Working Surfacelink · 16 Nov 2026
02
A row of small bound phrasebooks on a shelf, each tied by string to a different signpost pointing toward its own district of a city, while a figure seen from behind corrects a page in one with a pen, the whole scene drawn with wide empty margins on every side.
martinfowler.com

Generated Glossary

Martin Fowler · 12 November 2026

Company story · agents · artifacts · Agent context files

Read the original
Takeaway

Martin Fowler describes having an AI model draft a glossary of an organisation's private vocabulary, corrected by its experts, so that AI agents use terms the way staff do.

Summary

Martin Fowler, chief scientist at the software consultancy Thoughtworks, wrote a short practice note on his website about AI models working inside large organisations. The models do not know the organisation's specialised vocabulary. Drawing on the experience of his colleagues Eric Nagler and Unmesh Joshi, Fowler describes generating a glossary for them.

Key points
  • An AI model reads internal documents, emails and chat logs and lists new terms and familiar words used in unfamiliar ways, which Fowler says often cause the most confusion.
  • Human experts in the business review the draft and correct it before AI agents use it.
  • The glossary sits where AI agents can consult it cheaply, is kept in version control, and is refreshed on a schedule by agents and people.
  • Fowler advises separate glossaries for separate parts of the business, and entries for terms that should not be used, such as internal names kept out of customer screens.
Implication

A team whose AI agents misuse company terms can start with an expert-checked glossary for each part of the business; this is an inference from the note. The piece is a short practice note and reports no measured results.

Plays
Product

Keep a separate glossary for each part of the business in version control, with entries for terms agents must not use, such as internal names kept out of customer screens.

Engineering

Have an AI model draft a glossary of company terms from internal documents and chat logs, then have experts in the business correct it before AI agents use it.

Derived by Working Surface from the article; more in the Playbook. Source line: I'd urge everyone to resist any temptation to create a single, standard, glossary - I've yet to see that work.

Source issue

16 November 2026: A team learns what AI is worth by counting what its people kept from its work, by having experts correct what agents read, and by recording where work waits.