Member Exclusive: 4 questions to ask your PR agency before the next renewal

Agencies are restructuring around AI faster than many retainers reflect. The renewal conversation must change.

By Stephanie Nivinskus, principal, Ragan’s Center for AI Strategy

Your next renewal conversation with your PR agency cannot sound like the conversations you’ve had in years past. Scope and budget drove many of those conversations. While still valid, the bigger question now is one many agencies have been avoiding: who is doing the work, AI or human talent?

These four questions expose what matters in an AI-powered world: how the agency measures AI visibility, how it handles client data, how AI shows up on your specific account and how the team is being trained. Meiko Patton, an Advisor with Ragan's Center for AI Strategy, has more than 30 years of corporate communications experience with the United States Postal Service. Here is what she said when I asked for her thoughts.

1. How is your agency measuring AI visibility, and what is it doing about it? 

Clients should ask whether the agency is tracking how the brand appears in AI-generated answers in addition to traditional media coverage. If ChatGPT, Perplexity, Gemini or Google's AI Mode omit the brand, misrepresent it or cite competitors instead, that is now a visibility and reputation issue.

A Peec AI analysis of 30 million sources cited across ChatGPT, Google AI Mode, Gemini, Perplexity and AI Overviews found the top-cited domains are Reddit, YouTube, LinkedIn, Wikipedia, Forbes, G2, Yelp, Facebook, Medium and Techradar. The platforms diverge sharply, with ChatGPT favoring editorial sources like Forbes and Techradar while Google's AI Mode leans on Facebook and Yelp. That reshapes earned media strategy. A monthly AI visibility audit should run across the major answer engines, log omissions and misattributions, and feed into placements AI models cite.

This matters before renewal because buyers now research with AI before they touch a search box. An agency still measuring success by traditional press hits without tracking the citation surface AI pulls from cannot credibly claim to be moving the brand's AI visibility.

2. What is your agency’s AI use policy on client data, model training and disclosure?

Clients should ask for the written policy and read it. Three things must be named, including how client data is handled (zero-retention API access or enterprise tiers with training opted out), which tools are approved for which deliverables and the threshold at which AI use is disclosed to the client versus kept internal.

A concrete example of a real policy: Claude Enterprise and ChatGPT Enterprise approved for client work, free-tier consumer tools are prohibited entirely and written client sign-off is required before AI is used in any externally distributed deliverable.

This matters before renewal because regulatory and contractual exposure (EU AI Act, state-level disclosure laws and NDA breaches via consumer tool prompts) sits with the client. A vague policy means the client is carrying that risk without knowing it.

3. How will AI be used on our specific account?

 Clients should ask for a per-deliverable workflow map: where AI drafts, where humans verify and where AI is excluded entirely. Generic answers do not qualify.

An example: AI used for media list research and pitch variation drafting; humans owning reporter relationships and quote generation; AI excluded entirely from spokesperson statements and crisis communications. Every line item on the statement of work should carry a clear designation.

This matters before renewal because pricing should reflect AI's role on the account. If AI is drafting work that used to take a senior associate three hours, and the retainer has not moved, the client is paying senior rates for AI output. The workflow map is the basis for that conversation.

4. How is your team being trained to work with AI?

Clients should ask what structured AI training the team has completed and request the artifacts: custom GPTs and Claude Projects built for client work, tool certifications, a named internal AI lead and the cadence of workflow reviews. "We use AI" is not training.

A team-wide Coursera link is not training. A practical signal of real capability: a library of custom GPTs, Claude Projects or Skills built for specific deliverable types (pitches, bylines, briefing memos), monthly internal reviews of AI-assisted work for accuracy and voice and a designated AI operations lead who owns tool selection and policy enforcement.

This matters before renewal because the cost of untrained AI use lands on the client, not the agency. Hallucinated statistics, fabricated quotes and brand voice drift appear in published work. The best available predictor of whether that risk is contained is what the team has built: the custom GPTs, the Projects, the review cadence.

The bottom line: what to do before signing

• Ask for the AI visibility audit. Request the cadence, the answer engines covered and how findings feed into earned media placements. 

• Ask for the written AI policy. It must name how client data is handled, which tools are approved for which deliverables and the threshold at which AI use is disclosed.

• Ask for the workflow map. Every line item on the statement of work should designate where AI drafts, where humans verify and where AI is excluded.

• Ask for the artifacts. Custom GPTs, Claude Projects, tool certifications, a named internal AI lead and the cadence of workflow reviews.

Interested in the future of AI and communications? Your Council membership includes free enrollment in Ragan's Center for AI Strategy. Contact the Council Concierge at [email protected] with questions.