I’m writing this for forward-thinking CEOs who, like me, are feeling a bit overwhelmed by the rapid evolution of AI capabilities over the past six weeks — and wondering what it all means for your organization.

Here’s what I’m going to do: paint a high-level picture of the current enterprise AI landscape, and then give you a ground-truth framework for how to redesign your business to take advantage of these new opportunities and avoid the real threats. Let’s dive in.

The State of Enterprise AI

Starting in January 2026, things went from high-potential to high-actuality in a matter of weeks.

From the top down, the latest large language models — Opus 4.6, Gemini 3.5, Codex 5.3 (with Grok 5.0 on the way) — are all replicating human knowledge work at expert level for near zero cost. All of these companies now offer powerful multi-agent workforces with context and orchestration management built in. If you can design your business to take advantage of it, you can produce code, ads, analysis, workstreams, and customer interactions in a fraction of the time and cost you could before. Each vendor has their own flavor but all of these models are exceptional — and getting recursively better.

From the bottom up, an army of solopreneurs and startups are leveraging open agent models — like Clawdbot/Open Claw, with dozens more bursting onto the scene. They’re deploying virtual servers and local Mac Studios, Mac Minis, and Raspberry Pis to run N+1 agent workforces, often on Chinese open-source models. The scale is different but the outcome is the same: expert-level knowledge work at near zero cost.

From the side, a decimation of the B2B SaaS industry is underway. Major players — Microsoft, Monday, Salesforce, Atlassian, Adobe — are all scrambling to accelerate adoption of their own version of enterprise AI capabilities, and all are under intense valuation pressure from the large LLM providers above. Will they have to drop their per-seat pricing? Or will the fact that they “manage the mess” keep them relevant at 80% gross margins? We’ll soon find out. And filling every remaining gap, a wave of startups offering enterprise AI wrappers and consulting — most selling early capabilities more than a fully-fledged product yet.

Here you sit — in the center of it all.

If I were you, I’d be asking: “OK, what does all this mean for my business? There’s an infinite number of new AI tools and capabilities. An infinite amount of AI slop gumming up our go-to-market channels. An infinite amount of increasing pace to keep track of, integrate, or ignore. How do I lead my team — and organize my agents — to take advantage of it?”

I have an answer. Not that I know what you should do with your particular business without talking to you first — but I know the foundational pieces you need to have in place. This is true whether you’re Anthropic, a Chinese startup, or a more traditional company seeking to delight its customers no matter what.

There are three:

  1. Flatten your structure
  2. Build a system for fission-fusion
  3. Build and protect your proprietary context layer

There are many things you can do in this environment. What I’m telling you is that if you want to win, you must do these three. Here’s what each one means and why it’s essential to your success.

Flatten Your Structure

I’ve been writing and coaching about the beauty and functionality of flat hierarchies for over fifteen years. If you’re a solopreneur or a small team running an AI-native company, you don’t have this problem. Your structure is already flat. It’s you — or you and a few key leaders — and a team of agents. Congrats! This structure gives you very high velocity, and it will be hard for an incumbent stuck in an outdated structure to counterpunch. Expect a buyout offer. But then prime yourself for frustration and disappointment when they try to integrate you into their skeleton.

If you’re a more traditional business — and you’ve organized around people, span-breakers, or reporting lines — you need to adjust fast. Flatten. Flatten. Flatten. You don’t need layers of reporting, even without AI. You don’t need span-breakers — in fact, span-breakers can be very dangerous to organizational health. You don’t need to run your company around titles, one-on-one meetings, or other nonsense. Design for the strategy. Strip away competing accountabilities from both humans and agents. Build in total operational transparency and real-time feedback. Focus on delighting the customer, now and over time. Everything else is a distraction. True, some of your team clings to these distractions as if their careers depended on them — but that doesn’t make them any less of a drag. Unhook the anchors and design your business for throughput and velocity. Which means flat!

