Product Leadership · AI-Native
Most product leaders use AI. I build the operating layer an organisation runs on, ship AI-native products into market, and — the part almost everyone skips — get the capability adopted so it sticks after I leave the room.
Most companies treat AI like a rollout — a platform, a policy, a training plan, a login count. It fails, because AI doesn't change what people do so much as what they're capable of — and capability spreads socially, through people. The work is tending an ecology, not shipping a deployment.
The difference between AI that looks adopted and AI that actually changes the business.
| AI as rollout | AI as ecology | |
|---|---|---|
| The mental model | AI as a software deployment | AI as a capability that spreads socially |
| Success metric | Licences, logins, training completed | Changed work and commercial outcomes |
| The missing role | No owner at the people/tools/commercial junction | A gardener — usually product & tech leadership |
| How it spreads | Top-down policy and mandate | Move what people find; protect the risk-takers |
| What's left behind | A dashboard and unchanged habits | A team that keeps moving after I leave |
The operating layer
Original, published IP on how AI actually gets adopted and operated — the system, not a tactic.
A reframe of AI adoption from rollout to ecology. Companies treat AI like a software deployment — pick a platform, write a policy, count logins — and it fails, because capability spreads socially through people, not through a deployment plan. It names the missing role most orgs lack: the gardener.
Built on Howkins' Creative Ecologies and Snowden's Cynefin. Published via the Product Leaders Substack.
Read: AI adoption is not a rollout →The operating system that turns product strategy into revenue — built as a loop with a spine, not a ladder. Five stages move signal through the organisation; two more form the spine: the operating rhythm and the commercial seat.
The infrastructure inside which the adoption ecology is tended. A complete operating model, not a tactic.
Read: how to build a Product OS →Proof he ships
14+ products shipped across AI, LLMs, agentic systems, ML forecasting, NLP and voice — 0-1 and category-defining, not features bolted on.
Rebuilt the product around an AI Decision & Data Intelligence Layer — an 'agent at the core' that watches industrial signals, detects scenarios the customer never defined, and pushes decision recommendations. Shaped a RAG / context-aware architecture with an explainability layer as the moat, and secured a tier-one market-data partnership within months.
Shaped an AI-native product 0-1 for a profitable US services firm moving into software: 'the implementation intelligence layer that makes HCM delivery predictable, measurable and trusted worldwide.' Defined GTM positioning and delivered first-time product–market fit.
Ran discovery for an AI/LLM-powered GRC product to unlock a new category in SME cybersecurity — built on owning the specific workflows compliance officers and MSPs actually run, not generic AI features.
Built WFS's first data science capability alongside a new ML forecasting product. The ML scheduling product grew from 10% to 50% of total sales pipeline and supported additional PE investment.
Proof he builds
Working AI systems I configured and operate myself — including AI-native products I build for my own and customers' use.
A 7-agent software delivery team built in Claude Code (Anthropic's Fable model) — product-architect, frontend, backend, ai-engineer, ux, qa and code-reviewer — that lets a non-coding CPO ship production software. Agents run in parallel with self-verification, so I review finished work, not every step.
Reusable AI workflows that turn my own frameworks into repeatable tools: Play to Win, Value Proposition, Product Vision, Messaging Framework, Morning Briefing, Organize Notes, Substack Note and Demo Script — built with structured elicitation and benchmark reference outputs.
An AI-native voice-to-voice research product — for my own use and customers' — that runs and synthesises conversational research at speed, turning live dialogue into structured insight.
An AI-native social posting system that drafts, shapes and schedules on-voice content — productised so it runs as a tool, not a one-off prompt.
An AI outbound system (Make.com + Claude/OpenAI): lead enrichment → AI-personalised first touch → reply classification by intent → contextual follow-up, logged to CRM. Tuned for high-value selling — depth over blast volume.
An Obsidian + Claude Code + MCP setup giving AI persistent, user-controlled context across sessions — the same 'AI inside the system, not the system inside the AI' pattern I run for clients — feeding a live competitive and market-intelligence practice.
Proof it sticks
The part almost everyone skips: embedding the capability so the org keeps moving after I'm gone.
At CyberSmart, LLM tools were embedded across both engineering and GTM workflows — not a pilot in one team. The operating model was rebuilt so product, engineering, marketing, sales and CS worked in parallel from ideation; delivery shifted from 80% maintenance to 80% roadmap.
The gardener model is an adoption system: find who's quietly ahead, spread it socially, protect the risk-takers, steer the dead ends. That's how capability sticks across a team — instead of showing up as a login-rate slide and unchanged work.
The deliverable is a capable team running a working operating layer — customer intelligence, competitive monitoring, sales enablement — so the whole organisation moves at velocity and keeps running after any individual leaves the room.
Proof others learn it
The institutions and investors that bring Mike in to teach this to others.
More of the thinking: the AI Adoption Ecology, the Product OS, and moats in the age of AI — or browse all insights.
Common questions about AI-native product leadership and AI adoption.
Three things together, which almost no one can claim at once: original frameworks for how AI actually gets adopted and operated; AI-native products shipped into market, not just features; and durable adoption — capability that sticks across the team after the leader leaves the room. That triangle of framework, shipped product and durable adoption is the real bar — far past 'I use ChatGPT to write PRDs'.
It's a reframe of AI adoption from rollout to ecology. Most companies treat AI like a software deployment — pick a platform, write a policy, count logins — and it fails, because AI changes what people are capable of, and capability spreads socially through people, not through a deployment plan. The model names the missing role (the gardener, usually product or technology leadership) and attaches a four-move method: go and look, move what people find, cover the risk-takers, and have a view on which experiments to back.
It's the operating system that turns product strategy into revenue — a loop with a spine. Five stages move signal through the organisation: Sense, Frame, Commit, Deliver and Learn. The remaining two layers are the spine: the operating rhythm and the commercial seat. It's a complete operating model designed to keep running after you leave, not a single tactic.
Both — and the building is the point. Hands-on, I've built a 7-agent engineering team in Claude Code that ships production software, a library of custom Claude Skills that productise my frameworks, voice-to-voice research and AI social-posting products for my own and customers' use, and an AI-driven GTM automation stack. That's an operator who configures and runs agentic systems, not someone who only writes prompts.
By treating adoption as an ecology, not a mandate: find who's quietly ahead, spread what works person to person, protect the people taking risks, and steer the dead ends. At CyberSmart that embedded LLM tooling across both engineering and GTM and flipped delivery from 80% maintenance to 80% roadmap. The deliverable is a capable team running a working operating layer — a system that survives you leaving.
Yes. Being AI-native means building AI into two layers — how your team works and what your product does. A fractional Chief Product Officer with this background sets the AI operating layer, ships AI and agentic capability into the core product, and drives adoption that sticks across the team — concentrated into the decisions that move ARR and NRR.
If you're standing up an AI operating layer or shipping AI-native product — and you want it to stick — a short, no-obligation conversation will tell you where the real advantage is.