The winners will not be the companies with the most AI tools. They will be the companies that redesign execution so AI becomes part of the operating system.
By Nathan Rone — operator, ex-Amazon ($1B → $3.2B), six tech turnarounds, founder of One Advisory and Foresight.
I talk to companies every week who say they are “adopting AI.” Here is what that usually means:
They bought a few ChatGPT licenses. Someone built a demo. A team ran a pilot. There is a Slack channel called #ai-experiments where people share prompts. Maybe they hired a consultant who delivered a 40-page “AI readiness assessment” that sits in a shared drive.
None of that is AI adoption. That is AI tourism.
Real adoption is not about tools. It is about redesigning how the company actually operates — how information flows, how decisions get made, how meetings work, how institutional knowledge persists, and how one person creates leverage that used to require five.
Jack Dorsey wrote that hierarchy exists because humans were the only option for routing information. Managers summarize upward. Directors translate downward. VPs broker between departments. Every layer exists to compress and route context that the layer above cannot process directly.
AI changes that equation.
When the system can hold the full operating context of the company — what is moving, what is stuck, who owns what, which dependency claims are credible, what the customer actually needs — the coordination tax shrinks dramatically. The founder stops being the human router. The team stops rebuilding context from scratch every morning.
This is not about replacing people. It is about compressing the coordination overhead so the people you have can focus on judgment, relationships, and the work that actually requires a human in the room.
“Dorsey is building this for 12,000 people at Block. I deliver the same intelligence layer for teams of 5 to 50 — because the operating model shift works at every scale.”
— Nathan Rone
Not AI tools bolted onto old workflows. A redesigned operating system where AI is embedded into how the company thinks, decides, and executes.
A living model of the entire operation — what is moving, what is stalled, where dependencies are real versus claimed, who owns what, and what changed since yesterday. The founder stops being the human router for context.
What your customers actually need, built from real signal — not surveys and NPS scores. Support patterns, churn signals, deal velocity, and product usage assembled into a model of customer reality.
Escalation norms. Meeting standards. Decision-making frameworks. Risk tolerance. Communication preferences. The unwritten rules that make your company work — made explicit and applied consistently by the system.
Morning briefs that compress the operating truth. Meeting prep that arrives before the meeting. Decision framing that separates fast reversible actions from high-stakes commitments. Closeouts that preserve continuity.
Read-only observation. Draft-mode recommendations. Full autonomy within defined boundaries. The system earns trust incrementally — just like a new hire. Permissions, approvals, and audit trails keep it accountable.
Morning brief → execution → closeout → carry-forward. Every cycle enriches the context, sharpens the judgment, and builds institutional memory that persists across team changes and leadership shifts.
The teams getting 100x results and the teams getting marginal value are using the same models. The difference is never raw intelligence. It is the system that surrounds the model: what it knows about the operation, what rules constrain its behavior, how it routes decisions that require human judgment, and whether it remembers anything tomorrow.
Every serious AI implementation needs two layers working together:
Synthesis, ambiguity, context compression, judgment. Is this dependency credible? Is this meeting worth the time? What should the morning brief surface versus suppress? This is where the model creates value.
Rules, permissions, routing, approvals, audit, scheduling. When decisions are made, execution must be reliable. Permissions enforced. Actions logged. Data retention respected. This layer does not guess — it executes within boundaries.
Doctrine beats prompting. Prompting resets every session. Doctrine persists across every interaction, every day, every team member. The companies that encode their operating style into the AI system will compound advantage every week. The ones that keep prompting from scratch will stay on the treadmill.
There are five levels of AI maturity inside a services business. Every founder I talk to overstates the level they operate at. The cost of that mismatch is real, and it shows up in margin, retention, and team trust.
Top is not better than bottom. Each level wins somewhere. The goal is to match your level to your business and operate it cleanly.
Human judgment, end to end. AI is a sometimes-helper, not part of the operating model.
The founder uses AI inside their workflow. Drafts, code, research, summaries. Human signs every output.