Here’s what this looks like in practice. Below is a simplified before-and-after from one of my coaching clients — a company that went from a traditional layered hierarchy to a flat, strategy-driven structure. The before image shows what most traditional companies still look like: layers of reporting, competing accountabilities, and span-breakers slowing everything down. The after shows what happens when you design around the work that actually matters. The difference in speed and innovation is felt in weeks, not months.

 

 

If you’d like to see how to design a flat structure based on first principles, here you go: Designed to Scale Structure Change Videos.

Build a System of Fission and Fusion

Here’s an image of an Agent Mission Control Dashboard built by X user @pbteja1998, a small business owner. According to his post, he built this dashboard in a few hours using Open Claw. You can feel his enthusiasm for his newfound ability to manage multiple agents — an agent swarm — for his business.

Now, imagine that you are this solopreneur. Where is the next bottleneck going to emerge? I can’t say exactly, but I know one will emerge. That’s the nature of a bottleneck — the area of the business with the least throughput that prevents the rest of the business from moving forward.

There’s always a bottleneck somewhere. It will likely emerge in this example when navigating the nuances and capabilities of different models and agents doing different things. Just like people, agents forget. Agents lose context. Agents have bad days. Agents lose the plot line. Agents need to be corrected. Funnily enough, running a team of agents as a solopreneur is not too dissimilar from running a large company with a human leadership team. As things grow in complexity, there is always a complexity tax. Your job is to see ahead and prevent the complexity tax from costing you more than the gains.

The good news is that over billions of years, nature has evolved a way to manage complexity. You’ll see this system at work in every complex adaptive system, from a cell to a country. It’s how your brain works. It’s how your heart works. It’s how your company works — if it’s healthy. The system is called fission-fusion, and a nod here to one of my mentors, Howard Bloom.

Fission breaks things apart. Fusion brings things together. Your business, no matter its size, needs a regular cadence of bringing agents and humans together to synthesize the latest signals from the environment (fusion), make a smart decision, and then break apart again (fission) to execute with renewed shared consciousness.

The easiest way to conceptualize this pattern is to think of a bee hive. Every agent has a job to perform. There’s even a subset of lazy agents who just explore the environment. On a regular cadence, all the agents come back to the hive and share what they’ve learned. They bring evidence — pollen, water, information. The hive engages in a vigorous decision-making process until enough critical mass is gathered to decide and take action. Then every agent goes forth to do their job within this new context.

That pattern must exist in your agent swarm. It must exist in your leadership team. It must exist in your departments. Come together, synthesize, decide, act independently. Swiftly.

At Organizational Physics, in our Strategic Execution Team (SET) meeting process, I teach a simple method to create a “heat map” to enable fission-fusion. The heat map is an aggregation of the friction points, entropy, or issues that have arisen over the past week. We don’t try to classify the heat as it shows up during the week — we just track that it’s emerging.

We use two heat maps. One is for annual entropy reduction, called an Entropy Survey. The other is weekly, called the Open Hot Items to Allocate list. You can come up with your own designs and terminology, you could potentially increase the pace, but the important thing is that you’re not mixing short-range and longer-range into the same cadence.

Entropy Survey: Use for longer-range entropy reduction: Scan for persistent issues harming operational excellence and attack them as part of your strategic planning.
Open Hot Items to Allocate: Use for short-range entropy reduction: Don’t treat all issues the same. Allocate each hot item based on how it should be managed.

For short-range entropy reduction, during the SET meeting, the CEO — and it’s critical that one role, the highest authority on the team, owns this — decides how to allocate each hot item. It goes into one of three buckets:

  1. Not Going to Track It — we recognize there’s some heat here but it’s not a priority now. We don’t spend a dollar to save ninety-nine cents.
  2. Action Item to Assign — this is an important item. It goes into our centralized action plan where it will have an implementer (agent or human) who accepts it, a clear description of what we want to occur, a deadline, and a support team (agents and/or humans).
  3. Context Capture — here is new shared context that we all need to better understand and track, but it’s not a decision to make.

