Skip to content
Is there an AI for this?

Field note · Customer education

FDE, AI consultant, or software agency?

Three suppliers sell overlapping things and price them differently. This note sets out what each one is accountable for, what you get at the end, and the question that separates them: who is responsible when the system is wrong in production.

Source
Written by us. Method and judgement — no statement on this page rests on a fetched document.
Evidence
none on this page — it links to the pages that hold it

01The difference that matters

All three will meet you, ask about your goals, and propose a plan. The difference shows up at the end of the engagement, in what exists and who can operate it.

DimensionAI consultantSoftware agencyForward-deployed engineer
Priced inDays or a fixed advisory feeA scoped projectDays, against a deployment
DeliverableStrategy, roadmap, vendor shortlistA build against a specificationA working system in your environment
Writes the specificationOften, then leavesYou do, or they do firstThey do, after watching the work
Sees your real dataRarelySometimes, lateOn day one, under your controls
Accountable when the output is wrongNoAgainst the specificationFor the deployment
Ends withA documentA handover of codeA runbook and a named internal owner
What each supplier is accountable for — our assessment, not a vendor comparison chart

02Choosing between them

  • RECOMMENDATION

    We recommend a consultant when the decision is a portfolio decision — which of eleven candidate use cases to fund — and no single deployment has been chosen yet.

    BasisDepends on the brief having no single objective. Rests on the accountability table above: a consultant’s deliverable is the decision, and a decision is what is missing at that stage.

  • RECOMMENDATION

    We recommend an agency when the specification is already firm, the interface matters more than the model, and the work is mostly conventional software around a stable API.

    BasisDepends on a firm specification and a low share of retrieval or evaluation work. Rests on the same table: an agency is accountable against a specification, which is only a useful guarantee once the specification is right.

  • RECOMMENDATION

    We recommend an FDE when the constraint is where the data may go, when the workflow is specific to your organisation, or when nobody yet knows what a correct answer looks like.

    BasisDepends on the brief’s data-residency and confidentiality attributes and on the absence of an agreed quality bar. Rests on the observation step described in “What is a forward-deployed engineer?” — the diagnosis is the deliverable those three conditions actually need.


03Questions that separate them quickly

  1. 01Who will be sitting with our team, and for how many days?WhyNamed people and a day count are checkable. "A team of specialists" is not.A usable answerOne named engineer, on site or in your systems, for a stated number of days a week.
  2. 02What will exist at the end that we can operate ourselves?WhyThis separates a report from a system, and a system from a system somebody else can run.A usable answerA runbook, credentials we hold, an upgrade path, and a named internal owner who has already done a release.
  3. 03How will we know the output is good enough?WhyIf the answer is a demonstration, there is no measure. If it is an evaluation set built from your own documents, there is.A usable answerAn agreed set of real cases with expected answers, run before and after every change.
  4. 04Which of your recommendations pays you a commission?WhyA supplier who resells a platform is not neutral about the platform.A usable answerA direct answer, and a willingness to price the engagement without the resale.
  5. 05What would make you tell us not to do this?WhyA supplier who has no such answer has not thought about your case.