How Haymarket’s AI Tool ‘P3ter’ Addresses Pre-RFP Intelligence

Generative AI has made it easier to process and publish information, intensifying competition and flooding the market with data. For publishers, this presents both a challenge and an opening: they need to differentiate their content from automatically generated coverage and use AI in a way that uniquely addresses their audience's needs.

At FT Strategies, we’re seeing a growing number of publishers explore AI not as a replacement for journalism, but as a tool to extend it. Haymarket’s ‘P3ter’ offers a practical example of that approach in action.

Haymarket Media Group, a UK-based specialist publisher with more than 70 consumer and B2B titles including Campaign, PRWeek, What Car?, and Autocar, has long been recognised for its depth in professional and specialist markets. Within its Partnerships Bulletin brand, the company spotted an opportunity to use publicly available data to enrich its intelligence product. The aim was to address a specific need from its subscribers: identify emerging Public Private Partnership (PPP) opportunities before official tenders are announced.

From Customer Conversations to Concept

The idea for what became “P3ter” (short for “P3 Technologically Enhanced Research”) came from customer interviews, where the Partnerships team asked two key questions:

• How could we make the Partnerships brand more useful?

• What else would you like to see from us?

The most consistent answer was a  pre-RFP alert system that could flag likely PPP projects before they appeared on procurement portals. The team realised that while traditional reporting and research remained essential, the growing volume of online information made it possible to complement those methods with data-driven products.

Building ‘P3ter’

P3ter - short for P3 Technologically Enhanced Research - was developed in partnership with TMotions Global Limited, Haymarket’s technology partner. It consists of a set of tools and processes that help the editorial team detect relevant signals in public sources.

P3ter scans thousands of online sources, structures and filters the information, and surfaces only the most relevant leads, keeping humans in editorial control. 

It is developed through the following steps:

• Automated scraping and categorisation of online content

• Identification of new and non-performing URLs

• Regional and thematic tagging

• Publication of a bi-weekly feed of new project opportunities

Through continual refinement, the turnaround time for surfacing leads has dropped from up to six months to around two days.

P3ter enables the processing of data that would be unfeasible to analyse manually: as an example,  two US state governments publish about 1,500 master URLs and 300,000 sub-URLs. Each week, the system identifies ~5,000 new or updated pages, which are then processed through keyword filters and audio transcription tools. Approximately 50 qualified leads make it through to the editorial team each cycle, becoming story foundations and updates published on partnershipsbulletin.com.

“It’s not humanly possible to research hundreds of thousands of websites weekly. AI has made this possible,” said a Haymarket representative.

Lessons from the Build

The project started with a freelancer testing early prototypes. As the team learned more about the scale and complexity of the task, it became clear a more robust system was needed. TMotions then helped develop and expand the framework to 10 states in the first phase of the project.

A few lessons stood out along the way:

• Finding the right data is the hardest part. Sifting through thousands of potential sources took time and editorial judgment.

• AI needs tuning. Algorithms for keyword matching and ranking required a process of refinement and human oversight.

• Change needs managing. Some initial hesitation within teams reflected a broader learning curve around AI.

Haymarket was transparent about the use of AI, and subscribers responded positively.

Impact and What Comes Next

P3ter has allowed the Partnerships Bulletin team to double the number of pre-RFP projects shared with subscribers while speeding up lead validation and expanding the breadth of intelligence it can offer.

Encouraged by the success of P3ter, Haymarket is now exploring whether similar AI-assisted workflows can be adapted for other specialist titles.

Takeaways for Other Publishers

The P3ter project is less a technology showcase than a study in how AI can complement editorial practice. Three observations stand out:

1. Start from audience needs. The project began with client conversations, not a technology brief.

2. Pair automation with expertise. Machines can find the needles; humans decide which ones matter.

3. Provision for testing and refinement.  The system evolved through iteration - not a single “big launch.”

Final Thoughts

B2B Publishing provides some of the clearest opportunities for AI experimentation:

• B2B content often relies heavily on structured and/or public data sources, enabling an automated analytical layer. 

• Niche topics can command higher subscriptions but discourage competition, as the addressable market is limited and already served by established providers with dedicated sales teams. GenAI, however, lowers the cost of entry and enables low-cost niche-focused competitors, putting more pressure on publishers to differentiate. 

• B2B audiences have a more functional relationship with content: regular qualitative interviews can often yield good ideas for new products.