How We Built an AI Sales Coach on Bubble

The Problem: 67% of Salespeople Won't Hit Their Targets

Andy, the founder of SizzleKick, has been in B2B SaaS sales since 2003. He's built and exited companies. He knows sales.

And he kept running into the same problem, over and over again.

Every sales team has one or two people who crush their targets. And a whole group of others who don't quite get there. The frustrating part isn't that the gap exists. It's how to close the gap.

Salesforce is a great CRM. But pulling a clear picture of where every individual rep is at, their conversion rates, their call behaviour, their email quality, their activity patterns, on one screen? Almost impossible. And heads of sales don't have time to build 12 custom reports before their next forecast call.

The numbers back this up. In 2024, Salesforce predicted that 67% of salespeople would not hit their revenue targets that year. That's two thirds of a company's sales team falling short.

The question was: could an AI agent fix it?

The Insight That Made It Work

Andy didn't just wake up and think "AI + sales coaching = good idea." The insight came from his previous company, We Thrive, a platform built on the Human Givens psychological framework, helping organisations understand what their people need to perform at their best.

The key learning: coaching only works when it's personalised, timely, and reinforced.

Tell someone in a training session to ask more open questions in discovery calls. By week two, they've forgotten. They're back to old habits. Great coaching catches that, explains why it matters, and keeps returning to it until the new behaviour sticks.

The reason an AI sales coaching agent is so powerful here is variance. Every salesperson has different challenges, different metrics, different gaps to close. One-size-fits-all doesn't work. An AI agent can handle that degree of personalisation at scale, autonomously, in a way that standard software simply could not.

What SizzleKick's AI Sales Coaching Agent Actually Does

SizzleKick is an AI sales coaching agent that lives natively inside Salesforce. Here's what the agent does:

Identifies your most effective seller: The agent analyses all their metrics: conversion rates at every stage, average order value, call timing, email quality, talk-to-listen ratios. Then it autonomously benchmarks every other rep against that picture.

Builds personalised AI-driven coaching plans: Not "everyone copy Bob." The agent creates a specific plan for each rep based on where they are, what they need, and which improvement will have the biggest impact on revenue.

Analyses call transcripts autonomously: Via integrations with platforms like Salesloft, SizzleKick's AI agent pulls in what reps are actually saying in discovery calls and identifies patterns: are they pitching features instead of benefits? Skipping open questions?

Drafts personalised outreach at scale: Ask the agent which three leads to chase today. It picks the most likely to close based on your data, then writes a cold intro email aligned with your sales playbook, case studies, and competitive positioning. No blank page. No guesswork. The agent handles it.

Runs natively inside Salesforce: Not just connected to it. The AI coaching agent replicates the Salesforce UI environment and looks like it belongs there, because it does.

And it has a beautiful interface. Because as Andy put it: "It's hard enough to get salespeople to fill in Salesforce without getting them to use another product. If it was ugly, they'd look at it and go 'yuck, I'm not touching that.'"

It had to look right, work right, and earn its place in a salesperson's daily routine. See how we got there in our SizzleKick case study.

Why Bubble Was the Right Choice for Building an AI Sales Agent

When Andy first heard what Goodspeed could build on Bubble, how fast, and for how much, he didn't believe it.

"It fried my brain, to be frank. I couldn't reconcile that promise with my experience of building software."

He'd run companies before with sizeable tech teams. He knew the traditional process: whiteboard to requirements doc to spec to development. Long. Painful. Expensive. And the output still wasn't always right.

When Goodspeed delivered SizzleKick, a full AI sales coaching agent with a native Salesforce integration listed on the App Exchange, Andy couldn't quite believe it.

Here's why Bubble made building this AI agent possible:

Speed of iteration: When Andy came back after early client pitches and said he needed document upload functionality so clients could load their sales playbooks into the agent, it was live in the product within a week. In the traditional world, that would have meant a requirements gathering session and another significant development cost.

No technical debt: With traditional code, when something better comes along, you often can't use it. You've got a codebase to refactor. With the modular low-code architecture Goodspeed uses (Bubble, Flowwise, n8n) you can slot a new AI capability in, try a new agent behaviour, and pivot without burning the development budget.

Composability as competitive advantage: As Andy explained: "I think it means we can benefit from the latest, most effective tools that come out, whereas people without this architecture can't. And that gives us a competitive advantage and agility as a smaller player."

App Exchange listing: Getting listed on the Salesforce App Exchange is not straightforward. There are hoops. A lot of them. But the prize is access to 150,000 Salesforce customers. Andy's take: "It gives you credibility and removes a big barrier to sale."

The fact that an AI sales coaching agent built on Bubble could pass Salesforce's App Exchange approval process is a significant signal about how far the platform has come.

The Goodspeed Process: Go Slow to Go Fast

One of the biggest risks in building any AI agent product right now is the temptation to move immediately. Everyone sees AI generate full prototypes in minutes and thinks the same logic applies to building real products.

It doesn't.

At Goodspeed, we take care of Discover, Design, Develop, and Deliver. You take care of the business. Here is how that played out with SizzleKick.

1. Discover: We start by listening. Before a single screen is designed, we run a two-week Discovery Workshop to get deep into the business: the users, the workflows, the goals, and what is slowing things down. We do not start with features. We start with the problem. For SizzleKick, that meant understanding exactly what a head of sales needs to see, in what order, and what "great" looks like for the individual rep working with the AI agent every day. Andy described it well: the workshops felt like a lot of questions at the time, but they forced the team to get razor sharp on what they were actually building. You leave with clarity, not a guessing game.

2. Design: In a world where anyone can build, design is what makes people stay. Before development starts, every screen is mocked out. Every state. Every response. This step exists so that when Andy looked at a screen and said "that's not quite right," the fix happened in Figma, not mid-build. The result: when the AI sales coaching agent was handed over after QA, it looked how he expected it to look and did what he expected it to do. No surprises. And critically for SizzleKick, the UI needed to earn the trust of salespeople who already resented filling in Salesforce. Ugly was not an option.

3. Develop: This is where the AI sales agent comes to life. Built using Bubble, Flowwise, n8n, and OpenAI, with just enough custom code to keep things powerful, stable, and scalable. Because of the work done upfront in Discover and Design, development moves fast. And when Andy came back after early client pitches with a new requirement (document upload so reps could load their sales playbooks into the agent), it was live within a week. That is the speed that no-code and a solid AI architecture makes possible.

4. Deliver: Launch is not the end. Goodspeed stays involved as an extension of the product team, supporting real usage, improving agent performance, and helping the AI product evolve as the business grows. SizzleKick getting listed on the Salesforce App Exchange was not in the original brief. It came through that ongoing partnership.