Skip to main content
briefingagenciesGEOFoundationsAI tools

Comms With AI Briefing, 25 September 2026: Who owns the AI at your agency?

The first fortnightly Briefing. Four agencies put AI in a P&L role this fortnight and a client-agency panel named GEO audits as the new upsell. Plus Foundations, now live on Comms With AI: save how your organisation sounds and every Plan and Message Test starts from it.

Get the Briefing in your inbox

What is happening in comms and AI, every other Friday. Free.

Welcome to the first Comms With AI Briefing, which replaces the Comms With AI Monthly. So much is moving in comms and AI that once a month left too much out, so every other Friday we now cover what agencies, tools and models are doing and what it means for your team. It is new: tell us what helps and what is missing at michael@faur.site.

Four agencies hired someone this fortnight to run AI as a business line, and a client-agency panel named GEO audits as the trade’s latest upsell. If you buy agency time, the question running through this issue is what sits underneath the offer: counsel, a tool, or repackaged model output.

From Comms With AI

Foundations is live. Save how your organisation sounds, who it speaks to and what it has already said, and every Plan build and Message Test starts from it. What changes is the output. The tone rules match your tone of voice rather than a guess. The audiences are the ones you named. A brief that departs from a position you hold is flagged rather than quietly overridden, and the Message Test rewrite honours your red lines. Your foundations now carry what you have already said, so every plan is checked against it. A profile is nine short fields; paste an about page or a tone-of-voice guide and a helper proposes them for you to correct. A free account keeps one profile, members keep three, one per client if you are agency-side. Switch it on with the “Apply foundations” toggle. Save your foundations

On Applied / Comms With AI

Comms With Siri AI and Apple Intelligence. We put Apple’s rebuilt assistant, in beta with iOS 27, through real comms work (22 September). It proofread well and invented freely: asked for a LinkedIn post in our usual style, it cast the author as part of the Apple team launching Siri AI. It reaches staff phones with a software update, so treat it as AI your team already has. Read on Applied · the directory entry has the three settings to check first.

What the agencies are doing

Burson has named Jennifer Stearns global chief innovation officer (PRovoke, 17 September), replacing Chad Latz after 17 years (PRWeek US, 10 September). A consultant from Accenture now runs AI at WPP’s largest PR brand. If you brief Burson, ask what is tool, what is counsel, and what data sits behind the platform. Read

Penta has hired Dustin Johnson from Weber Shandwick as president of global intelligence and AI (PRovoke, 17 September), and says roughly half its revenue now comes from intelligence. If an adviser offers AI-driven monitoring, ask how it differs from the licence you already pay for. Read

Two more in the same week. Team Lewis gave Fairil Yeo a new global EVP of AI and growth post (PRovoke, 16 September), and APCO hired Harris Poll veteran Rob Jekielek to run Margy, its stakeholder-tracking platform (PRovoke, 14 September). Expect AI as a priced line on the next proposal. Team Lewis

Are GEO audits “the new upsell”? A client-agency panel named AI search audits as agencies’ latest retainer add-on (PRWeek UK, 17 September, paywalled). Days later, corporate reporting agency Emperor took a minority stake in GEO specialist Exceptional (Communicate, 22 September), so expect AI search bundled into reputation pitches. Before paying for an audit, ask what it measures, against which prompts, and which parts the agency subcontracts. PRWeek · Emperor

Tools and models

Two model releases on the same day. Anthropic released Claude Opus 5.5, which it says costs about 40 per cent less than Opus 5 and runs more than 30 per cent faster (Anthropic, 22 September). OpenAI released GPT-6 Sol and Luna, cheaper models it says bring “more clarity, less jargon” (OpenAI, 22 September). We are testing both on real comms tasks: results on Applied on Thursday 1 October, then we talk them through with Emma Ewing at Big Fish Training. Opus 5.5 · GPT-6 Sol and Luna

ChatGPT now carries sponsored agents. After clicking an ad, a user can open a labelled conversation with the advertiser’s own agent, in test with selected US advertisers (OpenAI, 16 September). A rival’s paid agent can now sit beside the answer about you, so add ChatGPT to where you check how you are described. Read

Google Meet starts AI notes automatically from 29 September. A new admin setting switches on “Take notes for me” for meetings of three or more people, on by default for Business Standard and Plus (Google Workspace Updates, 22 September). Before Tuesday, agree with IT which meetings stay unrecorded: crisis calls, HR conversations and embargoed briefings. Read

Test it yourself

Put the new models on your own work

Take one task your team does every week, such as a press release or an all-staff update, and run the same anonymised brief through your current tool and one of the new models. Strip the labels and ask two colleagues which version they would send. That is the Bench Test in miniature; the full template turns the answer into a one-page Tool Map.

Open the Bench Test →

Worth reading

  • The prompts your team optimised against were made up. Jonny Bentwood argues most GEO programmes score visibility against guessed prompt lists (Ragan, 17 September). Ask where the prompt list came from. Read
  • My organisation is asking me to use AI to reduce budgets. What should I do? Stephen Waddington on why AI-led headcount cuts return little (Wadds Inc., 14 September); his Summer Annual finds a formal AI strategy in 24 per cent of corporate affairs functions. Read
  • Who do Brits trust? Ipsos’s first measure of AI chatbots: 15 per cent of GB adults rate them trustworthy, ahead of politicians and influencers at 11 per cent (Ipsos, 18 September). Read

Events

Friday 2 October, 13:00: AI tools on trial, with Big Fish Training. Michael and Emma Ewing run AI tools, the new models included, through real comms tasks, live. Register with Big Fish

Friday 9 October, 12:00: Outputs or outcomes, with Stephen Waddington. Our rescheduled Leader Interview asks how comms should be judged once AI does more of the producing. Free and recorded; 16 September registrations stand and a new link comes by email. Register on Eventbrite

Elsewhere: Thursday 15 October, 14:00 to 17:00 CET: AI for Public Affairs masterclass. The Public Affairs Council’s three-hour virtual seminar on taking AI from individual use to team-wide adoption; from €295 plus VAT. Details

Talk it through. A Consult Comms With AI session is 45 minutes with Michael and a written advisory note you can act on. Book a session