Testing a policy message with Westminster and Brussels audiences

A policy position reaches Westminster and Brussels through channels that are public by design. A Commons departmental select committee gathers written and oral evidence and publishes its findings, and the government usually has 60 days to reply (UK Parliament, read 9 November 2026). In Brussels, feedback on a Commission call for evidence, published on the “Have your say” portal, is open for four weeks, and a public consultation for 12 (European Commission, read 9 November 2026).

Each channel has more than one reader. A Commission proposal goes to a lead committee of the European Parliament, where a rapporteur drafts the report and each remaining political group appoints a shadow rapporteur (European Parliament, read 9 November 2026). The Council’s preparatory work runs through around 150 working parties and committees (Council of the EU, read 9 November 2026). A trade association’s position on one file can be read by the rapporteur, the shadows, a Commission policy unit, a Council working party and the trade press.

The UK government’s consultation principles tell departments to consult while a policy is at a formative stage, so your position matters most when it is least settled (Cabinet Office, updated 19 March 2018). Testing that draft on the officials it is written for spends the first impression you are preparing. In Artificial Societies’ Washington engagement, putting draft messages before policy influencers “would have changed the very debate the client was trying to understand”, as the transport-policy case study puts it.

Ahead of the UK Budget on 28 October 2026, a date HM Treasury confirmed on 31 July, Artificial Societies simulated how five audiences would read the measures reported for it, from investors to Brussels, and published one audience each weekday from 19 to 23 October in its Budget-week simulation.

For Westminster, Artificial Societies asked 1,500 simulated MPs, peers, advisers, civil servants, lobbyists and journalists five questions on 14 October 2026. Of those 1,500, 579 (38.6%) named the extension of the mansion tax to English homes worth over £1.5 million as the measure most likely to be changed or dropped within 12 months. The same measure led on rebellion risk, named by 38.1% as the most likely to make Labour MPs vote against or abstain against the whip; 22.9% expected no rebellion. The comments behind the answers tied reversal to valuation disputes and homeowner backlash.

For Brussels, Artificial Societies surveyed 1,336 simulated MEPs, Commission staff, think tanks and lobbyists on 14 October 2026, and ran 10 simulated interviews from MEPs to the unions. Of the 1,336, 777 (58.2%) said delaying the UK’s target of spending 3% of GDP on defence would most damage its standing with EU partners. On the pledge to reach 3% by 2030, 621 (46.5%) called it not very or not at all credible, and 486 (36.4%) called it credible.

The same Budget choice read differently in London and Brussels. Asked on 13 October 2026 how the government should improve the public finances by about £35 billion, 543 of 2,963 simulated UK adults (18.3%) chose delaying the 3% defence target. The simulated Brussels audience named that delay as the choice most damaging to the UK’s standing. A team advising on the defence line needs both readings before the Budget.

For a global transportation leader, Artificial Societies built a society of 1,500 AI personas, each matched one-to-one to a real individual using public data and social activity, according to the transport-policy case study. The personas spanned more than 360 organisations and 800 job titles. The test ran in three phases:

1. 1.

   Baseline attitudes.

2. 2.

   Blind tests of two competing narratives and their proof points.

3. 3.

   Exposure to damaging news coverage and hypothetical scenarios.

More than 170 questions produced 250,000 responses, and the study set out to show how reactions differed by party, gender and role. It diagnosed a standing credibility gap, identified the strongest narrative and proof points, and showed where each held up or collapsed under negative coverage. Results came back in three weeks as an advisory report on positioning, proof points and risk, with full access to Radiant, our platform, down to individual persona reactions. The study compared reactions to wording the client could still change, which is a different job from forecasting a vote or a regulator’s decision.

A second Washington society, 1,364 personas of policymakers, lobbyists, think tanks and political influencers, was one of three that Artificial Societies built for Teneo in late 2025 to test six competing technology narratives. Across the three societies, the Teneo study gathered 189,756 responses in less than three weeks. Public affairs research walks through the transport-policy design.

Policy personas are built from the public record: what people have said, written and voted for, and their social activity. Artificial Societies’ Washington personas drew on public data and social activity, and each Brussels interviewee on public information about someone in that role. Hansard reports what members say in both Houses and transcribes select committee evidence sessions (UK Parliament, read 9 November 2026). The European Parliament publishes rapporteurs’ draft reports and the amendments tabled in committee and plenary through its Legislative Observatory (European Parliament, read 9 November 2026).

A persona built from a rapporteur’s public record answers as that record suggests. It cannot know what the rapporteur heard from a member state’s officials last week, or how the whips will handle a vote on the night, so the simulation prepares the meeting and the meeting decides the file. We would also expect a persona of a junior official with little public trace to have less to work from than one of an MP who has spoken in the chamber for years.

An engagement like Artificial Societies’ transport-policy work means one society built for the client, a three-phase test and, three weeks later, an advisory report plus full access to Radiant. Societies range from 12 to 3,500 personas. Bring the draft consultation response, written evidence or position paper, the alternatives you would accept, the committees or directorates-general whose reading matters, and the date it goes public. Three weeks fits inside a Commission public consultation’s 12 weeks but leaves little room in its four-week call for evidence (European Commission, read 9 November 2026). Book a call to scope it.