What mistakes did Charlotte Halligan make during an enterprise AI rollout — and what should others do differently?
Charlotte Halligan identifies three concrete mistakes made during a large-scale enterprise AI rollout: allowing conflicting announcements to land at the same time, defaulting to tech jargon that alienated employees, and involving too many stakeholders in shaping the narrative — which meant a clear, locked-down key messaging framework only existed roughly six months into the rollout.
Headcount reduction notices went out at the same time as upbeat AI messaging. Even when the two were entirely unrelated, employees drew a direct line between them. The damage to trust was immediate and difficult to undo. Halligan is clear: the sequencing and timing of communications during an AI rollout is not a logistical afterthought — it shapes everything employees believe about the programme.
Using the word "prompts" from day one created an unnecessary barrier. Plain language — something as simple as "ask your AI a question" — would have been far more accessible to a wide workforce. This is not a trivial point: the same organisation eventually rewrote its acceptable use policy from seven dense pages down to two short paragraphs after stripping out terms like LLM and large language model. The effect on employee confidence was immediate.
When too many stakeholders have a say in how AI is messaged internally, the result is delay, dilution, and inconsistency. In this rollout, it took approximately six months before a clear, agreed-upon key messaging framework was finally locked down. That is six months of ambiguous communication reaching a workforce already anxious about what AI means for their jobs.
The broader context matters here. Halligan was navigating a genuinely complex global organisation — multiple workforce types, different regulations, and vastly different levels of AI anxiety across departments. Learn more about how this rollout was handled across the full episode, available on Listenly.
"It's not necessarily that AI is going to take your job, but somebody who knows AI is going to take your job. So you kind of have to get on the bus or get off the bus."
— Charlotte Halligan, Unprompted: Real AI in IC, S01 EP4About Charlotte Halligan
Charlotte Halligan brings over 20 years of experience in internal communications, spanning many technology implementations and periods of organisational change. That depth of experience gives her a uniquely grounded perspective: she has seen how employees respond to poorly managed change programmes long before generative AI existed, which makes her account of this rollout both practical and pattern-aware.
By her own admission, Halligan was initially slow to adopt generative AI personally. That changed once she started using it — and the transition from sceptic to advocate happened fast. This matters because she took on a leadership role focused specifically on rolling out generative AI tools at enterprise scale at almost exactly the same time as her own adoption journey was beginning. She was learning and leading simultaneously, which is precisely what makes the mistakes she identifies so credible.
The rollout she describes was not a simple one: a complex global organisation, multiple workforce types with very different relationships to technology, and a patchwork of legislation and regulation to navigate. The three mistakes she names are not theoretical warnings — they are things that happened, with real consequences for employee trust and programme momentum. She presents them with the directness of someone who had to fix them in real time.