Vibe Coding for Sales

Why "Vibe Coding" Is Reshaping How Sales Leaders Operate

Three months ago, Matt Stinson didn't know what Salesforce Apex code was. Today, he's built a fully functional exec-threading bot that saves him four to five hours a week and has measurably increased the volume of executive outreach his team runs. His story isn't unusual - it's becoming the norm for a new generation of sales leaders who are discovering that the gap between "I have an idea" and "I have a working tool" has collapsed almost entirely.

That was the central theme of Unify's latest Grow & Tell session, where Neema Zarrabian, Senior Sales Manager at Unify, sat down with Jeff Kirchick, VP of Sales at Zorro, and Matt Stinson, CRO at Starbridge, to unpack what happens when sales leaders stop buying point solutions and start building their own. The conversation quickly moved past the theoretical and into the tactical - live demos, real failures, and a surprisingly honest debate about where automation helps and where it quietly makes reps worse.

Building Your Own Conversational Intelligence Stack

Jeff Kirchick made a case that raised eyebrows: he argued it's now "fiscally irresponsible" for sales leaders to keep investing in enterprise conversational intelligence platforms when they can build comparable - and in many ways superior - workflows themselves. Using Cassidy AI as his orchestration layer, Kirchick walked through four use cases he's deployed at Zorro.

The first is an automated post-call coach, trained on Kirchick's own coaching style by ingesting his blog posts, past standup notes, and real call feedback. After every rep call, the system sends coaching feedback that mirrors how Kirchick would review the conversation - surfacing missed opportunities like failing to ask a prospect why they raised a particular question. The second is a CRM updater with a human-in-the-loop design: rather than auto-populating HubSpot fields, it suggests what it heard on the call and waits for the rep to confirm before pushing data. This preserves what Kirchick called "valuable friction" - the rep still has to know whether the information is accurate.

Perhaps most compelling was his custom deal-scoring framework, a hybrid of SPICED and MEDDIC he calls "SPID." The model analyzes every call and email exchange with a given prospect, then calculates a zero-to-five score across each letter of the framework. The score flows into a custom HubSpot field, giving Kirchick a data-driven view of deal health across the entire pipeline - and, critically, visibility into where the team is consistently falling short in their sales process.

The Exec-Threading Bot and the Power of Proprietary Workflows

Matt Stinson's approach centered on a different pain point: getting executive outreach to actually happen. His team runs a standard play where, once a deal hits a certain stage, an AE drafts an email for Stinson or the CEO to send to a peer-level contact at the prospect account. The problem was that these draft emails lived across fifteen different Slack threads and Google Docs, and Stinson had become the bottleneck.

His solution was a Slack-native exec-threading bot built in a single Claude Code session. When a deal moves into the forecast in Salesforce, the bot automatically prompts the AE to fill out the details - prospect name, email, context about the deal. It drafts the email in Stinson's voice, sends it to him for approval, and when he hits send, it fires through Gmail and tags the opportunity in Salesforce so the play isn't duplicated. The entire pipeline - Salesforce webhook, Railway hosting, Slack bot interface, Gmail integration - was architected and built by someone who, by his own admission, couldn't run a customized Salesforce report three months earlier.

The key insight both leaders shared: the filter for what's worth building isn't "can I build it?" - it's "does this add leverage over what I can buy off the shelf?" Stinson's bot works because no vendor sells exec-threading automation tuned to his team's exact workflow. Kirchick's coaching bot works because no platform coaches the way he coaches. When the answer is proprietary leverage, build it. When it's not, buy it - because homegrown tools break when you change your Salesforce password, and you're the only one who can fix them.

The Friction Paradox: When Automation Makes Reps Worse

The most candid part of the session was a debate about a problem none of the speakers had fully solved: automation can quietly erode the very skills that make reps effective. Stinson described a pattern where reps show up to internal deal reviews with beautifully written notes - clearly generated by Claude - but can't answer the second question about their deal. Zarrabian shared a similar observation: after automating SPICED note generation, his reps stopped internalizing deal context because the friction of writing notes had been removed.

Kirchick offered a framework for navigating this tension. Rather than automating the process itself (telling reps what to do next), he focuses on measuring outcomes through his SPID scoring system. A rep with a 13 out of 25 on deal qualification is forced to defend what they know and articulate how they'll close the gaps - and if they can't, the coaching conversation surfaces it immediately. Zarrabian added that he's deliberately walked back some of his own automation after realizing he'd gone too far, reinstating manual coaching touchpoints like reviewing one call per rep per week and having live conversations about what he observed.

The consensus: the search for information is a problem AI should solve, but the internalization of that information still requires human friction. As Stinson put it, giving a kid a calculator means they never learn arithmetic.

Getting Started: From Zero to Your First Build

For attendees who hadn't yet started building, the speakers offered practical on-ramps. Kirchick recommended starting with a simple research agent - something that connects to your calendar and sends you a briefing email about each prospect an hour before every call. Stinson suggested an even lower-lift starting point: download some rep conversion data, open Claude, and ask it to surface patterns you're not already thinking about. That micro-iteration, he said, is where the first real skill ideas emerge.

Zarrabian shared that his first build was deliberately unserious - a call coach with a Boston accent that made too many sports analogies, built on a Friday night. But that throwaway project taught him the mechanics of building in Claude Code, and within weeks he'd replaced it with a production coaching tool that follows his actual management rules. The lesson: start with something small enough that failure doesn't matter, and let curiosity pull you forward from there.

About the Speakers

Neema Zarrabian - Senior Sales Manager at Unify. One of Unify's first AEs, Neema has scaled from individual contributor to leading the sales team, and has become one of the most prolific internal builders using Claude Code and Cowork to automate sales management workflows.

Jeff Kirchick - VP of Sales at Zorro. A 16-year sales leadership veteran and author of Authentic Selling, Jeff predicted AI's impact on sales six years ago and has since built a full conversational intelligence stack using Cassidy AI, replacing traditional enterprise tools with custom-built coaching and deal-scoring workflows.

Matt Stinson - CRO at Starbridge. With 15 years in SaaS sales across IC and leadership roles, Matt went from zero coding experience to building production Slack bots and Salesforce integrations in under three months, and is a vocal advocate for sales leaders learning to build their own tools.