Your follow-up · Big Fish Training × Comms With AI · Wednesday 8 July 2026
AI for Account Managers & Directors
Thanks for spending the morning with me. This is your follow-up: a short recap of what we covered, and, more usefully, a curated set of links so the frameworks and templates keep working for you and your team long after today. Bookmark it, and lean on the website rather than your memory.
Start here
Three things to put to work this week
Govern templates & Manage toolkit
An AI Use Policy, Tool Register, Review Log and the quality checks, plus the ongoing governance layer for an AI-using team.
Open Govern & Manage →
Messaging frameworks
The Message House and Key Messages Grid, drafted from a brief in minutes, then made sharp by you.
Open the library →
The five-part prompt
Role · context · task · format · constraints, with placeholders for any facts. Then meta-prompt: ask AI to critique and rewrite your prompt.
Jargon, in plain English
Six terms worth being able to explain
You don't need the jargon to lead AI well, but a client or your board might ask.
Shadow AI
the unsanctioned tools your team is already pasting client data into.
Consumer vs enterprise
free tiers may train on what you type; enterprise tiers don’t, and add controls.
Agentic AI
AI that takes actions and completes tasks, not just chats. It changes the risk picture.
Prompt injection
a hidden instruction on a web page that hijacks an AI doing your research.
Synthetic media / deepfakes
AI-made fake audio, video and images. A question of when, not if.
Data retention & residency
whether a provider stores your inputs, and where. Procurement will ask.
What we covered
The morning, and where to go next
The frame
At your level AI is a strategic capability, not a shortcut: the question is how the account uses it well, defensibly, and to clients' advantage. As PRCA CEO Sarah Waddington put it, the goal is “building confidence, capability and trust, not just efficiency.”
Prompting, going further, and Projects
Beyond the five-part prompt, two moves do most of the work: get AI to interview you and rewrite your prompt, and manage long chats by summarising and restarting rather than letting them drift. The bigger lever for a team is Projects (or workspaces): not one long chat, but a reusable space that holds your files and standing instructions, with house style and the quality check built in, so good practice is the default rather than a personal hobby. I wrote up how I use them in How project folders supercharged my AI comms workflow (Applied / Comms With AI).
AI in campaign planning
AI widens and pressure-tests the options faster than a room can: synthesise the landscape, find the angle competitors haven't claimed, ask it for the weakest assumption in a strategy, and generate alternative framings before you commit. In the session I built a campaign live from a brief to show the end-to-end. For fast, verifiable research, Perplexity is strong because every claim links to its source, so you can check at a glance.
The accuracy habit: AI is confident even when it's wrong, and at your level an error that reaches a client or the press is your reputation, not the tool's. Verify every name, figure, date and quote. It happens for real: Air Canada was ordered to honour a refund policy its own chatbot invented (2024), and a New York law firm was fined after ChatGPT produced six court cases that did not exist (2023).
Audience insight & messaging
AI gets you to a draft framework fast; you bring the differentiation. The message house is the workhorse: a single core message at the roof, three supporting pillars, and proof points in the foundation. Build it, then adapt the roof message for each audience without letting them drift apart. One of the sharper senior uses is building an AI ‘target reader’, a sceptical stakeholder or critical reviewer, to pressure-test how a message will land before it goes out. I documented the approach in How to build an AI target reader (Applied / Comms With AI).
Client conversations about AI
The least-discussed area for agency seniors, in five parts: disclosure (proportionate transparency, have an answer ready), policy (short, clear, living), expectations (cool the hype, warm the sceptic), positioning (lead with judgement; you're governed, not winging it), and recovery (own it fast when it goes wrong). Two newer conversations are here too: clients using AI on your work and passing it off, best handled as a ways-of-working conversation at onboarding; and clients asking how to show up in AI answers (AEO or GEO), where it's early and shifting, but clear, credible content still does most of the work. The PRCA Green Paper warns of a “two-tier industry”; being demonstrably on the right side of that is a genuine pitch advantage.
A go-to line you can borrow if a client asks: "We use AI to work faster and sharper, but a person always owns the thinking and the final word."
Quality control
Learn the tells of AI-assisted work gone wrong (generic voice, contrast clichés, confident wrongness) and build a light five-point check into the existing workflow: voice, facts, tone, confidentiality, ownership. The practical way to keep standards high, and to make it a team habit, is shared project folders with your house style and the check built in. For meeting notes, tools like Otter and Granola beat raw transcripts, but always verify.
Leading adoption
Set expectations (what good looks like), take objections seriously rather than dismissing them, build AI into the workflow so it isn't optional, and hire and grow for judgement, not tool tricks. A shared prompt library and shared agents, strongest on enterprise tiers, are how good practice becomes team infrastructure. The PRCA now frames AI fluency as basic workplace literacy; that's the direction of travel.