Repeatable steps run by the system. AI drafts, humans approve. The first level with real operational leverage.
Agents complete work end-to-end inside defined boundaries. Humans handle exceptions and judgment calls.
The system runs the business. Founders set direction and intervene on exceptions. Services margin becomes software margin.
The goal is not to reach Level 5. The goal is to know which level fits your business right now, and stop pretending you are at a level you have not actually built the discipline to operate.
Most founders are performing one level above where they actually operate. The market is starting to notice. These are the patterns I see in the wild every week.
The tell: The pitch deck says “AI-powered agentic delivery.” The actual delivery is the founder + ChatGPT + a Notion doc.
The cost: Charging Level 4 prices, delivering Level 2 quality, eroding trust with every project.
The fix: Either build the infrastructure or rewrite the pitch.
The tell: “We do bespoke, high-touch work.” But the work is actually repeatable and the founder is just scared of systems.
The cost: Capacity ceiling, margin compression, watching Level 3 competitors take the market.
The fix: Audit which 40% of your work is actually repeatable and move it to Level 3. Keep premium for what truly requires judgment.
The tell: Real automation under the hood, but the founder still talks about “craftsmanship” and “boutique attention” because they are afraid clients will balk.
The cost: Capturing a third of the margin the infrastructure can actually deliver.
The fix: Reposition. The market will pay for outcomes, not opacity.
The tell: “We let the agent run it.” But there is no ops discipline, no observability, no eval pipeline.
The cost: Same-day revenue events when the agent breaks. Client trust hits that do not recover.
The fix: Go back to Level 3. Build the discipline. Then climb when the foundation holds.
The tell: Beautiful delivery system. Founder still personally chasing every lead and discounting under pressure.
The cost: The whole business runs at the speed of the founder’s worst sales week.
The fix: Sales is a workflow too. Treat it like one.
The ones who win the next 24 months will not be the ones who climbed fastest.
They will be the ones who matched their level to their business and operated it cleanly.
Not advice from a distance. Hands-on operating model redesign with production systems that ship.
Where do you actually operate versus where do you say you operate? Which workflows are at which level? Where is the highest-cost mismatch costing you margin, retention, or team trust right now?
Not everything should change at once. Find the workflows where moving up one level creates the most leverage fastest: morning operations, meeting cadence, status reporting, dependency management, decision framing, client-facing prep.
Encode your operating style: escalation rules, meeting standards, decision frameworks, approval boundaries, risk tolerance, communication norms. Strong defaults first, your calibration on top. This is what makes AI feel like your company, not a generic assistant.
Production systems, not prototypes. Execution surfaces, intelligence layers, governance controls, and the daily operating loop that makes it all compound. I ship what I design — and I stand behind the results.
The operating model compounds over time. Monthly optimization, new workflow rollout, doctrine refinement, and quarterly reviews. As Fractional CAIO, I operate the AI layer as if it were my own company — because my compensation depends on the results.
The market has plenty of AI talkers and AI builders. Very few people can bridge operator judgment, AI architecture, and production implementation.
Amazon org growth. Built AI-powered systems inside one of the most demanding operating cultures on earth.
Cbeyond, Birch, Xspedius, Airband, Alpheus, Windstream. Operator experience across companies navigating chaos and reinvention.
I run autonomous AI systems in production today. Foresight is not a concept — it is a live execution system with paying customers.
Most advisors can talk strategy or build prototypes. Very few can redesign the operating model and ship production systems. I do both — and I bet my compensation on the outcome.
Three paths depending on where you are. All lead to the same place: AI as an operating layer, not a toy.
Two weeks. I map your operation, identify your real level, name your costliest mismatch, and hand you the playbook.
$5,000
I become your AI operations partner. Operating model redesign, system deployment, monthly optimization.
$20,000/mo
The productized execution layer. Morning brief, execution health, meeting intelligence, continuity. From $149/mo.
$149 – $1,495/mo