Most podcast outreach for coaches is either invisible (one careful pitch a week, sent by hand) or insulting (a thousand template emails a month, sent by a VA, none of them read). Neither produces bookings. The first runs out of hours; the second burns the sender's reputation faster than a single booked episode could justify.
There is a third path. It is what we built for a speaking coach over a single campaign window: 2,400 outreach emails sent, a 15% reply rate (3x the industry average), 55.56% of those replies expressing genuine interest, and a projected pipeline of $600k to $1M from booked appearances. The system runs on a workflow tool, an outbound platform, a language model, and a CRM. None of the components are exotic. The wiring is the work.
This post is the engineering view of that build, and a case for why podcast outreach automation is one of the highest-leverage outbound moves a coach can make. If you want the broader system view of how outbound fits into a coaching business, the AI automation playbook for coaches is the pillar that covers it.
Why podcast outreach is different for coaches
A coach selling £1,000 to £10,000 programmes does not need volume in the marketing channel. Ten thousand cold visitors a month is not the goal. Ten genuine conversations with the right host, in front of the right audience, will move more pipeline than any volume play.
That makes podcast outreach a near-perfect channel for high-ticket coaching:
- The audience is captive, attentive, and pre-qualified. Listeners have chosen this show, this host, and this topic. They are not scrolling.
- The trust transfer is large. A host saying "this is the person you need to listen to" is worth more than any ad.
- The format is long enough to demonstrate depth. Forty minutes of you talking about your work is a sales asset that keeps producing for years.
- The audience is reachable afterwards. Listeners who connect with you book calls, join lists, and apply for programmes.
The problem is the outreach. Doing it well at any meaningful volume is the kind of repetitive, judgement-heavy work that human assistants are bad at and AI is unusually good at. The output of a great VA at this is roughly 20 to 40 personalised pitches a month. The output of a well-built automation, with a human doing voice review and final approval, is closer to 400 to 600 a month, with personalisation that is actually personal.
That is the gap this post is about closing.
The case study: a multi-vertical podcast booking engine
The build was for a speaking coach with four distinct audience verticals: AI systems, spiritual coaching, business and entrepreneurship, and mompreneur. Each vertical has its own podcast ecosystem, host preferences, and tone of voice. The challenge was to run all four lanes in parallel without one bleeding into the other (a spiritual-coaching pitch landing on an AI-systems show is worse than no pitch at all).
The numbers from the campaign window are in the case study itself, and we cite them throughout below:
| Metric | Result |
|---|---|
| Outreach emails sent | 2,400 |
| Reply rate | 15% (3x industry avg) |
| Positive reply rate | 55.56% of replies |
| Pipeline generated | $600k to $1M projected |
| Audience verticals | 4 distinct lanes |
| Stack | n8n, Instantly, OpenAI, Notion, Calendly, PostHog |
The 15% reply rate is the headline number. Industry benchmarks for cold outbound at scale sit around 5%. The 55.56% positive reply rate is the more important number: it means the system is not just provoking responses, it is provoking the right kind of responses (genuine interest in booking, not "please remove me").
What the system actually does
The system runs five jobs in sequence. None of them are novel in isolation. Each one matters because the next one depends on it.
1. Identify qualified podcasts in each vertical
Starts with a sourcing layer that pulls candidate shows from podcast databases (Listen Notes, Podchaser, public RSS feeds), filters for audience size and recency of episodes (a show that has not published in six months is dead), and tags each one with the vertical it belongs to. Shows that match more than one vertical get tagged in both lanes.
The qualification filter is what makes the rest of the system work. A pitch list of 600 lukewarm shows does worse than a pitch list of 200 well-qualified ones. The filter is unglamorous work and most outreach systems skip it. We did not.
2. Craft host-specific pitches that reference real episodes
This is where the language model does the real work. For each qualified show, the system pulls the host's name, the show's premise, the most recent two or three episodes, and any guests the host has had on recently. The pitch is then drafted with explicit references: "I noticed your conversation with [recent guest] on [topic] last month and the angle you took on [point] resonates with how I work with my clients on [related theme]."
The reference is not decoration. It is the part that signals the email was written by someone who actually listened to the show, which is the bar that distinguishes a real pitch from a templated one. Hosts can tell the difference within the first sentence; the open and reply rates make this obvious in the data.
The voice the model writes in is calibrated per vertical. Spiritual coaching pitches sound nothing like AI systems pitches, and neither sounds like the mompreneur lane. We covered why lane-aware voice matters in the AI lead follow-up guide: the wrong tone in the right channel is worse than no message at all.
3. Send through a sender platform with proper warm-up and deliverability
The pitches go through Instantly as the sending layer. Instantly handles sender warm-up, IP reputation, deliverability monitoring, and inbox placement, which is the difference between landing in the host's inbox and landing in spam.