I’m seeing a lot of solopreneurs vote to skip meetings altogether and just run an asynchronous process of constant hot item allocation. I think they’ll find that constant short-range allocation without building shared context across the team leads to myopia and hits diminishing returns fast. There’s a reason that systems which take a rest, gather together to share information, and then decide as a team (not a democracy) tend to outperform systems that don’t.

If you’d like to see how to run a fission-fusion process in your own business, here it is: The Organizational Physics Strategic Execution Team (SET) process. Fair warning — this is one of those things that’s simple to understand and hard to implement without guidance. If you want help, let’s talk.

Build and Protect Your Proprietary Context Layer

Do you know what truly makes a business valuable? It’s actually hard to quantify. What makes Starbucks so successful? Is it the brand? Timing? Reach? Supply chain? Culture? Access to capital? It competes with a million and one other coffee shops, so what makes it so unique and valuable? How about Walmart? Anthropic? Apple? SpaceX?

The irony is that what makes these businesses thrive can’t actually be captured on the balance sheet. The outputs are captured, but not the inputs. Even though you operate in a different segment, your business is fundamentally the same. It has a unique set of worldviews, data, approaches, and capabilities — an amalgamation of the experience, knowledge, skills, and vision of the people involved, especially the early founders and leaders. That’s your real leverage point.

When it comes to AI, there’s infinite raw intelligence available. But it needs to be harnessed for your particular business strategy, stage, culture, and approach. The unique flavor of what you’ve experienced, what you know, and how you approach problems and opportunities — that’s your competitive advantage. It should be hard for other firms to replicate.

Let me give you an example. Say you’re a solopreneur running a team of agents. You can quickly spin out new digital tools for others to use. Well, guess what — the moment one of your tools gets any traction at all, you’re going to see a flood of “me-too” competitors, armed with their own agents, spin up cheaper and more capable versions. There will be enormous market noise, and it will be hard for most companies to break through.

Those that do break through — and do it sustainably — will be those who have developed an approach, a way of thinking about and pursuing things, that is unique to them. It takes advantage of their knowledge, experience, connections, culture, and vision, as well as their organizational structure and the capabilities of their agents and teams.

Your first challenge is to codify — for your human team and your agents — what it is that makes your approach different, valuable, and more adaptable than your competitors. In practice, this means codifying your decision-making frameworks, your institutional knowledge about what works and what doesn’t in your market, your playbooks, your customer insights, and your unique point of view — and structuring them so your agents can access and act on them in real time, not just your senior team over coffee. Think of it as building the operating manual for how your company thinks. Without it, your agents are just general-purpose intelligence. With it, they’re yours.

Where you store your context layer matters. The most defensible approach is to keep your proprietary knowledge local — on your own infrastructure — and feed it to models at query time rather than handing it over permanently. The major AI providers all offer enterprise agreements that protect your data, but the companies with the deepest moats will be the ones that maintain ownership and control of their context, not just contractual promises about it.

If you think forward a few years, in the ever-relentless pursuit of more data and more know-how from insatiable AIs, your company’s proprietary context layer may be its last and most valuable asset. You can hold on to it for your own competitive advantage, or license it to other models. If that sounds crazy or hard to follow — well, I get it. We are living in very interesting times.

The bottom line is that how you approach things may be as valuable as what you do.

Summary

Things are crazy out there. If they don’t seem crazy, well, you’re not paying close enough attention. We’re in a new world of recursive superintelligence. The models are now building themselves, better and faster than any human team could. While the inertia of the status quo is real, there’s going to be a reckoning. While I don’t have a crystal ball, I do know that these three things — flatten your structure, build a fission-fusion process into it, and develop and defend your context layer — are foundational. They’ll help you adapt to whatever comes next. And if you don’t have them, all your other efforts will be folly. The companies that thrive won’t be the ones with the best AI tools. They’ll be the ones with the best structure, the best decision-making rhythm, and the deepest proprietary context. That’s the physics of it.

📌 If you’re a CEO running a $10M-$100M company and this resonates, I’m taking on a small number of new coaching clients to work through this transition. Here’s how to start a conversation.