What automation actually looks like
The question almost everyone asked. The jobs worth automating are the repeatable, tightly-briefed ones: a scheduled sweep that checks new content against house style and reports back; the monthly reporting pull, drafted into your template for you to sense-check; and the admin layer of notes, actions, inbox triage and file wrangling. Start with one narrow, repeatable job. The narrower the brief, the better it runs unsupervised; loosen it and a human comes straight back into the loop.
Risk & governance
A lightweight, living policy covers four things: data (what never goes into a tool), accuracy (how errors are caught before anything ships), disclosure (what you tell clients, and when), and ownership (a named owner and a review date). Know the difference between enterprise and consumer tools, and sanitise by default: draft in Word with [CLIENT] and [FIGURE] placeholders, let AI work on that, then put the real detail back yourself. Because the field moves monthly, keep the policy flexible and review it regularly.
The prompts
Copy, paste, reuse
Everything we used on the day, in one place. Replace anything in [SQUARE BRACKETS], treat the first answer as a draft, and iterate two or three rounds.
Let AI improve your prompt (meta-prompting)
Here is a prompt I am about to use. First ask me what is missing, then rewrite it as a stronger version with role, context, task, format and constraints. Prompt: [paste].
Manage a long chat: summarise and restart
Summarise everything important from our conversation so far as a short brief I can paste into a fresh chat to carry on.
Review your work against the five-point check
Review this draft against a five-point check: voice, facts, tone, confidentiality, ownership. List anything that fails, and suggest the fix. Draft: [paste].
Stress-test a strategy
Here is our campaign strategy. Act as a sharp, sceptical planning director. Tell me the three weakest assumptions in it, and give me one genuinely different strategic framing we have not considered. Be specific and constructive. Strategy: [paste].
Build a message house
You are a senior messaging strategist. From the brief below, draft a message house: one core message (the roof), three supporting pillars, and two to three proof points under each. Then adapt the core message for two distinct audiences: [Audience A] and [Audience B]. UK English. Mark anything to verify. Brief: [paste].
Build an AI target reader / critical reviewer
Be my AI target reader. You are [role, sector, seniority, what they care about, what makes them sceptical]. I will paste content; read it strictly as that person and tell me what lands, what does not, what you would push back on, and what would make you act. First, ask me anything you need to play the part.
Practise the disclosure conversation
Act as a sceptical-but-curious client asking how my agency uses AI on their account. Ask me one question at a time, probe gently on confidentiality and quality, and react to my answers. After a few rounds, tell me where my answers were reassuring and where they were not.
Draft your team AI principles
Help me draft a one-paragraph set of AI principles for my comms team. Ask me three quick questions first (what we mainly use AI for, our biggest worry, who should own this), then give me a short, plain-English statement covering what good use looks like, what is off-limits, and who owns it.
Your 30-day plan
One client conversation, one team process, one experiment
Pick three things to put in motion this month. To get started, paste this into your AI tool:
You are my AI adviser for leading a communications team. I've just done a senior AI training course. Help me build a 30-day plan with three commitments: one client conversation to have, one team process to change, and one thing to try myself. Ask me two short questions about my team and our clients first, then give me the plan with a first action for each.
Coming up this Friday
Beyond the prompt: a practical framework for AI adoption in PR and comms
A Big Fish lunch and learn with Emanuela Giangregorio, a PRCA trainer whose practical framework for AI adoption expands on the themes from today. Friday 10 July 2026, 1pm London, and it's recorded for anyone who signs up, so it's worth registering even if you can't make the time.
Meeting ID: 852 9662 0139 · Passcode: 293397
Worth a read, and a few sources to keep up
Stay current on twenty minutes a week
Worth a read
- →PRCA AI Green Paper (2026): the profession's reference for client and board conversations.
- →Applied / Comms With AI: the ICP target-reader and project-folders pieces above, plus the Leader interviews and the State of AI in Comms.
A few sources to keep up
- →The Comms With AI Monthly (sign up below) and Applied: tested prompts and tool reviews, built for comms.
- →CIPR and the PRCA: the profession's guidance on using AI responsibly.
- →Wadds Inc. (Stephen Waddington) and PRWeek: sharp, ongoing reads on how the industry is actually using AI.
Comms With AI Monthly
One email a month. New templates, tested prompts, one tool review.
The easiest way to keep the momentum from today. Free, no spam, unsubscribe anytime.
Take it further
Want your team's AI practice sorted?
When you want it done with you, Consult Comms With AI is a working session plus a written advisory note. To deploy AI well across an account or the agency, Deploy and Manage are how we do that.
Michael MacLennan · michael@faur.site · commswith.ai · Hosted by Big Fish Training
Your follow-up from the session on 8 July 2026. Free to use, adapt and share.