Reputation scenario testing for technology companies

Pew Research Center found the US public and AI experts reading the same technology in opposite directions. In its survey of 5,410 US adults and 1,013 AI experts, people whose work or research relates to AI, 51% of the public said they were more concerned than excited about the increased use of AI in daily life, against 15% of the experts (fieldwork August–October 2024, published April 2025). In the same Pew survey, 64% of the public expected AI to mean fewer jobs over the next 20 years, against 39% of the experts. For an AI company, communications written for specialists can therefore read as a threat to the wider public, and a workforce change sharpens that reading.

The National Conference of State Legislatures reported in July 2025 that all 50 states, Puerto Rico, the Virgin Islands and Washington, D.C. had introduced AI legislation in the 2025 session, and that 38 states had adopted or enacted around 100 measures. The policy audience for a technology launch therefore reaches well beyond federal regulators. The Pew and NCSL figures describe the sector rather than either client discussed on this page.

In Artificial Societies’ two published technology engagements, the constraint was access. Teneo’s client was preparing a sensitive technology strategy on a tight timeline, and the Teneo case study describes the audiences that mattered most as impossible to reach. The hyperscaler’s investor relations team could not test its remarks or guidance on real investors, because of regulation on selective disclosure.

In late 2025, Artificial Societies built over 5,000 AI personas for Teneo’s project in three societies: Washington D.C. (1,364 personas), Tech Leaders (1,526) and General Population (2,381). The personas were grounded in real-world profiles and informed by social listening and deep qualitative research. Surveys and experiments in each society tested six competing narratives through 30 deep-dive questions, producing 189,756 unique responses. The study’s critical questions went beyond approval to how different stakeholders would react, who the true influencers driving the conversation are, which media sources hold the most credibility, and which partnership opportunities to explore.

“What we were able to accomplish with Artificial Societies would simply have been impossible with traditional market research.”

In the hyperscaler engagement, revenue was about to double year on year while the margin guide came down. Artificial Societies built four societies around the company’s stock, more than 2,400 personas in all, each matched one-to-one to a real individual using public data and social activity. The tests ran in three stages: the question each analyst was most likely to ask, investor reaction to the guide and to alternative versions of the prepared remarks, and how the story would travel through the company’s ecosystem and the press. The hyperscaler case study records nearly 125,000 responses within 48 hours.

Artificial Societies’ Teneo and hyperscaler engagements built seven societies between them, from Washington policymakers to a subgroup of the company’s own employees. Separate Tech Leaders and General Population societies let a test report the tech community and the public apart, which is where a gap like the one Pew Research Center measured between AI experts and the public would show.

Source: Artificial Societies’ Teneo case study (late 2025) and hyperscaler earnings-call case study.

The hyperscaler engagement gave the press a society of its own: more than 1,600 personas of business and finance journalists. Teneo’s study built no press society and asked instead which media sources hold the most credibility and who the true influencers are. That second question carries weight because a growing share of news arrives outside the press: the Reuters Institute’s 2025 Digital News Report found 54% of its US sample getting news through social media and video networks, a share it describes as sharply up. Teneo’s policymaker society represented Washington D.C.; the state-level scrutiny NCSL counts would need a society of its own, and neither published technology engagement built one.

Taken together, Artificial Societies’ two technology engagements show a reputation scenario test in five steps:

1. 1.

   Fix the scenarios. Teneo’s study tested six competing narratives; the hyperscaler had prepared remarks, a guidance range and a view of what the market would fixate on, and the tests covered alternative versions of the script.

2. 2.

   Build a society for each audience; a society holds 12 to 3,500 personas. Teneo’s were grounded in real-world profiles, social listening and qualitative research; the hyperscaler’s were matched one-to-one to real people from public data and social activity.

3. 3.

   Run surveys and experiments in each society. Teneo’s used 30 deep-dive questions across the six narratives.

4. 4.

   Follow the story outward, as the hyperscaler’s third stage did, through the company’s ecosystem and the press.

5. 5.

   Read the results at aggregate and individual persona level. Teneo could explore approval scores, emotional sentiment and verbatim reactions by audience segment and narrative.

When several drafts go to the same society, the first draft a persona reads can colour its reaction to the next. Our page on crisis communication simulation explains how resetting a society’s memory between exposures controls those order effects.

For Teneo’s client, the output was a choice among six narratives: Artificial Societies identified the strongest overall narrative and optimised the messaging for each audience segment, delivered as a written report and in Radiant, our interactive enterprise platform.

The hyperscaler test corrected the company’s own read of the room. The company expected the margin reset to dominate its earnings call; the simulated analysts largely looked past it and pressed instead on the durability of demand and the scale of the company’s financing commitments. On the live call, the questions ran the same way. Script testing showed which version of the prepared remarks landed best with investors, and for each analyst likely to be on the call, the team had the most likely question in that analyst’s own words.

We would weight that correction above the winning narrative. A ranking of drafts, even one optimised by segment, improves answers a team has already written, whereas a correction to its read of the room changes which questions the team prepares for at all.

Neither of Artificial Societies’ published technology engagements was a live crisis: Teneo’s was a pre-launch narrative test and the hyperscaler’s a pre-announcement disclosure test, so neither shows a simulated society following a story that is already breaking. A simulation tests the scenarios a team writes down. It cannot tell you what a journalist already knows, such as a source inside the company or a document you have not seen, and it cannot predict an event nobody thought to test.

Once a story breaks, the real coverage and the real reactions are the evidence to read first. For technology company reputation management, simulation is a pre-launch tool: it belongs in the period when the wording, the order in which audiences hear it and the spokesperson can still change.

Both published technology engagements started from one decision with a date on it: a technology strategy launch and a quarterly earnings call. You bring the drafts and scenarios you want to compare, such as competing narratives or alternative versions of a script, and name the audiences whose reading matters. We build a society for each audience, run the surveys and experiments, and return a written report with access to Radiant, so your team can re-run scenarios and re-test drafts as they change. To scope a test, book a call with the team.