What I do
Three ways I help.
Whether you need to define what you're building and why, turn scattered AI use into a system that actually runs your work, or fix how your product organization operates, I can help.
01
For the leaders setting direction
Strategic Foundations
Before you can operate well, you have to agree on what you're actually trying to do.
Most product organizations run hard without a shared answer to the basic questions: what are we building, why does it matter to the business, and how will we know if it worked. I work with leadership teams to define product vision, strategic bets, and the discovery and prioritization practices that keep a roadmap grounded in evidence instead of the loudest voice in the room. It's the foundation everything else, including your operating model, stands on.
What I Focus On
The difference between a strategy that holds up and one that dissolves at the first roadmap review comes down to a few things most organizations haven't made explicit.
Product Vision & Strategy
I help leadership articulate a clear, differentiated product vision and connect it to specific, measurable business outcomes. The test I use is whether someone outside the room can tell what the vision rules out.
Discovery & Evidence
I build the habits and cadence for continuous discovery, so prioritization rests on real customer and business evidence rather than on whoever pushed hardest in the room.
Roadmapping That Survives Contact
I translate strategy into an outcome-based roadmap that gives teams room to solve the right problems. Lists of features and dates tend to come apart the first time reality intervenes.
How I Work With You
I work directly with your leadership team, not around it. That means structured working sessions, pressure-testing the strategy against real evidence, and leaving you with a vision and a roadmapping practice your team can run without me. You keep ownership of the direction. I make sure it's built on solid ground.
02
For leaders whose AI stopped at individual productivity
Agentic Systems Design
A human touch is behind any great agentic system.
Your team has licenses. People paste things into a chat window, get a good draft back, and then do the real work by hand anyway. That's where most companies are, and it's why the returns have been so hard to point at. The next step isn't a better model or another subscription. It's design: deciding which work should be handed over, splitting it into agents that each own something specific, writing the instructions that tell them exactly how to do it, and defining how they hand work to each other and back to your people. That's what I do. I design the system and write the instruction sets that make it run.
I'm platform agnostic and I don't resell anyone's software. If you've already chosen a stack, I design for it. If you haven't, the design comes first and the choice gets easier once you know what you're actually building.
Where Most Companies Actually Are
One person has a prompt that saves them an hour a week. Someone else has a different one for the same task. A third team pays for a tool that does something similar. None of it is written down, none of it connects, and none of it survives the person who figured it out. The value stays trapped at the level of the individual, which is why the productivity gains are real and the business results aren't always visible. The fundamentals of how your teams work haven't changed.
What Changes When It's a System
A system is different in a specific way: the work moves without being asked. An agent owns a step and runs it when its trigger fires. It knows what good output looks like because someone wrote that down. It hands off to the next agent, or to a person, with everything that person needs to act. Nobody is prompting anything. That shift, from people using AI to processes running on it, is the whole game, and it is almost entirely a design and specification problem rather than a technology one.
What I Actually Do
I start from your work, not from a tool catalog. The question is never which platform to buy. It's which parts of how you operate should be handed over, and what has to be true for that to go well.
Work Mapping and Agent Candidacy
I map how a process actually runs today, then identify the parts worth handing to an agent and the parts that should stay with people. Most companies automate the easy, visible work and leave the expensive work untouched. I look for the opposite.
System Architecture
What agents exist, what each one owns, where the boundaries fall, and how work moves between them. An agent with a narrow job and a clear boundary is reliable. One asked to do everything is a demo.
Instruction Design
I write the actual instruction sets. Purpose, scope, the steps, the standards for good output, the edge cases, what the agent may decide alone and what it must escalate. This is the work that decides whether a system performs, and it is the part almost nobody does carefully.
Coordination and Handoffs
How agents trigger each other, what context travels with the work, what happens when something fails, and exactly where a person reviews, approves, or takes over. The seams are where these systems usually break.
The Part That Actually Matters
Every company evaluating agents is asking which platform to use. That's the wrong first question. The platforms are converging and they'll all work well enough. What separates a system that delivers from one that quietly gets abandoned is the quality of the instructions: whether the agent's job is defined precisely enough, and whether the edge cases were thought through before anyone met them in production. That's a specification problem, and it's what I've done for over 20 years. Writing the definition of the work clearly enough that someone else can execute it perfectly is product management. It turns out to be the core skill for building agents too.
What You Get
Agentic System Map
How the process runs today, which parts move to agents, and how the whole thing fits together once it does.
Agent Definitions
One clear spec per agent: what it's for, what it owns, what it can touch, what it decides alone, and what it never does.
Written Instruction Sets
The actual instructions for every agent in the system, written out in full and ready for your team to load into whatever platform you use, rather than a template for someone else to finish.
Coordination and Handoff Logic
What triggers what, what context travels with the work, and what happens when a step fails.
Human Checkpoints
Where a person reviews, approves, or takes over, what they see when they do, and how they hand it back.
Evaluation Criteria
How you tell whether each agent is doing its job, both before you turn it on and every week after.
Rollout Sequence
Which process to do first, what it has to prove, and what you build next once it does.
Ongoing Design Support
Optional retained work to tune instructions against real runs, add agents as the system earns trust, and keep the design coherent as it grows.
03
For the product organization itself
Product Operating Model Assessments
Sometimes the problem isn't what's being built. It's how the team behind it is organized.
AI has made it easier to build software and harder to ignore the cracks in how a product organization actually runs: unclear decision rights, titles that don't match the work being done, roadmaps that don't survive contact with the business. I run a structured assessment of your product organization, covering structure, decision rights, title to task alignment, and maturity, and deliver a written report and a roadmap you can act on. It's the same framework I used to scale a Product Center of Excellence for 600+ product managers at The Home Depot.
Also useful during private equity diligence or the first 100 days after an acquisition, when you need an honest, outside read on a product organization fast, before you commit to a plan, a budget, or a leadership team.
What You Get
A structured review of team structure, decision rights, and product maturity
A title to task and competency audit
A written assessment and a prioritized, sequenced roadmap
Engagements from a 2 week rapid read to a fuller 6 to 10 week assessment
Not sure where to start?
Most of my best engagements start with a single conversation about what's being built, or what should be. Book a 30-minute call. No pitch, no commitment.