How the work gets done
The principles behind every build, the method that keeps engagements on paper, and the three shapes an engagement can take.
What we believe about this work
Opinions, held in public. Every one of them has survived contact with production.
Accuracy is the product
A summary that cannot cite its source does not ship. We ground every extraction in the document it came from, and where a miss would cost real money, a person reviews before anything moves.
AI goes where the work is
Adoption dies in the extra tab. We build into the deal system, the inbox, the ERP, so the intelligence shows up inside a click your team already makes.
Every agent gets a supervisor
Autonomy for the routine. For the exception, a human sign-off wired into the workflow as a first-class step, with a record of who approved what.
Pipelines before predictions
Clean, boring data infrastructure comes first. A model built on a swamp makes confident guesses about the wrong numbers.
You own everything we build
Code, infrastructure, credentials, documentation, and the training to run it all. When the engagement ends you hold the keys, and nothing gets rented back to you.
Scope lives on paper
Hours and price go on paper with the deliverables before work starts. When something moves mid-build, the document moves first, and you approve the revision.
Everything important happens on paper
- STEP 1
Intro call
Thirty minutes, free, no deck. You describe the problem in plain language. We tell you whether it is really a data problem, what the fix looks like, and what it would take to build. Some problems end there, with advice and no proposal. That outcome is free too.
- STEP 2
Written scope
A statement of work with deliverables, hours, price, and the milestones where you see progress. You approve it before anything starts, and it stays the reference document for the whole engagement. When the work needs to move, the document moves first.
- STEP 3
Build
Working software at every milestone, run against your real data with the people who will use it in the room. Course corrections happen mid-build, while they are cheap.
- STEP 4
Handover
Documentation, credentials, and training sessions until your team operates the system without calling us. Then we leave. The door stays open, and nothing we built needs us to keep running.
Three shapes, priced against a written scope
Start narrow and expand, or bring us the whole system. Either way the scope document comes first.
Sprint
One problem, one working system. We take a narrow slice of the mess, build the fix, and put it in production, so you judge us on something that runs.
Duration2 to 3 weeksShapeFixed fee, fixed scopeYou getA system in productionBuild
The full system: platform, pipelines, AI features, agents. Working software at every milestone, documentation and training at the end.
Duration4 to 12 weeksShapeFixed scope, milestone billingYou getThe system, the docs, the keysAdvisory
A standing line to the people who build. Roadmap calls, architecture reviews, vendor evaluations, and help through the hard weeks in between.
DurationOngoing, monthlyShapeRetainer, cancel anytimeYou getSenior judgment on tap
Asked on most first calls
How fast does this move?
A sprint lands in two to three weeks. A full build usually runs one to three months, and you see working software at the first milestone, so you know early whether we are who we say we are.
What does it cost?
Every engagement is priced against a written scope with the hours on paper, and we bill the hours we work. If the honest fix is a spreadsheet and an afternoon, we say so on the call and you keep your budget.
We have no data team. Does that matter?
You do not need one. We build for the team you have, write the documentation in their language, and train them until the system runs without us on the phone.
Who owns what you build?
You do. Code, infrastructure, credentials, documentation. Everything lives in your accounts from day one.
What happens to the tools we already use?
We build into them. Yardi, QuickBooks, Google Workspace, whatever your operation runs on. Ripping out software nobody complained about has killed plenty of projects, so we save that advice for the rare case that earns it.
Is our data safe with AI in the loop?
Your data stays in your cloud accounts, under your access controls. Model calls go through enterprise APIs configured so your inputs train nothing, and the scope document names exactly which services touch what.
Thirty minutes settles most of it.
Bring the problem. We come back with a read on the fix, and if the fit is real the written scope follows within days.