Using LLMs to Prospect Sponsors for a Football Club

Most football clubs below the Premier League have one part-time commercial manager doing the work of three full-time roles. They write the deck, they cold email, they chase signatures, they renew sponsors, they activate matchday hospitality. The bottleneck is almost never strategy — it is research time.

This is the single best place in club marketing for an LLM workflow to pay for itself in the first month. Below is the workflow we have built and refined for clubs across the EFL and National League pyramid.

The problem this workflow solves

Cold sponsor outreach has two failure modes. One: generic emails sent to 200 companies with a 0.4% reply rate. Two: beautifully personalised emails sent to 15 companies because that was all the manager had time to research. Neither builds a pipeline.

A properly built LLM workflow turns that 15-emails-a-week ceiling into 60–80 personalised emails a week, with a reply rate that holds at 6–11% because each one is genuinely tailored.

The 5-stage workflow

Stage 1: build the prospect long-list

Pull 200–400 local businesses from a structured source. UK examples:

• Companies House filtered by postcode and SIC code.
• The local Chamber of Commerce member directory.
• A scraped list of recent planning applications (signals growth, signals capex appetite).
• Sponsors of any other club in your league (they already buy this category).

Output: a Google Sheet with columns company_name, website, companies_house_id, linkedin_url, postcode, industry.

This step is data engineering, not LLM work. Spend the £40 once on a freelance data person if you have to.

Stage 2: enrichment with an LLM

Now the LLM earns its keep. For each row in the sheet, run a single prompt:

You are a commercial researcher for [CLUB NAME]. For the company below, return a JSON object with these fields: industry_specific (one phrase), employee_estimate (number), sponsorship_history (any other clubs/teams they sponsor based on public info), decision_maker_likely_title, personalisation_hook (one specific recent fact about them — a new office, a hire, a product launch, an award), disqualified (true/false with reason).

Company: [COMPANY NAME], [WEBSITE], [LINKEDIN], [POSTCODE].

Run this through GPT-5 (best at structured output) or Claude Sonnet 4.5 (cheaper, marginally less reliable on JSON). Either works.

Cost: roughly £6–£14 to enrich 300 prospects. Time: a 20-minute batch.

What you now have is not a list of names. It is a list of qualified prospects each with a one-sentence reason to reach out this week.

Stage 3: tier and prioritise

Drop the disqualified rows. Sort the rest into three tiers using a simple LLM prompt over the enriched sheet:

• Tier A (top 30): strong personalisation hook + plausible budget + no existing sports sponsorship.
• Tier B (next 80): at least two of the three signals.
• Tier C (rest): add to a quarterly nurture list.

This is a 4-minute job for the LLM and saves the commercial manager from chasing the wrong companies.

Stage 4: drafted, not generic, outreach

For each Tier A prospect, run this prompt:

Write a 130-word cold email to [DECISION MAKER TITLE] at [COMPANY], a [INDUSTRY] firm in [CITY]. Reference this specific hook: [PERSONALISATION HOOK]. Propose a [PACKAGE] sponsorship at £[AMOUNT]. Mention our average attendance ([NUMBER]) and IG reach ([NUMBER]). End with a single soft CTA — a 20-minute call next week. Tone: confident, dry, not corporate. UK English. No exclamation marks.

The output goes into a draft queue, not your outbox. The commercial manager spends 3–4 minutes per email tightening the personalisation, fixing the CTA, and sending. Total: 30 personalised emails in roughly 90 minutes — a job that would have taken two full days manually.

The deeper sponsor pitch document, when those replies start coming in, is covered in our football sponsorship deck template.

Stage 5: log, follow up, iterate

Every reply (and every non-reply) goes back into the sheet. After 4 weeks of running this workflow you have a small but real dataset on which industries reply, which hooks work, and which package prices get traction. Feed that back into the Stage 2 prompt.

The agent gets sharper every month. The manual workflow does not.

The full stack we recommend in 2026

• LLM: GPT-5 for the structured stages (2, 3), Claude Sonnet 4.5 for outreach copy (4).
• CRM and sequencing: GoHighLevel or HubSpot — anywhere you can drop the drafted emails into a sequence with a human approval step.
• Sheet: Google Sheets is fine. Don't over-engineer.
• Optional: the Football Marketing Agency app for the pipeline view if you want sponsorship and content in the same dashboard.

What this is worth

For a typical National League or League Two club, replacing 20 generic outreach emails a week with 60 personalised ones consistently produces 2–4 incremental sponsor conversations a month and 1–2 incremental signed deals a quarter in our experience. At average package values of £4–£12k, the workflow is paying for the entire LLM and CRM stack many times over within the first signed deal.