What You’ll Learn

Most educators hate finding leads. James started his working life knocking on doors selling photocopiers, so he’s not romantic about it. In this session he installs an outreach agent team live and runs it end to end — find prospects, score them, enrich them, write the pitch.

By the end you’ll know what a real outreach pipeline looks like when agents run it, and why it isn’t the same thing as asking AI for a list of leads.

Why “Just Ask AI for Leads” Falls Short

Ask a model for leads and you’ll get somewhere to look, maybe a script, maybe a phrase to use on LinkedIn. What you won’t get is a system that finds people who are a genuine fit, checks whether they have the problem you solve, works out how reachable they actually are, and drafts something in your voice.

That gap is the whole point of running it as a team rather than a prompt.

Outreach Isn’t Only Sales

The most useful reframe in the session. James rebuilt his sales agent into a multi-target outreach team once he realised how many kinds of outreach a business actually needs:

• New customers and qualified leads
• Referral partners and affiliates
• Podcast guest spots, speaking, events
• Sponsors, media and press
• Community partners
• Win-backs, renewals, cross-sell and upsell on your existing list
• Testimonials and case studies

💡 In Plain English: The machinery for “find me five good podcasts” is identical to “find me five good customers.” Only the scoring rubric changes.

The Run, Step by Step

1. Install and onboard

James installs the plugin from Customize → Plugins → Add → Upload Plugin. On first run the team interviews itself into the business: who you serve, what you sell, your pricing, your voice. It only does this once.

Because the campus already holds shared context, most of it is confirmation rather than data entry. In the session it inherits the ICP directly and reports nothing to adjust.

2. Capability detection

The team checks what tools it actually has. In this run it finds a research scraper, web search and a connected CRM. Without those it falls back gracefully — web search instead of a scraper, a spreadsheet instead of a CRM.

✓ Check Your Work: If the onboarding doesn’t tell you which tools it detected, it hasn’t done capability detection. That step is what stops it inventing a workflow it can’t run.

3. Pick a target and run the scout

James asks for five podcasts to guest on. The team searches, scrapes deeper than a plain search would, and returns scored candidates with the reason each one fits.

A nice detail: it had already run that exact target earlier the same morning, remembered, and said so rather than repeating the work.

4. Scoring, and what gets rejected

The rejections are more instructive than the picks. A show with 50 million downloads scored down to 65 — strong, but hard to book, so not a first approach. Several large AI podcasts were filtered out entirely as anti-ICP because their audience was enterprise developers.

Fit beats reach. A big audience that isn’t yours is a bad booking.

5. Enrichment and access signals

For the shortlist it goes deeper: what the show covers, who guests, how big it is — plus the practical question of whether you can actually reach them. Public email? LinkedIn? A guest application form? A DM handle?

6. Briefs and drafts

Each candidate gets a one-page research brief — the show, why it’s an easy booking, the angle, your talking points, and the booking path with the real link. Then a pitch draft tailored to that show’s audience.

The best example in the session: for one host whose AI coverage was all “AI as a new side hustle,” the team proposed the opposite angle — not a new hustle, but the AI employee that runs the operations of a business you already have — and recommended stating that contrast explicitly rather than hoping the host noticed.

7. CRM write-back

With a CRM connected, the brief, the score, the fit reasoning, the pitch angle and the draft email land on the contact record. Without one, it all goes to a spreadsheet.

Nothing Sends

Worth being explicit, because it’s the design decision that makes this usable.

“Nothing’s been sent, these are drafts for your review edit before you pitch.”

⚠️ The agents do the finding, the qualifying, the research and the writing. You do the deciding — whether you actually want to do business with this person — and you send.

Key Takeaways

• Outreach is not just sales — the same pipeline handles partners, podcasts, press, sponsors, win-backs and testimonials by swapping the scoring rubric.
• Capability detection first: the team should tell you which tools it found and degrade gracefully when one is missing.
• Score for fit, not reach — a large audience that isn’t your ICP is a rejection, not a stretch goal.
• Include reachability in scoring; a perfect prospect with no public contact route is not an easy booking.
• Keep the send as a human gate — agents research, score and draft, you decide and send.

Your Next Step

Pick one non-sales target — podcasts, partners or win-backs — and define what a good fit looks like in one paragraph. That paragraph is the rubric the whole pipeline runs on.

The Campus AI OS is free and the outreach team drops into it: trainingsites.io/os