AI in Communications: Why Most Teams Are Stuck in Pilots-and How to Break Out
Comms teams have poured AI into their workflows, but most are still spinning their wheels. Fresh research shows 61% have no formal change communication strategy, 63% remain stuck in the "experimentation" phase, and 75% describe maturity as ad hoc or "discussion only." Just 5% report anything close to optimised integration. That gap isn't about tools. It's about missing strategy, governance, and resourcing.
The broader picture matches this. An international poll of CEOs in late 2025 found that while leaders expected to move beyond pilots in 2024, a year later 60% were still experimenting. Investors are starting to ask harder questions about ROI and value capture. See the IBM Institute for Business Value for context: IBM IBV research.
What's really blocking progress
It's not a lack of enthusiasm. It's a lack of a plan. In a Gallagher survey of 1,300+ communications and HR professionals, 57% said change management is their most valuable skill right now-yet 61% have no formal change communication strategy. Without a plan, AI stays a demo.
Resourcing makes it worse. 69% of firms have fewer than six people in comms, whether they employ 500 people or 50,000. 73% have no audience profiles. 37% don't have time to define tone of voice or an editorial style guide. Only 21% have toolkits or best practices in place. You can't scale what you haven't standardised.
A practical 90-day plan to move beyond experiments
- Weeks 0-2: Set the guardrails - Define business goals for AI (speed, quality, reach). Establish usage policy, human-in-the-loop rules, data security, and review/approval paths. Create a baseline style guide and message architecture.
- Weeks 2-4: Build the data you wish you had - Stand up lightweight audience profiles, a channel matrix, and tagging for content themes. Collect "gold" examples that show tone, structure, and outcomes to train prompts and evaluations.
- Weeks 4-6: Prioritise 2-3 high-impact use cases - Score by effort vs. impact. Pick cases with clear owners and measurable outcomes. Write simple SOPs so the pilot isn't person-dependent.
- Weeks 6-10: Run pilot sprints - Ship weekly. Create prompt templates, checklists, and QA steps. Track time saved, error rates, and engagement shifts. Adjust prompts and SOPs based on evidence, not hunches.
- Weeks 10-12: Scale what works - Fold winning workflows into your CMS, email platform, DAM, or chat tools. Publish your playbook, schedule training, and set a monthly review to keep it alive.
Start with use cases that actually move the needle
- Drafting and versioning internal announcements and newsletters with built-in style and compliance checks.
- Summarising long threads, meetings, or research into leader-ready briefs and Q&As.
- Media monitoring digests and early issue alerts for execs and spokespeople.
- Repurposing content across channels (email, intranet, townhall decks, short video scripts).
- Translation/localisation with human QA for nuance and risk.
Minimum viable toolkit for comms AI
- Policy and governance - Usage rules, approval flow, and records of what was AI-assisted.
- Style and message pack - Tone guidance, banned phrases, reading levels, message pillars, and example "gold" outputs.
- Prompt library - Reusable prompts for briefs, announcements, Q&As, summaries, and adaptations by audience.
- Templates and checklists - Draft-to-publish workflow, QA steps, bias checks, and legal/compliance review.
- Vendor scorecard - Security, integration, quality, and cost criteria to avoid shiny-object churn.
Measure what matters (or it didn't happen)
- Speed - Time-to-first-draft and time-to-publish vs. baseline.
- Quality - Error/rollback rates, approval cycles, and readability scores.
- Engagement - Open/click rates, intranet dwell time, sentiment shifts, and question volume post-announcement.
- Cost - Cost per deliverable and reuse ratio of assets.
- Compliance - % of outputs passing policy, accessibility, and brand checks on first pass.
Resource smart: a small-team operating model
Most teams are five people or fewer. That's workable with a hub-and-spoke model. Stand up a small "comms AI hub" that owns governance, prompts, playbooks, and training-then enable content owners in the business to execute with clear guardrails.
- Hub - 1 Comms AI lead (governance, roadmap), 1 content ops manager (SOPs, tooling), and access to a shared data/analytics partner.
- Spokes - Channel owners and subject-matter teams trained on the playbook and measured on agreed KPIs.
If you need structured upskilling, explore the AI Learning Path for Public Relations Specialists to standardise skills across your team without pausing delivery.
Don't skip the human work
AI can format and accelerate. It can't replace listening. As one industry leader put it, employees want to know their feedback is being heard-townhalls and in-person events remain the best venue for that. Keep your feedback loops tight, use AI to process the input, and let your people do the persuading.
Make it stick with simple rules
- Ship small, measure, repeat - Weekly improvements beat grand launches.
- Default to transparency - Tell employees where AI helps and where humans lead.
- Document the win - Capture before/after metrics and examples so budget talks are easy.
- Audit quarterly - Refresh prompts, update the style pack, and prune tools you're not using.
For deeper context on adoption headwinds and comms readiness, see Gallagher's research on internal communications maturity: State of the Sector. Pair that with the IBM Institute for Business Value to set realistic timelines and expectations with your leadership team.
Your membership also unlocks:
