilty pleasure into a business model, all while deploying code hundreds of times daily in a relentless cycle of optimization that killed their profitable DVD business without hesitation because they saw where the world was heading.

Luckin Coffee built their entire business model around the assumption that most coffee purchases are routine and predictable, so they created a mobile-first experience where ordering ahead, paying digitally, and grabbing your drink without speaking to anyone isn't just an option - it's the whole point, allowing them to run 22,000 stores with minimal staff because the app handles the complexity that would typically require three times as many people.

The frontier AI labs are the most interesting case because they're using their tools to build better tools, creating a recursive loop of improvement. Their coding assistants, such as Claude Code, Gemini, and Codex, help build and train the next generation of models. These models learn from millions of user interactions to continuously refine not just their products but also their entire approach to building and deploying AI systems.

Your 30-Day Challenge

Here’s a 30 day challenge for business owners and executives thinking about implementing AI in their business: pick one process that makes everyone groan when they hear about it, whether it's processing orders that somehow still takes three days, handling returns that require five different approvals, or approving expenses through a system that seems designed to punish people for spending company money on company business.

Week 1: Watch and Learn
Before you make any changes, follow 10 examples through your current process from start to finish, counting the number of people who touch each one and timing how long they spend on it. I guarantee you'll discover that what you think happens and what happens are two very different stories. Ask your team what drives them crazy about this process, and then listen - listen, because they've been thinking about how to fix it for years, but nobody ever asked.

Weeks 2-3: Fix Something
Start by tackling the most obvious problem, such as data that gets copied between systems that should have been integrated years ago, or decisions that someone makes fifty times a day using the same criteria that could be written down in five minutes and done with the help of a custom GPT, N8N workflow or Gemini Gem. Find a few volunteers who are frustrated enough to try something new, while everyone else continues with the old way. Forcing change on people who don't want it is the fastest way to guarantee failure.

Week 4: Check Your Work
Compare the new approach to the old using the metrics that matter most to your business: Is it faster? Does it require fewer people to touch it? Do customers complain less? Are employees less likely to quit? If the answer is yes to most of these, expand carefully to more people; if you're close but not quite there, adjust based on what you've learned; and if it fails, congratulations - you've learned what doesn't work without betting the company on it.

Why Most Companies Fail This

Transformation projects fail for boring, predictable reasons that everyone knows but somehow ignores: leadership treats it as an IT project when it's really about changing how and why work gets done, the CEO delegates ownership to someone two levels down who lacks the authority to make fundamental changes, and everyone expects revolutionary results in three months when the reality is that meaningful change takes years of consistent effort.

The worst mistake, and it's so common it's almost a cliché, is automating chaos - taking a process that's already broken and making it fail faster with more expensive technology, as if bad data somehow improves when you move it to the cloud or broken workflows magically heal themselves when you add AI to the mix.

Watch for these warning signs that you're about to waste a lot of money: different teams handling the same task in completely different ways because no one ever standardized anything, rework rates above ten percent because the process breaks so often that fixing it has become someone's full-time job, and procedures that exist in beautiful documentation that no one has followed since the day they were written.

If your business can’t take a small project from idea to deployment in 30 days, you have an even bigger problem, and the emergency sirens should be flashing red. You’re in an all-hands-to-battle-stations emergency and urgently need to streamline processes and controls to enable flexibility and innovation.

This mindset issue is a significant problem I keep hearing about and seeing. Every business in every industry needs to rapidly accelerate its pace so that it can shorten feedback loops between experiments and learn the lessons of what does and doesn’t work in its business. There isn’t time to waste in multi-month project approval processes anymore unless you’re happy to lose market share and margin to competitors.

Building Your System

You need three layers, and while none of them are technically complex, getting them right requires the kind of sustained attention that most organisations struggle to maintain. Again, operating model compression in the AI era is about rethinking how problems are solved.

Everything that can be automated and integrated should be, allowing your teams to focus on more complex problems and interact with customers to better understand their needs.

The connectivity layer is essentially plumbing: it defines how systems communicate with each other, what data flows where, and who can access what. However, most companies skip this foundational work and then wonder why their digital transformation feels like pushing water uphill.

The memory layer captures what happens in your business, not for compliance reports that no one reads or dashboards that everyone ignores, but to create a living memory that learns from every interaction and makes tomorrow's operations slightly better than today's without requiring constant human intervention.

The decision layer starts with encoding the obvious decisions that humans make over and over (if the customer has been with us for five years and the amount is under $500, approve the refund) then gradually adds intelligence for edge cases while maintaining clear boundaries about what requires human judgment, because the goal isn't to eliminate people but to free them from decisions that don't require thought.

The Clock Is Ticking

Here's the uncomfortable reality that most executives don't want to hear: companies that build self-running operations will compete against companies that don't, and the difference between generating $2 million per employee and $642,000 per employee isn't a gap that traditional efficiency improvements can close. It's a fundamental difference in how work happens.

Based on current adoption rates and the speed at which these AI technologies are improving, organisations may have only five years before Operating Model Compression becomes a table-stakes requirement, just as having a website went from a competitive advantage to a basic requirement in a decade.

The companies moving now aren't waiting for perfect conditions or complete certainty; they're moving because they understand that the alternative is competing against businesses that require one person for work that takes their competitors five times as long, and that math doesn't work, no matter how talented their people are.

Operating Model Compression isn't just about reducing headcount; it's about minimising the number of times humans have to interact with something before it's complete, thereby eliminating the handoffs and coordination that consume most of everyone's day. You can select one process where you can measure both the number of touches and the total time from start to finish, and then see if you can cut both by a third without compromising quality or making customers unhappy.

Success doesn't require perfect technology or massive budgets; it requires measuring what happens in your business, rather than what you think happens, fixing the broken parts before automating them, and documenting everything. This is because automated decisions need to be explainable when something goes wrong, and something will inevitably go wrong.

Start with one process that would prove to the sceptics that fundamental change is possible in your organisation, measure it, and automate what works. When people see it working, they'll stop resisting and start asking when they can improve their process. That's how you build an Adaptive Operating System: one success at a time, with momentum building from the ground up rather than being forced from the top down.

How I Can Help

If you’re ready to move from an AI transformation idea to executing on an AI transformation in your business, but need help working through the challenges you face, book a complimentary 30 minute call with me.

Regards,

Brennan

PS: In case you missed it: