Examples

Worked examples: what we would build, and why it holds up

Find the situation that looks like yours.

Showing all 14 examples. These are worked examples. They do not describe real clients.

Finance

The pricing model for life cover and the date of 2 December 2027
An insurer prices life cover with machine learning, so we would build in the file of a high-risk system.
The assistant on the participant portal answers in the name of the fund
A pension fund finds an assistant on its portal; we would build it as a private deployment, logging every answer.
Two registers that do not join
A bank cannot match its AI to its supplier register; we would build a tested route to leave its provider.
What a European bank's questionnaire asks of a software company in the United States
A software company in the United States is signing a Dutch bank; we would log and explain every credit decision.

The reprimand of 27 July 2026 and the office of two lawyers
Two lawyers research with a private chatbot; we would build an assistant whose every reference is read at source.
Which files may leave a notary office and which stay inside
A notary office must decide which files may leave the building; we would build a route for each kind.
When the client's guidelines say no model provider may see the file
A law firm's client forbids model providers, another expects AI; we would build a route per matter, set at intake.
A New York legal technology company and the questionnaire from Amsterdam
A New York company sells a drafting assistant to Dutch lawyers; we would let each customer choose where it runs.

The bank's questionnaire now asks about your model provider
A software company never declared its model provider as a subcontractor; we would put every feature's route in one table.
Speech driven reporting arrived before the governance did
A mental health organisation already reports by speech recognition; we would log who generated and who approved each note.
The ranking your recruiters asked a chat assistant for is on the high risk list
Recruiters rank applicants with a chat assistant; we would build screening overseen by named people and tested on past applications.
What a municipality must be able to show about the algorithm behind a benefit decision
A municipality's model picks benefit applications for extra checks; we would build it so a caseworker can deviate from it.
What the board of a water company approves when it approves artificial intelligence
Engineers give pumping station settings to an unlisted chat assistant; we would build an air-gapped assistant on the premises.
A British company, a marking tool and the long reach of Article 2
A British company marks pupils' work in the EU; we would build it so a teacher can overrule every mark.