Welcome back to The Customer Continuum. Issue #32, and chapter two of the champion series.
Last week’s build was the scoring agent, the one that reads your customer base and hands back a ranked shortlist of who’s worth pursuing as a champion. This week is the harder half of the same problem: what you do once you have the names, when starting the relationship is the actual work. And the clearest way I know to explain it is to tell you about the time I had to do it for a product that was quietly churning.
The advocacy work I’m proudest of isn’t a single program but the blueprint my team and I built for standing up advocacy at scale across Adobe’s portfolio, the model that became Advocate Nation. Champions was the cornerstone, but it was one piece of a much larger engine: user groups, the customer reference program, community, lifecycle and retention programs, peer reviews, voice-of-customer, and a customer advisory board, all working together as one system rather than a scatter of disconnected initiatives. It started at Marketo, with an established champions program we reinvigorated and grew, and the real test came when we had to prove the whole model could work somewhere it had never existed.
That somewhere was Adobe Analytics, and it wasn’t a clean assignment. Analytics was struggling. Adoption had gone stagnant, churn was a real problem, and both customer success and product came to us with the same question: how do we build the Marketo-and-Adobe story here, how do we drive the deeper adoption this product needs. The answer we landed on was to rebuild the whole Advocate Nation engine inside Analytics, and we started with champions, before anything else, because we already had customers who loved the product. They just needed to be harnessed in the right way. The product was suffering, but it was ripe for advocacy.
The problem was that I was starting completely cold. No existing program, no base, no awareness. Nobody was going to apply for a program they’d never heard of, so it had to be nomination-based from the ground up, which meant I had to figure out who the real champions even were before a single one of them knew the program existed.
I did it by hand. I sat down for informational interviews with customers one at a time, and I pulled context from product, from customer success, from anyone across Adobe who actually knew these accounts, trying to assemble a picture of who was worth pursuing and how to reach them in a way that wouldn’t feel like a canned response. It took weeks. Champions became the cornerstone, and the user groups and the rest of the engine grew out from there, until what started as one replicated program was the full Advocate Nation model running inside a product that had never had it. Sourcing champions from a base that already knows you is one kind of problem. Building a program from scratch for a struggling product nobody’s heard of is a much harder one, and it’s nomination and outreach the whole way.
That cold start is the situation most customer marketers are actually handed, so it’s the one worth automating. This week’s build does by machine what I did by hand at Adobe Analytics: it gathers the scattered signal from customer success, product, and community, decides who’s worth pursuing, and drafts the outreach to reach each person individually.
What I built
The Champion Outreach Agent takes the shortlist from last week’s scoring agent and turns each name into something you can act on. It runs in three stages per candidate.
First it synthesizes the scattered signal. The information you need on any one customer lives in four different systems: usage and health in the customer success platform, advocacy history in the CRM, community behavior in the community tool, expansion and executive sponsorship in sales. The agent pulls all of it into one short profile per person, and where a piece is missing it says so rather than guessing. That assembly is the week of manual work I did at Adobe, compressed.
Then it drafts a nomination brief, the short internal case an account owner reads before any outreach goes out: why this person, what specific strengths they bring, what role they’d play in the cohort, and every gap you still need to fill.
Then it drafts the invitation itself, personalized to the real person. It references something that customer specifically did: the user meetup they run, the review they wrote, the questions they answer for other customers unprompted. The test it holds itself to is whether you could swap in a different customer's name and have the invite still read true, and if it would, the draft failed.
That whole shape, the agent drafts and the human decides, is the same spine as last week’s scoring agent, moved one step further down the funnel.
The honest run
I ran the agent cold against a slice of eighteen synthetic accounts for a fictional analytics platform, the same customer set behind last week’s scoring work, with nothing but the spec and the data.
The synthesis held on all eighteen. Pulling scattered signal into a clean profile is mechanical, and the agent did it every time without inventing anything. The personalized invitations held too, as long as there was real history to reference. For the customer whose director already runs a local user meetup and moderates a sub-forum, the agent wrote an invite that named both and framed the ask around the community work she was visibly already doing. You couldn’t have sent that invite to anyone else. For the customer who quietly mentors new users in the community without being asked, the agent wrote something specific and warm, and the brief behind it told the account team the honest internal truth, that this account is worth recruiting for advocacy rather than revenue, because there’s no room for the account to grow.
It also caught a trap I was watching for. One account was a regulated bank that can’t advocate publicly at all. The agent didn’t send it the standard speak-and-review ask. It switched to a private-track invitation built on internal enablement and product feedback, with nothing that would put a compliance-bound customer in a bad position.
Then it got to the accounts where it stopped cold, and those are the ones worth your attention. The three highest-value accounts in the set were big, executive-sponsored, and hungry to expand, and not one of them had any advocacy history. The agent refused to draft an invitation for a single one. On the biggest account it was blunt: there’s nothing specific to reference, so any invitation would read generic, and the account should go to an informational conversation first to learn whether a champion actually exists inside it. It would rather hand the account back to me than send a hollow ask to a customer that valuable.
