ENGAGE

Design partnership — co-building, not a pilot in disguise.

Fifteen weeks in which you shape the platform while you evaluate it — on a decision you already have to make. Four phases, a named deliverable and an exit point at each.

Partners are selected for learning leverage, not logo value. You do not need perfect data or a mature AI strategy. You need an executive who owns the outcome, a team in transition, and a decision the evidence has to change.

How it runs

T0 is contract signature. Each phase closes on a named deliverable and a decision point; either side can stop after any review.

  1. Phase 1

    Scope — the first two weeks

    Context, objectives, candidate use cases and exploitable data. Operational constraints, for the partnership and for production. Success criteria agreed in writing. This phase is the paid opening; the fee is invoiced here and credited against setup if we continue.

    Deliverable. Scoped use cases, a data map, agreed success criteria.

  2. Phase 2

    Live — by week three

    The platform operational for the named squads, sources connected, tokens on your own keys. A forward-deployed engineer works alongside your team from here: configuring, integrating, and carrying what they learn back into the product.

    Deliverable. A platform your teams can use.

  3. Phase 3

    Evidence — to week thirteen

    Real usage, bi-weekly reviews with a named counterpart, iterative adaptation, with the forward-deployed engineer still embedded. An interim check against the success criteria, and a decision log.

    Deliverable. The use cases demonstrated, with evidence.

  4. Phase 4

    Decide — week fifteen

    Cost and benefit characterised, final evaluation, continuation decision.

    Deliverable. A learnings report — either way.

Who you work with

A forward-deployed engineer is embedded with your team through Live and Evidence — not a support contact and not a success manager. They configure the connectors, integrate the memory into the way your teams already work, and carry what does not fit straight back into the product. Palantir named the model and OpenAI and Anthropic run it too; it is how a platform reaches a real workflow without a services layer between the two.

It is why this is a design partnership rather than a pilot: the engineer is there to change the product, and the bi-weekly review is where you say what has to change.

How success is measured

One headline, and the supporting families both sides check at exit so that an observed improvement rests on a system that is safe and actually used.

The headline
The actions taken on the memory's recommendations, and the improvement observed on the goal named at scoping.
Evidence coverage and freshness
The memory covers the scoped workflows and is current enough for the decisions it feeds.
Agreement, correction and objection
Where a self-view is exposed, people recognize themselves in it and can fix it.
Permission tests with zero critical leaks
Production-like permission checks pass. One critical leak fails the partnership regardless of every other number.
Adoption without coercion
The intended users come back without being told to.

Who we look for

Engineering-heavy organizations of roughly fifty to five hundred people, running on the work-tracking, source-control and collaboration systems the connectors read. The criteria protect both sides from an engagement that cannot produce learning.

An executive owner who can clear a path
Not only approve a budget. Someone who owns delivery outcomes and the decision the evidence has to change.
One or two teams genuinely in transition
Typically a delivery or platform team of ten to fifteen people — not volunteers with spare time.
Read access to the systems they already use
Work tracking, source control, documents, calendars. No new tooling to adopt.
One named decision, this half
Concrete enough to have a date and an owner — who moves into which role, or whether a transition is working.
An AI initiative already underway
The memory measures and steers a transformation; it does not sell one into existence.
Tolerance for co-building
A product still being shaped, honest feedback — including criticism — and the willingness to have an assumption invalidated, ours or yours.

You commit

  • —The sponsor, the squads, and read access
  • —One named decision the evidence has to change
  • —Bi-weekly reviews with the people who matter
  • —Honest feedback, including criticism

We commit

  • —The platform, setup, usage and support throughout
  • —Bi-weekly reviews with a named counterpart
  • —A fee deliberately below catalogue, credited on conversion
  • —Model tokens on your own keys, or re-billed at cost
  • —Stop after any review, without penalty — a learnings report either way

What a design partnership is not

If expectations drift toward any of these, the partnership gets reset or declined. Saying so in advance is cheaper than discovering it in week six.

  • Nota free pilot
  • Notcustom development or consulting
  • Nota sales demo in disguise
  • Nota long-term dependency

Exit is acceptable

Exit without documented learning is not. Every partnership ends with a decision — continue, transition, or stop — and what was learned, written down.

If that fits

The next step is a conversation, not a form. Book a call and we will work out together whether the criteria above are met.