How Iterable VP of Product Marketing Gray Hardell Built an AI Sales Coach That Increased Rep Confidence by over 65%
VP Product Marketing & GTM Strategy at Iterable
Templates Included
Summary
Most PMMs want to support every important deal. The problem is that the more useful PMM becomes, the easier it is to become the bottleneck.
In this playbook, youβll learn how Iterableβs VP of Product Marketing, Gray Hardell, built an internal AI sales coach to scale product marketing judgment across the field. The goal was not to create another content library. It was to give reps deal-ready access to the messaging, positioning, product context, proof points, and customer nuance they usually needed Gray to provide live.
But the biggest lesson was not about AI. It was about codifying PMM judgment.
Youβll learn how Gray mined repeated sales questions for product requirements, turned product and business context into modular context packs, tested the tool against real deal scenarios, drove adoption through selective access, and connected usage back to sales outcomes.
The early signal: when the tool was used, Iterable saw 42% faster sales cycles, a six-point higher win rate, and more than 65% of reps report higher confidence across AI POV, roadmap, and messaging.
If youβre trying to scale PMM support without joining every sales call, this playbook gives you a practical blueprint for turning your field knowledge into an AI coach reps can actually trust and use.
The triggerThe project started because I was spending too much time on individual deals.
That was a good problem. Reps trusted me for product context, proof points, messaging, roadmap nuance, and help applying those details to specific customer situations.
But the more useful I became to Sales, the more I became a bottleneck. The business needed my focus on broader GTM initiatives, not just one deal at a time.
I needed to build a way for the field to get that product marketing context without waiting on me.
Key internal stakeholdersThe project touched more teams than I expected. I needed input, validation, adoption, or operational support from:
Competitive Intelligence
Product Management
RevOps
Marketing Ops
Customer Success
Sales / Revenue leadership
Field reps across Sales and CS
Executive sponsors, including my CMO and CEO
IT / internal systems support for API access and tooling
That cross-functional spread mattered because the agent had to work for the real field motion. If it only worked for one type of rep, one segment, or one use case, it would not scale.
GoalThe goal was to increase sales efficiency. I wanted reps to get faster answers, move faster within deal stages, use better assets, and apply consistent messaging and positioning without waiting for me to join a call or build something custom.
I also wanted the work to be measurable. "Reps like it" was not enough. I needed to know whether the agent was helping with sales cycle time, win rate, confidence, closed-won and closed-lost patterns, and ACV.
Framework
To scale product marketing context across the field, I used a five-part framework for turning repeated sales questions into an AI field coach reps could trust. The sequence is:
Diagnose the bottleneck. Mine the questions reps already ask to identify the repeatable, message-sensitive problems slowing deals down.
Codify the context. Turn product knowledge, business context, sales process nuance, proof points, and PMM judgment into modular context packs the agent can use.
Test in real selling situations. Validate the agent against actual field prompts, real deal scenarios, and trusted users before opening access broadly.
Drive adoption through trust and exclusivity. Roll it out selectively, use influential reps as champions, and make trust visible through sources, approved examples, and timestamps.
-
Instrument usage and connect it to revenue outcomes. Instrument the workflow so adoption can be connected to sales cycle time, win rate, confidence, ACV, and deal outcomes β without asking reps to do extra logging.
.png)
Case Study
The sections below walk through how I moved from being a deal-by-deal PMM support layer to a field-facing AI system that could answer with our product context, proof points, and deal-ready information β starting with the questions reps were already asking and ending with usage tied back to revenue.

1. Diagnose the bottleneck by mining the questions reps already ask
The first phase was treating sales questions as data. The problem had been annoying for a while, but then it became overwhelming. I was getting pulled into too many deal-specific moments: "Does anyone have a deck for X?" "How do we position against competitors?" "What proof point should I use for this industry?" "Can Gray join this call?"
Over time, I realized those questions were effectively the product spec. The repeated asks revealed which product marketing knowledge needed to become scalable, self-service, and deal-ready.
Substep 1.1 β Treat repeated rep questions as data, not noise
I looked at the channels where reps were already asking for help: Slack DMs, ask-product channels, deal channels, and recurring sales requests over email. I looked for the questions that kept coming back in slightly different forms. Those repeated questions became the early spec for what I needed to build.
π₯ Hot Tip: The Help-Desk Burden Is the Spec
When the same field question shows up over and over, I treat it as evidence instead of a nuisance. My help-desk burden is often the clearest signal of what needs to be codified.Β
Sign up free to read the rest of this playbook.
Join Sharebird to access proven product-marketing playbooks from teams at the fastest-growing companies.