Method · laboratory
Make AI competent in your context.
Expertise encoding is the work of turning professional judgement into AI systems that keep context over time. It is the difference between asking a generic model for help and building a system that understands your documents, decisions, workflows, examples and quality standards.
What this page is: a method case, not a client programme. Most organisations do not lack access to AI. They lack a way to make AI competent in their specific context. NiallOS is the working laboratory behind this pattern.
The problem
Professional knowledge is fragile.
The most valuable knowledge in an organisation rarely lives in one clean place. It is spread across people’s heads, old proposals, meeting notes, slide decks, emails, policy documents, spreadsheets, and repeated judgement calls that senior people make without writing them down.
Generic AI tools can produce fluent output, but they do not automatically know which sources matter, how your team makes decisions, what quality looks like, or what would be risky to get wrong.
The design bet
The goal is not to replace judgement. It is to make expert context available at the point where work happens, and to keep that context reviewable.
What it means
Capture how experts think, then make it reusable
Expertise encoding combines five kinds of material. None of them is enough on its own.
Source material
Documents, policies, research, examples, client work and reference material that the system is allowed to use, and that people can still check.
Professional judgement
How experts decide, prioritise, assess quality and handle ambiguity, including the cases where a fluent answer would still be wrong.
Workflow
The repeated steps people follow when producing reports, proposals, workshops, analysis or products, so the system supports the work rather than inventing a new process.
Quality gates
The checks that stop plausible but wrong output from reaching clients or stakeholders. Review is part of the system, not an afterthought.
Retrieval and synthesis
Ways for AI to find, compare, cite and reason across relevant context, instead of relying on whatever happened to be in the last prompt.
Human in the loop
Critical decisions, client outputs and judgement-heavy work stay reviewable and auditable. People still specify, supervise and reject the work.
How the process works
Five moves, in order
This is a working sequence, not a product roadmap.
A simple model
Four layers, not a clever chat
The useful unit is an operating layer. Each layer depends on the one above it.
4. Reusable AI capability. Not a generic assistant. A system that can work in this context, under review.
Where this applies
Education first, and anywhere expertise is the product
Most of my work is in education. The pattern applies wherever professional knowledge is complex, document-heavy and context-specific.
Education organisations
Making AI useful across curricula, policy, professional learning and school leadership, without flattening local context.
Professional services
Teams whose value depends on expert judgement and client context, and who cannot afford fluent but ungrounded drafts.
Learning and research teams
L&D, research and policy groups working across large bodies of evidence, with a need to cite and defend what they produce.
Development partners
Organisations operating across countries, languages and institutional contexts, where generic tools miss the setting.
Proof of concept
NiallOS is where I test the pattern.
NiallOS is my working laboratory for AI-assisted knowledge work. It combines a structured database, project workspaces, document ingestion, verified extracts, reusable skills, semantic search, review queues and agent workflows.
It is not a generic productivity app and not a public product. It is a live example of what happens when professional context is made persistent enough for AI to work with it.
What it is not
Not a chatbot. Not a product page. Not a client case study.
- Not a replacement for experts. Encoding judgement makes it available. It does not make it optional.
- Not a tool you can log into. There is no public expertise-encoding service, and this page is not a SaaS pitch.
- Not a substitute for the programme pages. DELP, KFAS, ECOSOCC and the other selected work pages are delivered programmes with public sources. This page is the method underneath.
Confidentiality: this page describes a method. It does not publish client files, internal databases or screenshots that would expose other people’s work. NiallOS holds working material from consulting, research and personal practice; that material stays off this site.
Trying to build this kind of capability?
I work with education organisations and professional teams on encoding expertise, grounding AI work in sources, and putting review around delivery.