Journal
The six-week proof
Agentic Engineers · 4 min read
- Engagements
- Agentic Workflows
One agentic workflow shipped to your production environment in six weeks. Why a short, tightly scoped proof de-risks the decision better than a multi-month retainer.
Most engineering engagements ask buyers to commit to a multi-month retainer before anyone has seen the work ship. That makes sense for the seller. It is a terrible deal for the buyer, who is taking on a quarter of cost against a deck and a reference call.
A short proof engagement is the alternative we built so the first proof point lives in your production environment, not in the slide pack. Six weeks. Fixed scope, fixed window. One agentic workflow shipped into your stack, plus a discovery doc that maps the next three, plus a metrics baseline so you can measure compounding over time.
How the six weeks run
The structure is deliberately rigid. Rigidity is the feature; flexible discovery engagements drift and ship nothing.
- Week 1 to 2, discovery and selection. We map your operational surface, pick one workflow where AI-forward automation will compound (code review, ticket triage, customer-facing data fetches, recruitment screening, sprint reporting are the common five), and define the success metric in writing before we touch the code. The discovery doc is part of the deliverable, not a separate phase.
- Week 3 to 5, build, integrate, validate. We build the workflow against your real stack, integrate into the tools your team already lives in (GitHub, Linear, Slack, your repo, your data warehouse, whatever your operators actually open), and validate against the metric defined in week two. No standalone demo. The workflow runs where your team actually works, against real data, with real failure modes surfaced and handled.
- Week 6, handover, baseline, decision. The workflow is yours. The discovery doc maps the next three candidates with effort estimates. The baseline metric is recorded so you can measure compounding over time. The decision at the end of week six is yours: end the engagement, or continue into an ongoing embedded engagement with the system already running.
The engagement is scoped to make the decision easy. Either way, you are net ahead.
What this engagement is not
A few things this is not, because the category is crowded with adjacent offerings that look similar from the outside and feel nothing like it from the inside.
It is not a discovery doc that no one reads. We have read enough of those that we know the shape, and we refuse to produce one without a live workflow attached.
It is not a proof of concept that breaks the moment it touches production. The engagement ships into production, against your real auth and your real data, by week five. If it cannot survive contact with production, it is not finished.
It is not a billable timeline where the meter keeps running past week six. The scope is fixed and the work does not move with the hours on a timesheet; we commit to capacity, not a meter. If we underestimate the lift, we wear it. If we overestimate, the remaining capacity gets pushed into the second candidate workflow from the discovery doc.
Eight weeks lets the team drift. Four weeks is not long enough to ship a real workflow against a real stack with real edge cases handled. Six is the smallest window that consistently produces a system the team can run unsupervised at the end. We have tested both edges. We landed here on purpose.
Who this is built for
A short proof engagement is for the founder or operator who is one of three people. The first is the buyer who has heard the AI-forward pitch a dozen times this year and wants to see one workflow actually run before signing a longer engagement. The second is the team that has tried installing an internal agentic workflow, watched it drift inside a month, and wants someone who has shipped this pattern across multiple stacks to install it correctly the first time. The third is the founder approaching the 1-to-N break who needs a low-risk proof of fit before committing the team to a multi-quarter operating model change.
If you do not fit any of those three, this is probably not the right entry point and we will say so. The fastest engagement is the one we decline before it starts.
What you walk away with
At the end of week six, regardless of whether you continue, you have:
A live agentic workflow running in your production environment, owned by your team, with the code in your repo. A discovery doc mapping the next three workflows worth installing, with effort estimates and expected impact. A baseline metric for the workflow you shipped, recorded so the next operator (us or you) can measure compounding from a known floor. The internal vocabulary your team needs to talk about agentic workflows without falling back to "AI" as a black box.
If you continue into a longer embedded engagement, all of that becomes the foundation of the operating system we keep installing. If you do not continue, all of it is still yours.
How to start
We take on a small number of these each quarter, because each one draws on the same shared infrastructure our active engagements rely on. If you want one scoped for next quarter, that is the conversation to start now. Book a call.