Skipping this layer is how most coaches damage their domain. A few hundred pitches a month sent through a normal Gmail account, even if every one is well-written, will trigger spam filters quickly. A dedicated outbound platform is not optional once volume crosses a threshold. We covered Instantly in more depth in the best AI tools for coaches list.
4. Classify every reply with AI
Every reply that comes back is read by a language model and classified into one of a small set of categories: positive interest (book the call), positive but later, soft no, hard no, out of office, removal request, and edge cases (not the right host, suggesting another guest, asking a clarifying question). The classification fires the next step in the workflow.
The classification model is trained on the coach's preferred response patterns and the lane-specific voice. A "positive but later" reply in the spiritual coaching lane gets a different follow-up than the same reply in the business lane.
5. Draft on-brand responses with lane-aware voice
For positive replies, the system drafts the response back to the host with the right tone, the right Calendly link, and the right context from the original pitch. The coach reviews and approves the draft (a one-click step in Notion), and the response goes out.
For non-positive replies, the system handles them automatically: removal requests get processed, soft nos get marked for a 90-day re-engagement window, hard nos get archived, edge cases get flagged for the coach to handle directly.
The result is an inbox that does not need to be touched on most days. The coach sees only the replies that need their judgement, not the long tail of administrative responses.
What makes this work that most outreach systems get wrong
The build is not technically exotic. The components (n8n, Instantly, OpenAI, Notion, Calendly, PostHog) are all available and reasonably priced. The thing that makes the system produce the numbers it does is the calibration, not the components.
Personalisation that is genuinely personal
Most outreach automations claim personalisation. What they actually do is mail-merge the host's first name and the show's title into a template. The reader can tell within the first sentence. A real reference to a real episode, with a real opinion attached, changes the response rate by an order of magnitude. The 15% reply rate in this build is doing exactly that work.
Lane-aware voice
A four-vertical campaign run with one tone-of-voice template would have produced a fraction of the reply rate. The four lanes are written by the same coach but with materially different language registers. The model is given vertical-specific style guides; the output sounds like four related but distinct voices, each tuned to the audience that show serves.
Reply classification doing real work
The reply classifier is the part that lets the system scale. Without it, a 15% reply rate on 2,400 emails is 360 replies for the coach to triage by hand, which kills the whole point. With it, the coach sees roughly 30 to 50 replies a month that genuinely need their attention. The rest get handled by the system.
Volume calibrated to deliverability
The 2,400 emails over the campaign window is not a hard volume cap. It is the volume the system can sustainably send through Instantly without hurting the sender reputation. Pushing beyond that breaks the deliverability of every future email. Volume is a function of inbox placement, not just send count.
CRM-as-pipeline
The CRM layer is in Notion, and every interaction (sent, opened, replied, classified, responded, booked) writes back to the contact's record. This is what lets the system run as a pipeline rather than as a queue of emails. The coach can see at any moment where a given relationship is, what was last said, and what is next.
Common mistakes coaches make in podcast outreach
A short list, in case any of these are familiar from your own attempts.
Pitching to the wrong shows. Quality of the list matters more than the size of it. Two hundred shows that genuinely match your work outperform a thousand shows in adjacent niches.
Sending generic pitches. The host receives 50 pitch emails a week. Every one of them says the host's first name and the show's title. None of them reference a real episode. Yours has to.
Pitching once and giving up. The win rate on first pitch is small. The win rate after a one-touch follow-up two weeks later is meaningfully larger. Most coaches stop at the first pitch.
Damaging the sender domain. Sending hundreds of pitches a month through a personal Gmail account will mark you as a spam sender within weeks. Once that happens, even your warm replies start landing in spam. Use a dedicated sending domain through a proper outbound platform.
No pipeline view. Coaches who treat outreach as "send and hope" never see the patterns: which lanes work, which subject lines lift the reply rate, which hosts are higher-yield, which cadence converts best. The CRM layer is what makes the channel improve over time.
Trying to outsource it without a system. A VA without a system produces 30 pitches a month with mediocre personalisation. The same VA inside a system that handles the qualification, drafting, classification, and pipeline can review and approve 300 pitches a month at higher personalisation than they could ever write by hand. The leverage is in the system, not the headcount.
What this looks like as a build
The shape of a comparable build for a coach with one to four audience verticals:
The qualification layer pulls candidate shows from the podcast databases, filters by audience size, recency, and topical relevance, and tags by vertical. Owned by n8n, with sourcing fed by API integrations into Listen Notes and similar sources.
The drafting layer is OpenAI (or Claude) writing each pitch with a real episode reference. Lane-specific style guides are part of the prompt. Output goes into a draft queue in Notion for human review on day one (and unattended once the calibration is dialled in).
The sending layer is Instantly, with a dedicated sending domain warmed up over the first three to four weeks. Inboxes rotated to keep deliverability high. Daily volume capped at the level the platform can sustain without reputation damage.
The reply classification layer is the language model again, reading every inbound reply and tagging it. Tagged replies fire the next workflow step.
