Field Notes
Field Notes
Notes on deploying AI inside a company: what the work is, how the systems are built, what the rules require, and how an engagement is run. Every factual sentence cites a document we fetched; everything else is labelled as our assessment or our recommendation and says what it rests on.
- Articles
- 8
- Citing a fetched document
- 5
- Classes
- 4
Customer education
2 articles
What the work is, what it costs, and when it is the wrong thing to buy. Written for the person deciding, not for the person building.
- 01What is a forward-deployed engineer?A forward-deployed engineer works inside your organisation for the length of a deployment: watching the actual workflow, choosing the smallest system that fixes it, building it on your infrastructure, and handing it over. This note describes the role, what it is not, and when it is the wrong thing to buy.reviewed 24 Aug 20264 sectionsMethod and judgement — nothing on this page is stated as a fetched fact.4 statements0 cite a document
- 02FDE, 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.reviewed 24 Aug 20263 sectionsMethod and judgement — nothing on this page is stated as a fetched fact.3 statements0 cite a document
Technical
2 articles
How a deployment is actually put together, and what the choices trade against.
- 01How to deploy a private company ChatGPTThe shortest honest path from "we want our own ChatGPT" to a system employees use: what the four moving parts are, which decisions are irreversible, and the two places every one of these deployments gets stuck.reviewed 24 Aug 20264 sections7 statements5 cite a document
- 02Open WebUI, AnythingLLM, Onyx: which is for whatThree self-hostable projects that all get described as "private ChatGPT" and are built for three different jobs. What each one is actually shaped for, the licence differences that matter, and how to choose without installing all three.reviewed 24 Aug 20264 sections11 statements6 cite a document
Compliance
2 articles
Issue-spotting for the questions that decide whether a deployment is allowed to exist. Not legal advice.
- 01Data residency is not the same as complianceChoosing an EU region does not make a deployment lawful under the GDPR, and keeping data in Hong Kong does not answer the PDPO. Residency answers one question — where the bytes sit. This note sets out the other questions it leaves open, with the text of each instrument behind them.reviewed 24 Aug 20264 sections14 statements10 cite a document
- 02Questions to ask before adopting an AI vendorEleven questions, in the order that saves the most time, with what a usable answer looks like. Written for the person who has to sign, and structured so that four of the eleven can be checked from the vendor’s own published pages before the first call.reviewed 24 Aug 20263 sections8 statements6 cite a document
FDE practice
2 articles
How the work is done: observing, scoping, deploying, handing over.
- 01Observe the workflow before choosing the technologyThe first three days of a deployment decide most of it. This note describes what to watch, what to write down, and the four questions that turn an observation into a scope — plus the two signals that mean the project should not be an AI project at all.reviewed 24 Aug 20264 sectionsMethod and judgement — nothing on this page is stated as a fetched fact.3 statements0 cite a document
- 02Handing over an AI system responsiblyA deployment that only its builder can operate has not been delivered. This note sets out what has to exist at handover — runbook, evaluation set, ownership, upgrade path, and the decision log — and why human oversight is a duty to staff rather than a line in a slide.reviewed 24 Aug 20264 sections6 statements2 cite a document