Quick summary: To build a customer advisory board more effectively, use AI agents to surface hidden advocates, flag at-risk relationships, and dramatically improve your invite conversion rates.
When I took over our Customer Advisory Board program earlier this year, I inherited a blank canvas with a deadline.
The mandate was to stand up a CAB that would give ChurnZero’s executive team real, unfiltered insight from the CS leaders who know our platform best.
We needed to choose approximately 20 members: strategic thinkers with senior titles who wouldn’t be afraid to tell us what we’re doing right, and wrong.
Straightforward, right?
We already knew the right invitees… or did we?
To build a customer advisory board, you start with a list. Most of the time, it comes from intuition: CSM nominations, executive relationships, or the accounts that show up in every reference conversation.
However, this entirely reasonable starting point comes with three structural blind spots. Most programs never solve them, and I didn’t want to replicate them.
- Visibility bias. You invite the customers you already know. Advocates who haven’t been on a reference call, haven’t been flagged by their CSM, or haven’t made noise tend to go under the radar and don’t make the list.
- Hidden risk. Manual review can’t detect when a customer relationship is quietly fraying. Send an invitation to someone whose sentiment has been shifting downward for months and you’ll erode goodwill, not build it.
- Over-indexing on enthusiasm. Enthusiastic customers are valuable, but they’re not always the most strategic voices in the room. The danger is that if your CAB skews toward fans instead of nuanced critics, it will simply validate your direction.
The solution: a two-layer, human-AI selection framework.
We built our CAB selection framework with two distinct sets of filters. The first was “standard” and structured; the second was AI-powered and signal-based.
Layer 1: standard attributes
Our first layer provides the best-practice baseline required by any rigorous customer advisory board selection process. We filtered to identify influential, leadership-level candidates within specific market tiers, product editions, customer segments, and ARR.
Next, we scored each qualified candidate against a 100-point rubric across six weighted categories: strategic account value, role seniority, customer health, industry influence, product adoption, and diversity of segments and geographies.
We determined that candidates who scored above 90 would be considered must-haves, while those who scored under 80 would be “not yet.” In between, we would exercise our judgment.
Layer 2: AI Agent signals
The first layer above is as deep as most CAB selection processes get. If your selection process stops after layer 1, you’re only seeing the empirical, product-focused data, leaving you blind to qualitative insights.
Luckily, we had another layer of understanding to apply: a handpicked team of four ChurnZero AI agents, each built to surface a specific kind of nuanced signal. For this team, we picked:
- Herald for assessing advocacy readiness. Herald analyzes engagement history, sentiment patterns, and reference behavior to assign each contact a status: Advocate Ready, Advocate Potential, Neutral, or Do Not Solicit. Note: that last category matters as much as the first.
- Pulse for evaluating participation depth and engagement focus. Pulse was there to help us distinguish strategically engaged contacts from those who are simply active. This doesn’t show up in login frequency but emerges in behavioral patterns over time.
- Vibes for emotional sentiment signals. A sentiment shift is often the earliest signal by far that a relationship is moving positively or negatively. Vibes surfaces latent enthusiasm that health scores or CSM notes might miss, and flags deteriorating relationships before they become visible problems.
- Harbinger for a critical eye on hidden risk. Harbinger identifies behavioral warning signs. It scans for stakeholder shifts, engagement drops, and signs of relationship strain, then surfaces them to our team as risk signals.
The MVPs of this project: Herald and Harbinger.
Several candidates who made our final invite list were not on our team’s original nomination sheet. Herald surfaced them for our team to consider.
These customers were quiet advocates. They passed every filter, and their advocacy readiness, engagement focus, and sentiment were strong. They just weren’t visible enough to be considered at first.
First results: how the numbers validated our human-led AI approach.
Out of the 41 candidates we invited, 25 said yes: a 61% RSVP conversion rate.
If you’ve ever made cold asks of senior executives, you’ll know that’s a strong number for a new program. Our final CAB cohort:
- 100% VP-level and above
- 7 industry clusters represented
- 4 countries across North America and Europe
- 4.4 years average tenure with ChurnZero
- 96% daily platform usage
- Distributed across three scoring tiers, giving us a meaningful spread for measuring contribution quality
What else are we measuring?
1: RSVP signal quality. We tracked our human-identified and AI-suggested invitees separately throughout the invitation process. Once the cohort is fully confirmed, we’ll compare their conversion rates directly.
2: Session contribution quality. After our first session, we’ll analyze the transcript for how our invitees contributed new ideas, reinforced themes, challenged assumptions, or asked sharp questions.
3: Community engagement. We’ve also built an exclusive member portal to keep CAB conversations going between events. The participation rate will indicate whether the cohort is genuinely invested. It also gives us a living data source for measuring cohort health heading into our annual ZERO-IN 2026 summit this fall.
What else can you take from this as a CS leader?
In this very focused task, Herald, Pulse, Vibes, and Harbinger proved a valuable extension (not replacement) of CSM judgment in an aspect of customer success that has traditionally been driven by slow-moving data and gut feel.
My question to you: where else are these blind spots showing up in your team’s work and results?
Consider which accounts are prioritized for expansion conversations, which customers are invited to reference calls, and which relationships are considered healthy.
If your judgment is based on scores alone, without considering sentiment and behavior, it’s worth exploring how to surface the hidden signals that made all the difference here.
Our cohort of AI agents—purpose-built for CS teams—is growing every day.
Peter Adams is the customer marketing manager at ChurnZero.
