Testing an employee announcement before you send it

The consulting team in Artificial Societies’ integration engagement faced three obstacles, and a restructure, a new operating model or a return-to-office policy is likely to meet them too. Testing a rollout sequence with real teams would have changed the integration before the comparison was complete. Informal influence sat outside the organisation chart, leaving no reliable map of who would move adoption. And internal communications data was difficult to obtain, while the senior people who mattered most were the hardest to recruit at scale.

A global consulting firm was advising on the integration of a 200-person professional-services firm into a 5,000-person acquirer, with one attempt at the real integration. Artificial Societies modelled the 1,500 people across both firms with first- or second-order touchpoints with the acquired team, as AI personas built from real, anonymised professional profiles in public social-interaction and employment-history data. Artificial Societies ran 12 integration strategies in parallel on independent copies of that network, each across four intervention phases, producing more than 340,000 responses in six weeks. The case study reports three findings:

1. 1.

   Sequencing mattered more than messaging. Strategies that began with the technology rollout before partner alignment produced the largest retention declines among senior people in the acquired firm, even when the message itself tested well.

2. 2.

   Roughly 40 mid-level employees whose employment histories connected the two firms, none with formal authority, were the people through whom adoption moved fastest when they were engaged first.

3. 3.

   The messenger changed the reception. Firm-wide announcements drew simulated responses along the lines of “this is being done to us, not with us”, while the same message met less resistance when it came from a partner the recipient already knew.

The consulting firm used those findings to recommend a sequence that, according to the case study, met its talent-retention target and reached its technology-adoption target one quarter ahead of schedule.

Settle the order before the wording. In the integration Artificial Societies simulated, putting the technology rollout ahead of partner alignment produced the largest retention declines among senior people in the acquired firm, even when the message tested well. Rewriting the memo is the cheapest fix, and on this evidence the less decisive one.

Part of the order is set by law. In the UK, an employer proposing 20 or more redundancies at one establishment within 90 days must notify the Redundancy Payments Service on form HR1 before consultation starts, then consult employee representatives for at least 30 days before the first dismissal, or 45 days for 100 or more (GOV.UK, read 9 November 2026). In the US, the Worker Adjustment and Retraining Notification (WARN) Act requires employers with 100 or more employees to give at least 60 calendar days’ written notice of a plant closing or mass layoff (US Department of Labor, read 9 November 2026). Your employment counsel fixes those steps. A simulation tests the order the law leaves to you: who hears first, and whether the operational change arrives before or after its carriers are on board.

The people who carry a change need not be the people who announce it. In Artificial Societies’ simulation of the integration, adoption moved fastest when roughly 40 mid-level employees with no formal authority were engaged first. What marked them out was an employment history spanning both firms, a connection the organisation chart does not record.

Network research agrees, with a caution. Julie Battilana and Tiziana Casciaro analysed 68 change initiatives in the UK’s National Health Service and found that change agents whose contacts were not connected to one another helped changes that diverged from established practice get adopted, but hindered the adoption of less divergent changes (Academy of Management Journal, April 2012). Which connectors to engage first probably depends on how far your announcement departs from current practice, a question to test rather than assume. We built the integration network from public professional data, so it approximates the informal network rather than reproducing it: a working relationship with no public trace is missing from it.

The evidence comes from one integration engagement. Applying it to a restructure, a new operating model or an office policy is our reasoning about why the same mechanics should carry over, not a separate study. Each of those announcements asks people to accept a change they did not choose, from a messenger, in an order, through an informal network that shapes what colleagues believe about it. All three mattered in the integration simulation, and none is specific to a merger.

The result itself does not transfer: the sequence that worked for one acquisition, or the roughly 40 people who mattered in it, says nothing about your organisation. Testing internal announcements elsewhere means building each organisation’s own network and treating the integration findings as hypotheses. A simulation also returns simulated reactions, and your consultation process, engagement surveys and managers’ conversations remain the record of what employees think afterwards. Our pages on post-merger integration and testing corporate strategy announcements cover the integration itself and how outside audiences read a deal.

A first engagement for employee communications testing follows the shape of Artificial Societies’ integration work. Decide who is in scope: that engagement modelled the 1,500 people closest to the acquired team rather than all 5,000 at the acquirer, and a society can hold 12 to 3,500 personas. Express each option as an intervention sequence, as the integration’s 12 strategies were. The strategies run in parallel on independent copies of the network, so one does not colour the reaction to the next, and the consulting team could follow each strategy, see where it began to fail and refine the sequence before recommending it.

Clients receive a report and full access to Radiant, our platform, to re-run the analyses themselves. To scope a test, book a call.