How We're Scaling Sales Coaching with AI and Sigma
Table of Contents

Sales coaching has always been tough to get right. Most solutions engineers (SEs) and account executives (AEs) only get a couple of hours of feedback per month, and even then, the feedback isn’t always consistent. We’re all moving fast, and it can be difficult to factor in time for detailed feedback and continuous improvement.
I’d been trying to fix that issue for my team for some time. I wanted a way to make feedback consistent, fast, and tied to how we actually sell. So I started building a custom data app, in Sigma, to track discovery questions across Stage 2 calls. It wasn’t perfect, but it was a start.
Then one of our SEs, Khush, showed me a sales coaching data app he’d built for his own team: SE Buddy. That moment completely changed how I thought about coaching at Sigma.
The sales coaching problem no one solved… until now
Let me rewind a little bit to a time before SE Buddy. I’d had this aspirational goal for a while to review sales calls for each solutions engineer on my team at a pace of one per week, per SE. As the team grew (there are now eight solutions engineers on my team), it quickly became clear that my earlier goal was unrealistic, so I adjusted it to review one call every two weeks. I failed at that too.
More recently, I was trying to figure out with our sales leaders how to provide coaching in early stage sales calls with my counterpart in a scalable way, since we keep growing our teams. Specifically, I wanted to figure out how we could help our teams move from Stage 2 to Stage 3 in the deal cycle, faster.
This is real, usable coaching, delivered at scale. We designed it around how we sell, not around some generic framework.
There were five questions I thought we should be asking during those calls to surface the specific pains that Sigma uniquely solves for, but I wasn’t sure they were actually being used. So, I listed them out and built a tool where each row represented a Stage 2 call, and each column was one of the five questions. It was a simple “yes or no,” “red or green” system. If you covered more than two questions, it was green. Two was yellow. Zero was red.

At first, I tested it with an account executive who hadn’t asked any of the five questions, but had run a great call. He’d built strong rapport, identified the decision-maker, walked through timelines, scoped the proof process, and even ended the call chatting about MMA with the prospect, whose son practiced jiu jitsu. Yet, by my framework, he scored red. I realized my five discovery questions weren’t enough because a call could skip them all and still be great.
Then Khush showed me SE Buddy, and I realized—I was working with Duplo Lego blocks. He was building the Millennium Falcon.
Building AE Buddy: Same engine, new formula
Khush had structured SE Buddy so that the AI could evaluate real calls against custom knowledge bases like discovery best practices, demo structure, Sigma differentiators, and competitive positioning. It showed you what went well, what didn’t, and how to improve. I immediately thought, “We need to build something similar for Account Executives.”
I took the framework and adapted it, starting with what I called the Stage 2 Accelerator. This stage is all about business discovery—surfacing pain, identifying ROI, and getting aligned on the value of solving that problem. So, I started building out knowledge bases that captured the real elements of a strong Stage 2 conversation like pain identification, champion building, process management, and stakeholder alignment.
It wasn’t a one-off. It was a new muscle for the business.
I also added a knowledge base for multi-threading, which asked questions like, “Are we engaging more than one person in the deal? Are we reaching decision-makers, influencers, blockers? Are we building broad support, or just talking to one friendly champion?” That kind of context can make or break a deal, and now the AI could coach on it directly.
Then, I created an AI data app specifically for Champion Building to coach reps on how to connect personally, understand the "what's in it for me," and create real internal momentum.
The sales call scoring system built in Sigma
Sales score equations
To give you an idea of the automated scoring system, here's an example of the components that go into the Champion Building score:
- Personal connection (25% weight)
- Active listening (20% weight)
- Champion development (30% weight)
- Trust building (25% weight)
The data app calculates a score based on these dimensions and weights. For champion building, the score looks like this:
Final Score = (Personal connection × 0.25) + (Active listening × 0.20) + (Champion development × 0.30) + (Trust building × 0.25)
Score component adjustments
Additionally, each component that goes into the score has its own specific guidelines. For example, these are the adjustments that factor into the active listening score, with a baseline score of 5/10:
- +1.0: Uses reflective statements
- +1.0: Asks relevant follow-up questions
- +1.0: References specific client statements later in call
- -1.0: Interrupts client
- -1.0: Misses obvious cues/pain points
- -2.0: Talks over client repeatedly
I spent about 12 hours building, improving the accuracy and repeatability of the output from the LLM. That was all. Then I retested it with that call where the rep talked about MMA with the client to see if it would capture all of its goodness, and the app actually scored it really well. That’s when I knew we had something unique.
The smartest coach you’ll ever meet
What we built with the Stage 2 Accelerator isn’t another dashboard. This is real, usable coaching, delivered at scale. We designed it around how we sell, not around some generic framework. Every knowledge base is tied to Sigma’s methodology, whether that’s discovery, champion building, objection handling, multi-threading, or competitive positioning. The system coaches against the exact motions that matter in our sales process. No generic LLM model knows that level of detail for the way we work.
To make the process more user-friendly, I pulled in help from our analytics team to build an interface that reps can actually use. Now, every Stage 2 call gets 10 scores. Nine are component areas—like discovery depth or stakeholder alignment—that roll up into a weighted executive score. Instead of one vague rating, sales reps can see precisely where they’re strong and where they need to focus on improvement.

And the real magic is how it compounds. We’re analyzing 18 months of calls to see which scores actually correlate to win rates. Every call teaches the system what really drives success.

We’re already on version 6 of the Stage 2 Accelerator—and version 7 is coming out soon. Each iteration bakes in feedback from the sales team. For example, AEs wanted to see scoring not just at the call level but at the opportunity level, across multiple conversations. So now it can roll everything up by deal, showing what’s been covered and what still needs work. It’s literally removing every traditional barrier to great coaching.

The spark that started 10 fires
Once the Stage 2 Accelerator went live, the requests started rolling in. Customer Success wanted a Customer Health Buddy to analyze sentiment across onboarding calls, customer QBRs, and escalations so customer success managers could get ahead of risks before they turn into churn. Marketing asked for a Campaign Intelligence Buddy to break down campaign calls, or track how messaging landed. RevOps pitched a Deal Inspection Buddy to audit deal hygiene, stage consistency, and pipeline risk automatically.
Pretty quickly, everyone saw the same pattern: if we could do this for SEs and AEs, we could do it everywhere. And when they saw we were building these in hours—not weeks—they realized it wasn’t a one-off. It was a new muscle for the business. That’s the real transformation here, where manual coaching becomes scalable, tribal knowledge becomes institutional, and every new "coaching Buddy" makes the next one faster to build. What took me twelve hours for the Stage 2 Accelerator could take one hour in the future.
The spark from SE Buddy lit a fire. Now, every team is asking the same question: what Buddy should we build next?
The only question left: What’s stopping you?
SE Buddy and AE Buddy didn’t start as AI projects. They started with sales teams who understood the challenges of coaching and wanted a better way to improve their outcomes. Sigma made that possible by giving them the flexibility to turn individual feedback into a consistent, data-driven system the whole team could use and refine.
With these tools, coaching isn’t an isolated event anymore. It’s a system that runs on real data, built by the people closest to the work, and refined through every conversation. Once we proved it in sales, every team saw how they could do the same. Sigma made that repeatable, and one idea became a model for how the company learns.
If you want to see what this looks like in practice, join our webinar on Building AI Systems with Sigma. You’ll see how teams are using Sigma to operationalize what they know—turning expertise into systems that scale.