ENGAGEMENT / 03 · IMPLEMENTATION PARTNERSHIP

We embed with your engineers and ship it into production together.

Two to six months, one or two week sprints, one shared backlog. Your people build it with us, so when we leave the system runs and your team owns it.

Book a call All engagements →

2–6
MONTHS: 1–2
WEEK SPRINTS: 100%
CODE IS YOURS: 45K
FROM, IN EUROS

01 / What this actually is

What this actually is.

This is where something gets built. The word that matters is embedded: we do not disappear into a side room and return with a deliverable. We work in your sprints, your repositories, your ticket system and your channels, beside the engineers who will still be there next year.

The target is a production system carrying real load — connected to your real data, reviewed by your security people, and documented well enough that a new starter can pick it up. Not a demo that impresses a steering committee and then dies.

WHAT IT IS NOT

It is not an agency handover. Nothing is thrown over a wall at the end, because there is no wall — your team wrote half of it.

02 / Who this is for

01You know which system you want and you want it live, not demonstrated.

02You have at least one or two engineers who can work alongside us.

03The thing touches real systems: your ERP, your CRM, your data warehouse, your ticket queue.

04You want the capability to stay in-house afterwards, not to be rented forever.

WHO THIS IS NOT FOR

If you have no technical people at all, this model loses half its value — start with advisory, or let us run it as a managed service instead.

03 / How it runs

How it runs.

01 · WEEK 1–2

Solution architecture

Systems to connect, data sources, security and residency constraints, failure modes, scaling profile, and the evaluation set we will judge quality against. Signed off before a line of production code.

02 · WEEK 2–3

Environment and access

Repos, pipelines, secrets handling, staging data. We fit into your tooling — Jira, GitHub, Azure DevOps, Slack, Teams — rather than asking you into ours.

03 · ONGOING

Joint delivery sprints

One to two week cycles with a working increment at the end of each. Mixed team, one backlog, one demo. Priorities can change every sprint; the direction cannot change silently.

04 · ONGOING

Knowledge transfer as we go

Pair programming, code review both ways, written architecture decisions, and short internal sessions. Transfer happens during the build, not in a workshop at the end.

05 · FINAL 3–4 WEEKS

Hardening, testing, go-live

Evaluation against representative data, load and failure testing, security review, rollout plan with a rollback path. Then live, with us watching the first weeks.

06 · CLOSE

Handover

Formal handover against a checklist: your team can deploy, debug, retrain and extend it without calling us. If they cannot, we are not finished.

03B / The cadence

What a sprint looks like.

No monthly status theatre. You see working software every week, and the honest number every second week.

MON

Sprint planning, one shared backlog

TUE–THU

Pairing, build, review both ways

FRI

Demo of a working increment

EVERY 2ND FRI

Retro, scope and cost check

04 / What you hold at the end

What you hold at the end.

01A production systemRunning on your infrastructure, carrying real traffic, with monitoring already attached.

02Source code and full accessEverything built in the engagement belongs to you. No black boxes, no licence trap.

03Architecture documentationDiagrams, decision records and the reasoning behind each trade-off we made.

04Evaluation suiteThe test set and metrics that define 'working', so quality is measurable after we leave.

05RunbookWhat breaks, how you notice, what you do about it, and who to wake at three in the morning.

06A trained internal teamNamed people who have shipped this system with us and can extend it on their own.

05 / Who you work with

Who you work with.

Delivery lead

Owns scope, cadence and the honest status. One person, one number to call.

AI engineers

Two to four, embedded in your sprints, writing production code beside your team.

Data engineer

Pipelines, access, quality and the unglamorous plumbing most pilots die on.

Architect on call

Part-time oversight on security, scaling and integration decisions.

06 / Price

from 45,000 €

PER BUILD · TYPICALLY 8–14 WEEKS · SCOPE FIXED UP FRONT

INCLUDED

Architecture, delivery, testing and go-live
Knowledge transfer and documentation throughout
Two weeks of hypercare after launch
All code and artefacts, transferred to you

NOT INCLUDED

Cloud, licence and model usage costs, billed by your own providers
Ongoing operation after hypercare — that is managed AI services
Scope added mid-flight, which is re-priced openly rather than absorbed quietly

07 / Questions

Questions we get.

How long does a build take?

Most land between two and six months, with eight to fourteen weeks typical for a first system. We size it in the architecture phase and tell you if it does not fit.

Do we need our own engineers?

Ideally one or two. The embedded model is built around shared work and transfer. Without any internal capacity it still ships, but you should plan for managed services afterwards.

What happens if scope changes mid-project?

It usually does. Scope is fixed at the start, and changes are priced and scheduled openly in the next sprint planning. No silent overruns, no surprise invoice at the end.

Who owns what we build?

You do. Code, models, prompts, evaluation sets and documentation are yours, in your repositories, from the first commit.

What if it turns out not to work?

The evaluation set exists so that is a fact rather than an argument. We agree a kill gate before the build; if the numbers are not there, we stop and you keep everything produced up to that point.

08 / Where this sits