Part of The AI Communications Hub · part 2 of the Running AI Tools Inside a PR Agency series
Edited on Oct 7, 2026.
A PR team can go from no AI tool use to a working daily habit in one week, structured as one focused session per day rather than a single long training that nobody retains. This guide lays out the five-day structure, what to cover each day, and the two mistakes that most commonly derail a team's first week with a new tool.
A single two-hour training produces a burst of enthusiasm that fades within days, since nobody retains a tool's full range of uses from one sitting without practicing it against real work. Spreading the rollout across a week lets each day build on the last with the team's own actual tasks, which is what makes the habit stick past the first week rather than requiring a second training later to re-teach the same material.
Day one is account setup and the single most common task, not a tour of every feature. Get every team member logged into the approved tier, confirm training is turned off where the tool allows it, and walk through one task the whole team already does daily, drafting a press release or a pitch email, using the tool together in real time. Ending day one with everyone having produced one real draft, not a demo draft, is the actual goal.
Day two introduces prompt specificity using the team's own recent work as examples. Take two or three real prompts from day one that produced weak output and rewrite them together as a group, showing exactly what changed and why the output improved. This is more useful than an abstract lesson on prompt engineering, since the team sees the improvement against work they actually care about.
Day three covers the data-handling policy directly, not as a compliance formality but as a working session. Walk through the agency's actual AI use policy line by line, and run one real example of genericizing a client-related prompt together, since this is the single skill most likely to prevent a real confidentiality mistake later. See AI and Client Confidentiality for the rules to walk through.
Day four introduces the second tool, if the agency is rolling out more than one, and the specific task split between them. Show the team exactly which task goes to which tool, drafting in one, research in another, using real examples from the week so far rather than a generic comparison chart. For the underlying logic behind this split, see Three Bots. Three Jobs. One PR Team.
Day five is a review session built entirely from the week's actual output. Have each team member bring their best and worst result from the week, and discuss as a group what made the difference. Set one standing rule before the week ends: every AI-drafted piece gets a human review before it ships, stated explicitly rather than assumed, since this is the rule most likely to erode quietly once the initial training excitement fades.
The first mistake is training on hypothetical tasks instead of the team's real work, which produces a training that feels useful in the room and evaporates the moment someone returns to an actual deadline. The second is skipping the policy conversation because it feels like a separate compliance topic, when in practice the confidentiality mistake is the most costly failure mode and the easiest one to prevent with fifteen minutes of direct practice.
A team can build a working daily habit in one week if each day uses the team's own real tasks rather than generic examples, and if the structure spreads learning across five short sessions instead of one long one.
After. The vendor and tool decision should already be settled, since a team learning a tool that changes mid-training loses the momentum the structure is built to create.
That every AI-drafted piece gets a human review before it ships. Stating this explicitly during training prevents it from eroding quietly once the excitement of a new tool fades.
Roughly 45 to 60 minutes is enough for most days, since the goal is one focused skill built on real work, not a comprehensive feature tour.
Sarel Doglu is a technology and product executive with more than 25 years of experience building software and leading technology teams. He is a technology and marketing expert.
Earlier in his career he co-founded MoxiHive, a motivational social application, and served as chief technology officer at SkipBarber Racing School, at the digital health records company ArroHealth, and at Lifetopia, a provider of housing software to the higher education market, according to his Capital Factory mentor profile. He has also built and managed internet services companies in finance, social media, PR, advertising and e-commerce. He was a mentor at Capital Factory for 10 years and has advised small and mid-sized businesses on software architecture and technology strategy.
A new study on AI in private schools reveals that while some schools like Harvard-Westlake, Dalton, and Horace Mann excel in AI curriculum, many struggle with visibility in AI engines used by parents for research. This poses a significant communications challenge for admissions.
FleishmanHillard's Intel program is the canonical modern tech-launch playbook. Inside the disciplines that make a major hardware launch work — reviewer relationships, benchmark methodology, analyst relations, retailer editorial, trade press — and what brands beyond Intel are worth studying.
Duolingo learned the hard way. The rule is simple — build AI-first inside, sell outcome-first outside. And run the LLM check before you ship.
James Cadwallader, co-founder and CEO of Profound, built the AI visibility leader to a $1.8B valuation. Background, funding, platform and lessons for PR teams.
Twenty-five mobile game marketing campaigns from 2012 to 2025, from Clash of Clans' Super Bowl ad to Monopoly GO's $6B in 1,275 days, each with a named source for its result and the marketing move behind it.
The best press release distribution services ranked by AI citation share across five wires, with vertical winners, budget options and how to choose one.
EPR publishes the data every week.
Free. Weekly. Unsubscribe anytime.
