I redesign service firms around AI: legal, accounting, agencies, healthcare ops, property management. Org design first, then production build, then workforce training. Fractional Applied AI Leader. Not an advisor. Not an automation shop.
Same headcount, different topology. The left chart routes every decision through layers: executive sets direction, directors translate it, managers approve work, ICs do the work. The right chart puts operators and agents around one shared knowledge base. Both are producers. The base is what makes them peer.
The AI-Native Org. AI is not a better reviewer of human work. Humans know the business best, and humans-in-the-loop will still verify the calls that matter. What AI absorbs is the coordination cost, not the judgment cost. Status updates, handoff translation, approval queues, the meeting before the meeting. That is the layer the left chart is built to carry, and it is the layer that collapses first.
Bain's operating-model research calls this the decoupling of headcount and output. The old chart assumed one more unit of output required one more seat. Once agents carry coordination, that assumption breaks, and the org chart has to be redrawn around the work that is left: production and judgment. Anthropic frames the operating principle as manager and IC at the same time. Every person manages agents and ships work themselves. The middle layer is not removed by force, it is removed by redefinition.
If the job is approving someone else's work, an agent approves it faster and cheaper. Managers don't disappear. They become producers again: write the spec, ship the feature, talk to the customer. The flatter org isn't a headcount cut, it is a role redefinition. Humans become AI-augmented producers. Agents become human-augmented producers.
Direction stops traveling through approvals and starts traveling through the knowledge base. North star: what we are optimizing for, measurable. Guardrails: what is off-limits, written and enforced in code. Skill bar: what gets shipped, not who reviews it. Context becomes portable. Anyone can act with full context, not only the person who has been there five years.
Middle managers without producer accountability use AI to generate decks no one reads. The redesign closes that loophole: if your output is not a shipped artifact, the chart does not have a seat for it. Memo issue 01 covers this in full.
I ran Director-level programs at eBay before AI was the tool. Now I orchestrate the agents that ship the work. One brain on both altitudes. No handoff between the person who thinks the strategy and the person who ships it. Bain calls this the operating-model rewrite. Anthropic frames the principle as manager and IC living in the same head. The wedge is the same thing said two ways.
Three things clients hire for. Same person across all three.
Director-level TPM at eBay on the Product Strategy and Operations team. I think the rewrite with you: decision flows, humans and agents as co-actors, headcount decoupled from output.
Same person designs the model and writes the integration code. No translation layer between what you want and what gets built. Strategy and shipping live in one head.
Every engagement ships with the eval suite, the playbook, and the team trained to run it. The system stays after I leave.
Six engagement shapes. The flagship is the Fractional retainer. Most clients start with the AI-Native Org Audit to map the redesign, then graduate to ongoing fractional leadership, a scoped production build, or workforce training. AI Care keeps the system running after the active engagement ends.
For companies that need a Director-altitude AI leader without a full-time hire. Owns the org redesign, the production builds, and the team handoff. Hands-on across builds, eval, hiring, and vendor calls. Ongoing retainer, 6-month minimum.
Learn more →For COO or VP Ops in pilot purgatory or facing board pressure with no pilots shipped. Maps the org redesign required to ship AI past the pilot. Deliverable: decision-flow map, AI-native role design, 90-day execution plan. 3 to 4 weeks.
Learn more →For execs who need to defend AI budget and pick where to allocate next. Board-ready roadmap classifying every AI bet across H1, H2, H3. Named failure modes per horizon. 3 to 4 weeks.
Learn more →One AI use case, scoped to production. Multi-agent workflow shipped, eval suite the team owns. Observability, token tracking, regression tests on the prompts that matter. 60 to 90 days.
Learn more →For teams hitting the 20% adoption ceiling. Three tiers: exec workshops, IC hands-on, embedded coaching. Anchored on Mollick's Leadership-Lab-Crowd framing (Wharton, May 2025). 4 to 8 weeks.
Learn more →For clients post-Scoped Build or post-Fractional who shipped production AI and need someone watching it. 2 to 3 days a month: eval monitoring, prompt regression triage, vendor escalations. $3-5k per month, 3-month minimum.
Learn more →Fractional Head of AI Transformation for operators and founders. AI broke the link between headcount and output and most org charts haven't caught up yet, so I redesign the org around how the work actually changes, then ship the production systems and train the team that runs them.
Former Technical Program Manager at eBay on the Product Strategy and Operations team, working with Director and VP-level engineering peers across the marketplace. Mechanical Engineering at SFSU. Founder of Baxie, shipping production AI publicly today. Runs Baxie and Guzman Applied Technologies in parallel out of one shared knowledge base.
Proof is built, not pitched. Below: my own SaaS shipping publicly, plus Director-altitude artifacts you can read before you book a call.
Margin OS for residential general contractors. Multi-agent orchestration across PDF extraction, estimate generation, scheduling, and rough estimating. Multi-model Claude routing. Closed-loop learning that gets sharper with every job. In paid beta.
Read moreA 1-page scoring framework operators and founders can run against their own portfolio in an afternoon. Identifies pilot purgatory, frozen transformation bets, and the next H1 efficiency win.
Read moreWorked example for a mock $50M SaaS. Shows how to allocate AI investment across H1 efficiency (~70%), H2 new capabilities (~20%), and H3 transformation (~10%), with named failure modes per horizon.
Read moreThirty minutes. If there's a fit, we scope the audit. If there isn't, you leave with a sharper read on your own AI portfolio. Capacity is 3 fractional retainers and 1 audit per quarter. If the quarter is full, you'll go on the next quarter's list.