The method, then the machine.
I spent a career inside the old model. I still use those tools. What changed is the client, and the speed.
How an AI implementation engagement actually moves.
Nine practices I bring to the work. We use them as the situation needs, not as a checklist, and not always in this order. One conversation can touch several. We return to them as the work unfolds.
Understand
Nothing changes yet. The system is measured as it truly runs.
Intake evaluation
Where the business is, where it hurts, and the 1-to-10 question that finds the gap worth closing.
Value-stream mapping
How work actually flows: where effort pools, where it waits, and where it quietly leaks.
Identify the pain
The constraint named and measured. The most obvious pain point, not the most elegant one.
Design
The fix is scoped from evidence, and reviewed before anything is touched.
Solution design
The smallest system that removes the constraint, with opportunity solution trees keeping the options open.
System architecture
How the pieces fit: your existing tools, the new system, and the seams between them, built to outlive any single contributor.
AI strategy
Where AI belongs in the design, where it doesn't, and the operating rules that govern it in your organization.
Deliver
Working software in daily use, and the practice that keeps it that way.
AI implementation
Built in short, reviewable passes, every change tested before it ships. Staged, reviewed, reversible.
Testing and feedback
Verified against the real environment, with the people who use it, not a demo harness.
Retrospective
What worked, what didn't, and the lessons log that keeps the practice improving after I leave.
Work that holds you back, and work that fits.
Same hours in the week. Different system underneath them.
Work that holds you back
The week as it usually feels
- Work lives in six tools and one person's memory
- AI output with no rules and no review
- Meetings that exist to find out what's going on
- Fixes that leave when the consultant does
- Growth that feels like more noise
Work powered by a practice that fits
The week after the method is in
- One place that shows the state of the business
- A practice with rules, review, and a method you keep
- A ranked surface for what actually needs you
- The method stays after I leave
- Growth that feels calm
The tools I actually use.
Not a menu of services to buy by the hour. A typical working set of practices and methods.
AI practice
- AI strategy
- AI design
- Agentic workflows
- AI support
Systems
- System architecture
- Software design
- Requirements gathering
- Constraint identification
Improvement
- Value-stream mapping
- Systems and design thinking
- Opportunity solution trees
- Continuous improvement
Delivery
- Change management
- Automated testing and release
- Business improvement analysis
- Efficiency
Two things I learned by getting them wrong.
A real site does not take a department or a year.
If the copy is ready it can be live in days, with no build step and no database. Most people have been told the opposite, and that belief is now the expensive part.
Detection can be bulk. Correction cannot.
I once wrote a data correction that ran by itself on every restart and quietly overwrote records that were already right. The repair is never a cleverer sweep. It is per-record verification against something outside the data, and the ability to put it all back.
One person is a risk. Here is how I handle it.
Everything I build runs on your accounts, not mine. The documentation, the runbooks and the source live in your repository from day one, not delivered at the end. The practice I install is written down in your language, so someone else can pick it up and read it. On my largest engagement the repository, the hosting and the domain were transferred into the client's own accounts, which is what leaving the method behind actually looks like.
If an engagement genuinely needs more hands than I have, I will say so rather than stretch. I have thirty years of contacts who do this work and I would rather bring one in than deliver you half of something.
See the method on real work.
Every case study includes a wrong turn. That's deliberate. A case study with no wrong turns is marketing.