That refusal is the whole issue. The agent can write a genuinely specific invite when the relationship is already in the record. It cannot manufacture a relationship that isn’t there yet. Those high-value, no-history accounts are exactly the Adobe Analytics situation, the ones where I had to pick up the phone and have a real conversation before any outreach was worth sending. The agent found that boundary on its own and stopped at it, which is better than drafting twelve confident invites and letting me discover the hollow ones after they’d gone out.
Then it did two things I didn’t expect, and they’re the reason this run is worth more than a clean one would have been. The first was about my program design. The two strongest advocates in the whole set, the customers who’d given the most, were the two the agent flat refused to send to. Their invitations were the most specific ones it drafted, and it held them back anyway, because both people work at agency or partner accounts, and my program design never actually says whether a partner employee can be a customer champion. So the agent set both aside and asked a human to settle the rule before either invitation went out. It had found a hole in my own blueprint, on the two names I could least afford to get wrong.
The second was about my data, where not one advocacy act in the whole file carried a date. Every review, reference, and talk sat there with no record of when it happened, so the agent pointed out that every invitation it had just written quietly assumed the advocacy was recent. Thanking a customer warmly for a review they wrote three years ago and have long since forgotten lands worse than saying nothing, so it told me it couldn’t vouch for how recent any of this was and asked me to check before anything went out. That’s one gap in my program and one in my data, both caught before either could reach a customer.
So the human stays involved at a clear and predictable point. The agent drafts the outreach for the people you already have a relationship with, it sends the strangers back to you for the first conversation that only a person can have, and along the way it points out the gaps in your program and your data before any of them reach a customer.
What this means for your team
This matters for you even if you’re nowhere near building agents yet.
Most customer marketers aren’t handed a beloved program with a warm base lining up to reapply. They’re told to build one somewhere the company has no relationships yet, in a new segment, a new product line, a new region, with a customer base they’ve never met, and often for a product that’s struggling and needs the advocacy most. That’s the Adobe Analytics problem, and it’s the default, not the exception.
When you’re in that situation, the work is nomination and outreach: figuring out who the real champions are before anyone’s heard of your program, and reaching each of them in a way that proves you actually know them. The lesson from this build is where the line sits between what you can systematize and what you can’t. You can automate the gathering, the synthesis, and the drafting for the people whose history is already on record. You can’t automate the first conversation with someone you don’t know yet, and you shouldn’t try, because a generic invitation to a customer you’re trying to win is worse than no invitation at all.
The version worth building does more than draft. It reads your customer data and your program design closely enough to catch what’s missing, an advocacy act with no date, a rule your program never wrote down about who’s even eligible, and it stops to flag those before they reach a customer. The drafting saves you time, and the catching saves you from a send you’d regret.
So build the outreach system for the relationships you already have, and protect your own time for the ones you still need to build. That’s the guidance whether you’re running one agent or none.
What’s next
Chapter three moves from reaching champions to running them: the charter they sign, the tiers they move through, and the annual renewal that decides whether the program compounds or fades. The tier and renewal machinery already exists in the kit, and next week connects the outreach engine to it, so a nominated candidate has somewhere to land.
This week’s free starter: the Champion Outreach Agent
The free starter is the agent from this issue. Paste it into a fresh Claude Project’s custom instructions, hand it your scored shortlist plus whatever context you can pull per candidate, and it returns synthesized profiles, nomination briefs, and personalized invitation drafts, with a human-judgment section that names every candidate it wouldn’t draft for and why.
It’s built to sit directly behind last week’s scoring agent, which tells you who’s worth pursuing while this one drafts how you reach them. Together they cover the front of the champion funnel, from a raw customer list to an invitation you can actually send.
Download it at the link below and save it to your Copilot pack alongside the scoring agent.
For paid subscribers
Here’s exactly what you’re getting this week and where to find it, no digging required.
Everything for this build lives in one folder: the Champion Program Toolkit. I’ve just added this week’s outreach agent to it, so the whole champion funnel now sits in one place. Open the folder, start with the file named Start Here, and it walks you from the scoring agent to the outreach agent to the Blueprint to the Charter in order. You don’t have to hunt through the 135-agent kit and assemble it yourself. That’s what the folder is for.
What’s inside: the scoring agent that ranks your base, the outreach agent that drafts the nominations and invitations, the full Champion Program Blueprint with the 100-point rubric the scoring agent runs, and the fillable Charter you put in front of the champions you land. The Start Here guide connects them and points you to the one agent already in your kit that runs the program after an invitation is accepted, Agent 20, the Champion Program Agent, which handles the tiers, the benefits, and the annual renewal.
By the end of this week you can have the front of the funnel running: score your base, draft the asks, and formalize the champions who say yes.
Three folders in your paid Drive bundle:
Next week, chapter three: the charter, the tiers, and renewal.
— Kevin
P.S. If you’ve ever been handed a program to build somewhere you had no relationships yet, forward this to one customer marketer who’s staring down the same cold start. That’s how this newsletter grows.