Embedded Discovery
Duration · 7–10 days
We embed, map the workflow with the people who run it, and de-risk the approach with a working prototype.
- –Workflow map
- –Problem framing
- –Build-ready scope
- –Working prototype
Approach
How forward-deployed
works.
No discovery deck. No offsite. Discovery ends in a working prototype and a build-ready scope, and handover ends in runbooks and architecture docs, because the people who stay need them. We start where your work happens and stay until the system runs without us.
01The weeks
Week by week.
A typical Production Build, from the first week inside the team. Embedded Discovery comes before it, and Hardening, when the system must survive enterprise reality, after. The full durations are in the phases below.
W1 · Embed
We sit where the work happens
Week one is spent with the people who run the workflow, mapping it as it actually happens: the exceptions, the workarounds, the volume. We find where AI pays and, just as important, where it does not.
W2–3 · Ship
A real system in front of real users
A working system is in the hands of the team by week two or three, integrated with the tools you already run. Feedback goes in daily. Scope stays fixed. That is what keeps it honest.
W4 · Harden
It earns trust by being boring
Evals, monitoring, access control, documentation. The system behaves the same on Friday as it did on Monday. This is the difference between a pilot and production.
THEN · Hand over
We leave. The system stays.
Your cloud, your repos, your prompts and evals, documented. If we stay, it is because you chose it.
02Engagements
Three phases. One finish.
Production Build
Duration · 2–4 weeks
The first real system, shipped to real users on real data.
- –End-to-end workflow
- –Integrations
- –Auth & access
- –Deployment
Hardening
Duration · 3–6+ weeks
For when the system must survive enterprise reality.
- –SSO + RBAC
- –Audit logs
- –Evals & monitoring
- –Reliability & security
→ Handover
Not a phase you buy: every engagement ends here, whichever phases it ran. The checklist is in the next section. Aftercare after handover is optional and by agreement: senior attention at a cadence, ended whenever you choose.
03Ownership
What handover means.
Ownership isn't a clause in the contract. It's the deliverable.
Every engagement ends the same way: your team holds everything, and nothing routes through us. The five items on the right are checked off together, in your repo, before we call it done.
[x] Handover complete
Handover checklist
Doc H-0 · every engagement
- H-01[x]Source code in your repositoriesFull history, transferred to your org. Not a zip file.
- H-02[x]Infrastructure in your cloud accountsVercel, GCP, AWS: under your billing and your access controls.
- H-03[x]Prompts and evals, documentedThe reasoning behind every prompt, with eval suites so your team can change them safely.
- H-04[x]Runbooks and architecture docsHow to operate, debug, and extend, written for the people who stay.
- H-05[x]Zero dependencies back to usNo shared accounts, no our-keys, nothing that breaks when we leave.
04Fit
An honest filter.
No public pricing means the filter has to be real. If the left column sounds like you, the Workflow Audit is worth thirty minutes. If the right one does, we'll save you the call.
Buying for an enterprise team? The enterprise page covers procurement, security, and starting with one team.
Built for
- →Real operational volume: support, ops, sales ops, back office
- →A workflow owner who can decide
- →Data that is reachable
- →Appetite to own the system after handover
×
Not for
- ×AI strategy decks with no build behind them
- ×Tool-vendor roadmaps and lock-in
- ×Committee-first decision making
- ×Pilots that are meant to stay pilots
05Questions
Asked before the audit.
Q-01Where's the pricing?+
Scope is fixed per phase and quoted after the Workflow Audit: a 30-minute call, not a sales cycle.
Q-02What stack do you use?+
Yours, where possible: the system has to live in your operation. Where we choose, we're production-deep on TypeScript and Next.js, Postgres and Supabase, Google Cloud, and the Claude API.
Q-03Do you replace our team?+
No. We build with them. Capability transfer is part of handover, and the goal is that your team runs the system without us.
Q-04How do you handle our data?+
Systems run in your cloud under your access controls. We do not use your data to train models, and nothing needs to leave your accounts.
Q-05What if AI isn't the answer?+
We say so in the audit. Sometimes the fix is boring automation: cheaper, faster, and better than AI for the job.
Different question?
The Discovery Agent knows our scope, our stack, and our limits. It answers at 3am.
Talk to our agent06 · Start
Put an engineer
inside your workflow.
Start with our Discovery Agent: a handful of questions, a structured brief, and a Workflow Audit booked straight into the calendar. Prefer to skip the conversation? Book the audit directly.