16 articles
Building and running it
Architecture decisions that are hard to reverse, and what changes once the thing is live. Retrieval permissions, structured output, agent tool access, long documents, and the cost of the year after launch.
-
Fine-tuning, retrieval or a better prompt
The three levers fix different problems. Why fine-tuning does not reliably teach facts, what it is genuinely good at, and the order to try them in.
-
Why retrieval misses the answer that is there
The passage is in your documents and the assistant did not find it. Six reasons retrieval misses, the test for each, and why hybrid search is the missing piece.
-
Guardrails: what to check going in and coming out
A rule in the prompt is a request. A rule in code is a control. What to check on the way in, what to check on the way out, and the false positives nobody measures.
-
Qualifying inbound enquiries without losing the good ones
An assistant on an enquiry form should route fast and gather context, not decide who is worth talking to. The disqualification you never see is the expensive error.
-
The jobs that should not use a model at all
A lookup, a parser and a rule beat a language model on the tasks they cover. How to tell which half of a workflow is which, in one question.
-
Multi-agent systems: when the complexity pays
The three conditions that justify splitting one agent into several, the costs nobody budgets, and why most multi-agent designs should be a workflow instead.
-
Structured output: getting reliable JSON from a model
Schema-constrained decoding, permissive parsing with strict validation, and why a valid object with wrong values is the failure that actually costs you.
-
AI training that changes what people do on Monday
Most AI training produces enthusiasm and no behaviour change. What separates the sessions that stick: your own documents, one workflow per role, a number.
-
Lead scoring that sales actually trusts
Most scoring models are ignored within a quarter, and the cause is never the algorithm. It is no feedback from closed deals, no explanation, no definition.
-
Tool-calling agents: what to let them touch
How to tier the actions an agent may take, why the dangerous combination is read plus write plus untrusted input, and the controls that survive a bad plan.
-
RAG permissions: stopping the assistant leaking documents
Why filtering results after retrieval leaks, how to enforce permissions at query time instead, and the four leak paths that survive a correct filter.
-
Document extraction: what to verify before you trust it
Field-level accuracy hides which field fails. The validation layer, confidence routing and reconciliation checks that make extracted data safe to post.
-
The integration that silently stopped syncing
A timeline of the failure nobody sees: what each person observed, why none of them connected it, and the three checks that would have caught it on day one.
-
Context windows and why long documents still fail
A large context window is not a document strategy. Why retrieval accuracy drops in the middle of long inputs, and what to do with a 200-page contract instead.
-
Better decisions, not just faster ones: where AI actually helps
Speed is the easy win. The harder question is which recurring decisions are worse than they need to be, and which of three failure modes you are facing.
-
Where the hours actually go: finding automatable work
A method for locating the work AI can take off your team without a six-month process-mining exercise, including the four shapes that look automatable and are not.
The other folders
- What will this cost, and who should build it? Buying and budgeting AI 15 articles
- Which tool, and why do these things break? Workflow automation 14 articles
- How do we know it is any good? Measuring whether it works 11 articles
- What do we have to be able to show? Governance and regulation 4 articles
- Why does ChatGPT name a competitor? AI search visibility 5 articles
Let's talk
Prefer a conversation to an article?
Most of what is written here started as a question a client asked on a call. If you have one, bring it - we answer questions before contracts.