Methodology
How I work: scope, build, verify, hand off
Every K.AI Consulting engagement runs the same way: a 30-minute call about the workflow that hurts, a written scope with a fixed price and a date, a build you can watch, verification against work you already did by hand, and a handoff with documentation and a recording. Support after launch is agreed in writing before the build starts.
Published 2026-09-30 · Updated 2026-09-30 · By Kai Myers, K.AI Consulting, Anchorage
1. The call
30 minutes
You bring the paper form, the spreadsheet nobody trusts, or the portal somebody checks by hand. I ask what happens to the data after it's captured, who reads it, and what goes wrong. You leave with a straight answer on whether it's worth building, roughly what it costs, and how long it takes. If a spreadsheet or an off-the-shelf tool is the right answer, I say so.
2. Written scope, fixed price, date
Within a few days
A short agreement in plain language: what's being built, what's explicitly out, the price, the milestones, the timeline, and who owns what. Your data and your proprietary logic stay yours. A deposit is due on signing and the balance on delivery; larger builds split the balance across milestones, each accepted on its own.
3. Build in the open
1 to 3 weeks for a sprint, 4 to 12 for a production build
I build with AI-assisted tooling, which is why it's fast. What ships is ordinary, testable software. You see working versions early, on your own files, so scope questions get answered against something real instead of a document. Changes go through a short written change order.
4. Verify against your hand-done work
Before acceptance
The honest test for a tool that replaces a manual process is to run it on batches you already did by hand and compare line by line. The disagreements become the question list, and the rules are refined from your answers, not from guesses. For a field app, that means test data and a demo account before real data goes in.
5. Hand off
At delivery
A written guide and a recorded walkthrough for whoever runs it next. Standard stacks (React, Supabase, Expo, Cloudflare, Python), a repository, and no proprietary black boxes, so any competent developer can pick it up. You are never dependent on one person.
6. After launch
Monthly retainer or scoped engagements
Bug fixes, small changes, and updates when lab formats, cleanup levels or app-store rules change, on a monthly retainer sized to the tool. New features are a separate scoped engagement. Either way it's in writing, and you can stop the retainer with notice.
Principles
Deterministic by default
If the job can be done with fixed rules, it is. AI runs inside a product only when it's the right tool, and then with fixed source material and guardrails, so it only does the job it was given.
Decision support, not a substitute for judgment
Tools that touch regulatory or safety work are built and labelled as decision support. The qualified professional still signs.
Offline is a feature
Alaska crews work where there is no signal. Field tools capture offline and sync in order when connectivity returns, without losing or duplicating records.
Numbers I'll stand behind
Outcomes on this site are stated as what clients report. When there is a measured number, it's given with its source. When there isn't, it isn't invented.