Five gates.
Each phase ends with a decision that is yours. You can stop at any of them.
Five phases, five decisions.
Each tile is a phase. At the end of it the record drops to you, and the decision is yours. Say yes and the next phase starts. Say no and it ends there.
Gate by gate.
What happens in each phase, what the 8 checks do there, what you receive, how long it takes and on what condition, and the decision at the end.
-
Gate 1
Review
Free
30 minutes, on my calendar
- What happens
- Nothing is built. I look at one workflow with you: where the work starts, where it stalls and where it needs a decision. Then I say whether it is a fit, and what the next step would be.
- The 8 checks
- Nothing is run. The fit decision says, in one line, which checks the candidate workflow would face.
- You receive
- One candidate workflow, a fit decision, the next step, and the checks that workflow would face. Not a full map or a build plan.
- How long
- 30 minutes.
- Stop point
- Either side says no.
-
Gate 2
Diagnostic Sprint
$1,000 USD fixed
Paid before work begins, not credited toward a build
- What happens
- I map the workflow as it runs today, measure its baseline, write the ROI case and set a priority action plan. Nothing is built yet.
- The 8 checks
- Each check is marked "Applies" or "Not needed", with the reason, for the chosen workflow. The acceptance rule is written down: every applicable check must pass before handover. Nothing is built, so nothing is run.
- You receive
- Workflow map, baseline, ROI case, priority action plan, stress-test plan. Yours to keep, with or without me.
- How long
- 1 to 2 weeks with timely access.
- Stop point
- Is this the workflow to build?
-
Gate 3
System Build
$2,500 USD per agreed automation
50% before build, 50% after acceptance
- What happens
- One defined workflow with one primary outcome. Integrations, decision rules, failure paths and acceptance criteria are fixed in writing before the build starts. Then I build it: fixed rules check, decide and route; an AI step reads, judges or drafts only where a rule will not do; a person approves anything sent outside your team.
- The 8 checks
- Every applicable check runs on test data before the system touches real data. The model-review date is recorded for each AI step. A failing check blocks handover.
- You receive
- Architecture, the working automation, the stress-test record, monitoring, documentation, runbook, handover, 14 days of stabilization. Excluded: new workflows, material scope changes, third-party outages, vendor pricing and usage costs.
- How long
- 2 to 4 weeks after scope and access are agreed.
- Acceptance
- The written scope is met and every applicable check has passed, or the handover does not happen.
- Stop point
- Does it meet the written scope, and did every applicable check pass?
-
Gate 4
Launch and 14 days of stabilization
Included in the System Build
- What happens
- The system goes live on production accounts that stay under your control. For 14 days after launch I fix defects inside the agreed scope.
- The 8 checks
- The record from the build stands, dated. If a fix changes the AI model or a prompt, the applicable checks run again before the change goes live.
- You receive
- A live system, the runbook, the documentation and the stress-test record, in your team's hands. The handover is designed for operation without the retainer.
- How long
- 14 days after launch, for defects inside the agreed scope.
- Stop point
- Your team owns the delivered workflow, code, configuration, prompts, documentation, runbook and data. The system must stay operable without the retainer.
-
Gate 5
Optimization Retainer
$2,000 USD a month
Billed in advance, optional
- What happens
- I watch the system, review its health each month, and run one improvement cycle on the change that matters most. You get a reply within one business day.
- The 8 checks
- The applicable checks re-run after any model or prompt change, after any third-party change that touches the workflow, and at least every 90 days. The record gets a new date each time, and the monthly health review lists the last run date per check.
- You receive
- Monitoring, a monthly health review with the last run date per check, one prioritized improvement cycle, a reply within one business day. Excluded: new systems, material scope expansion and third-party costs.
- How long
- Monthly. Optional after launch.
- Stop point
- What single change matters next?
Paid work begins with the Diagnostic. Third-party software and usage costs stay separate and need your approval.
What to bring, and what you leave with.
The review is free and takes 30 minutes. It works best when you arrive with one workflow in mind.
One workflow that costs time, money or attention. Before the call, note where the work starts, where it stalls and where it needs a decision. Name the tools it runs across.
A fit decision and a recommended next step. One candidate workflow and the checks it would face. Not a full map or a build plan.
What you own.
Your team owns the delivered workflow, code, configuration, prompts, documentation, runbook and data. Production accounts stay under your control. Your data is not used to train AI models.
The monthly retainer is optional, and the handover is designed for operation without it. The system must stay operable without the retainer.
Bring one workflow that costs time, money or attention.
Before the free 30-minute review, note where the work starts, stalls and needs a decision. You leave with a fit decision and a recommended next step.
Free. You get one candidate workflow and a fit decision. Not a full map or a build plan.