The response layer drafts on-brand replies for the coach to approve. Calendly link inserted automatically. Booking confirmations write back to the CRM.
The pipeline layer is Notion (or Airtable), with PostHog tracking the funnel: sent, opened, replied, classified, booked. The data shows where to optimise next.
A typical engagement runs 6 to 10 weeks from scoping to live, depending on how many lanes are in scope and how much voice calibration the coach wants to do up front. The first 4 to 6 weeks of live data are usually the calibration window; the system improves week-on-week as the coach approves or rejects drafted pitches and the model learns the voice.
How this fits with the rest of the coaching automation stack
Podcast outreach is one of the highest-leverage outbound moves a coach can make, but it is not the only system in the stack. The full picture, from inbound to delivery, looks roughly like this:
Outbound (this post) brings hosts and listeners into the orbit. The AI lead follow-up guide covers what happens when those listeners come back to your site or DMs. The discovery call automation guide covers what happens when they book a call. The coaching client onboarding pillar covers what happens after they sign. The client accountability system pillar covers what holds them in the programme.
Each layer feeds the next. Outbound that brings in the right kind of listener and then dies because the lead follow-up is slow has wasted the leverage of the channel. The reason we ship the layers together when we can is that the pipeline is end-to-end or it is broken.
FAQ
How many podcasts should I pitch a month?
For a single-vertical coach, 100 to 200 well-qualified pitches a month is a sustainable target with a working system. For a multi-vertical coach (the case-study shape), 400 to 600 across all lanes is achievable. Above that, deliverability becomes the bottleneck before pitch quality does.
What reply rate should I expect?
Industry benchmarks for cold outreach sit around 5%. The case-study build hit 15% with proper personalisation, lane-aware voice, and qualified targeting. Most coaches running this themselves see 8% to 12% in the first three months, climbing as the system calibrates. Below 5% indicates the qualification or the personalisation is broken; above 18% is unusual and usually indicates the volume is too low to be representative.
How long does it take to see bookings?
The first replies typically come within 48 hours of the first batch going out. The first booked episode usually lands within two to three weeks. Steady-state booking volume (a handful of episodes booked per month from the channel) is typically reached within 8 to 12 weeks. The pipeline impact compounds because each booked episode produces listeners who become leads for months.
Do I need a dedicated sending domain?
Yes, once volume crosses roughly 50 pitches a week. Sending from your primary domain (the one you use for client communication, payment notifications, contracts) is risky: if the outbound activity damages reputation, your transactional email starts landing in spam. A dedicated sending domain (e.g., yourname-pitches.com) isolates the risk. Instantly and similar platforms make warm-up and rotation straightforward.
Can I use this approach for direct outreach to clients (not podcasts)?
Yes, with some adaptation. The same system architecture (qualification, host-specific drafting, lane-aware voice, reply classification, CRM pipeline) maps directly to direct outreach into your ideal-client lists. The personalisation reference shifts from "I noticed your recent episode" to "I noticed your recent post / case study / hire / launch." The build pattern is the same; the prompts and the source data change.
What is the cost of running this?
Tooling cost for the case-study shape lands around £200 to £400 per month: Instantly subscription scaled to the sending volume, n8n self-hosted or low-tier cloud, OpenAI or Claude API usage proportional to pitch and reply volume (typically £30 to £80 per month at coaching outreach volumes), Notion or Airtable seats, Calendly subscription. The build itself is a per-project investment; the case-study build paid back inside the first booked engagement.
Why not use a podcast booking agency?
Agencies are a different model. The good ones charge £2,000 to £6,000 per month and book a handful of episodes per quarter, mostly through their existing host relationships. They are a good fit if you want to be fully hands-off and the per-month cost is comfortable. Automation is a better fit if you want a higher pitch volume, more control over targeting, and the pipeline data to live in your stack rather than the agency's. Some coaches use both: the agency for warm-relationship pitches, the automation for breadth.
Will the AI-drafted pitches sound like me?
If the voice is calibrated properly, yes. The model is given style examples from your past writing or speaking, lane-specific style guides for each vertical, and a review queue where you approve or edit drafts in the early weeks. After the calibration window (typically four to six weeks of live data), the drafted pitches sound enough like you that the review step becomes a five-second skim rather than a rewrite. If you skip the calibration window, they will sound generic; that is the mistake to avoid.
Where to take this
If you want the broader system view of how outbound fits into the rest of the coaching stack, the AI automation playbook for coaches covers it. If your bottleneck is what happens when these listeners come back to your DMs and form, the AI lead follow-up guide covers the inbound side. If discovery calls are the next link, the discovery call automation guide covers that layer.
If you want to see the case study this post is built on, the multi-vertical podcast booking engine walks through the architecture with the metrics in context. And if you want to see how the same pattern would map to your coaching business, see how Praxail works. The systems are bespoke, but the building blocks are the ones in this post.
