Draft for review · not yet live · to be checked by Matt before publication
Use of AI

AI does the searching.
People decide.

Companies are unsure about AI. We are not. We know what AI is worth, where AI is strong and where AI is weak, and we use AI deliberately: as a tool in service of our clients and the people our clients serve. This page says exactly how.

What AI does in our work

AI searches, structures, connects and drafts at a scale people alone cannot reach. In practice that means finding and reading sources, proposing how evidence might be organised, summarising large volumes of material, and producing first drafts for a person to work from. That lets a small team give a client an evidence base no single brief could otherwise afford.

What AI never does

AI does not decide what is true, what matters or what a client should do. Those are human judgements, and those judgements stay with people. Nothing reaches a client that a person has not read. Every claim we put in front of a client links to a source and carries a confidence rating, and where the evidence is weak we say so rather than let a fluent sentence stand in for a fact.

Mixed intelligence, declared

Work that used AI says so, in every sector we work in. A report, a prototype, a map or a campaign that AI helped to make carries a notice saying the work was created with mixed intelligence, meaning AI and human input together, that the work draws on public and, where permitted, proprietary data, and that content should be verified before reuse. This website was written by people with AI assistance and reviewed by people before publication. We do not publish AI-generated testimonials, and we do not present synthetic people as real evidence.

How your information is protected

Client and research participant information never goes into consumer AI tools. We use business or enterprise tiers whose terms exclude training on our data. Identifying or sensitive content is removed before any material is processed by AI. Outputs are reviewed by a person before use. Access is restricted to named people and logged, and the platforms we control keep data in UK and EU hosting. Sub-processors are named in the data protection documentation we share with clients.

Which tools

We use a small number of leading AI providers on business terms, chosen for the task, and we review the list as the field changes. We name the providers in the data protection impact assessment we give to clients, rather than here, because the list is a security matter and changes faster than a website should.

Proportion

We use the smallest model and the shortest run that does the job well. That is good practice, and respect for the environment that sustains the people we serve.

Questions

If you are a client and want the detail, ask for our data protection impact assessment. Anyone else, write to hello@uservox.com.

Version for review, 24 September 2026. This page belongs to Uservox Limited and applies to uservox.com.