Every distributor has the same untapped asset: the knowledge inside their best reps' heads. The rep who knows which complementary items go on every order, who reads a slammed operator's tone in the first ten seconds, who never gets rattled by a tough gatekeeper - but that expertise gets passed down one ride-along at a time, and most of it never makes it past the truck.

What if a new hire could practice against that knowledge whenever they wanted: running the same cold open, the same gatekeeper, the same tough close as many times as it took to get it right? That's the promise of a voice AI agent, and in our latest webinar, Nick and Michael built one live to show how approachable it has become. 

Here's what we covered:

Different Kinds of Agents

Before building anything, you decide what the agent is for. 

• Role-playing agents let your team practice a skill in a safe place. Things like cold opens, getting past a gatekeeper, and upselling complementary items. As Michael put it, the best reps know an operator ordering buns and ground beef probably needs tomato, lettuce, and ketchup too. A role-play agent runs that a hundred times without burning a real prospect. 
• Documentation agents are the other side: an ask-me-anything bot for product knowledge, playbooks, and process. The fastest lookup you'll ever run against your own information.

The System Prompt: Where the Persona Lives

The single biggest lever is the system prompt: the agent's identity and persona. Start it blank and it knows nothing; give it a persona, and it comes alive. Remember though, you don't have to write the perfect prompt. Simply Claude, ChatGPT, Gemini (or whatever AI tools you use)  and have a conversation. 

In the demo, the team asked Claude for a composite restaurant operator and got Dana Winfield of Ardent & Oak, a 140-seat Southern American spot, on her third broadline rep in two years, direct and impatient because she came up on the line, not in an office. That texture came out of a short back-and-forth, not a blank page.

And you almost certainly have better raw material than a composite. Pull a customer list from your ERP. Feed in your persona docs, PDFs, and order guides. Use call recordings if you've got them! A conversation with an AI gets you richer content than you'd type from scratch.

The Knowledge Base: Turning Ride-Alongs Into Assets

If the system prompt is who the agent is, the knowledge base is what it knows. The session built one from a single 32-minute interview with a DSR veteran, pulling out how to handle "I'm happy with my current rep," the exact scripts to use, and how hard to push on target accounts. 

Simply Record the interview on your phone, drop the transcript into your AI of choice, and a ride-along becomes something every new hire can talk to.

Analysis: Scoring and Data Collection

In the analysis tab, you write call-scoring criteria in plain English. Things like "mark as successful if the rep uncovered at least one pain point." 

In one test that we covered, the agent flagged a failed pitch: good discovery, but the rep jumped to price and never proposed a next step. That's coaching a manager used to deliver by hand, generated automatically. Data collection goes further: capture which questions reps ask, and you'll see exactly where your documentation has holes.

Where to Start

The honest answer to "how do I begin" is: begin. Build a composite persona in a few minutes, drop in one ride-along transcript, and poke at it — you'll be surprised how good it is before you've spent a dollar. Pepper runs a whole stable of agents, including Gail, a notoriously tough receptionist newer reps have to get past, and every one started exactly that way. The knowledge in your best reps' heads doesn't have to stay there.

‍