Your CRO Wants to Run the Business in Chat. He's Not Wrong.

Two years ago, Robert Zimmerman, CRO at Qualified and a former Salesforce early employee, said something that made Kieran Snaith mentally roll his eyes.

"I'm going to run my business in chat someday," Zimmerman told him. What's my forecast? Where's my risk? Who do I need to call this week? All from a single interface.

Kieran's reaction, in his words: "At first I didn't believe in that vision. I knew it wouldn't understand our data model. It wouldn't handle the errors. That's where these systems often trip and fall."

Two years passed. Last week, Kieran sent Zimmerman a Slack: Go login to Von. You said this two years ago. You can now do it.
‍

Kieran Snaith is SVP of Revenue Operations at Qualified. He's evaluated over 100 AI vendors since GPT first dropped, and he joined us this week for a webinar on what it actually looks like to build an AI-forward revenue tech stack, what he'd do differently, and what he's most excited about going into the rest of 2026.

Start with the problem, not the technology

The mistake most RevOps teams make when they get the "go do something with AI" mandate from above: they start shopping before they know what they're buying for.

Kieran's rule is simple. Pick two to three critical revenue problems first, not a wish list or a technology category, but actual problems your business needs solved.

For his team, that was forecasting accuracy, process automation, and inside sales outbound. They picked those because they were critical business problems that were concrete, measurable, and owned by someone.

"Forecasting was an easy fit," he said. "Everybody wants to do it better. So great, AI could come in, have a massive impact, and get executive visibility immediately."

The second piece that trips teams up: time to value. Some things move fast. A simple PDF extraction, pulling data off an order form into Salesforce, can be done in hours. But if you're trying to replicate what a human BDR does, or automate a CSM's renewal motion? That's six to nine months before you have something you'd call finished. And then you're already on V2.

Understanding which category you're in before you sign makes a significant difference.

Assign ownership like you would any other hire

This one surprised people in the room, and it shouldn't.

When Qualified introduced an AI forecasting engine, Kieran didn't hand it to IT. He gave it to his VP of Sales Operations, the person already accountable for forecasting. Every Tuesday, they'd compare: who called it better, the AI or Nick?

"He felt ownership over it," Kieran said. "Because he owns forecasting. You want people to feel that."

He did the same thing with AI BDRs. He put one on the leaderboard alongside 20 human reps, measuring it on the same activity metrics and conversion rates as everyone else. It wasn't a separate experiment running in parallel. It was a direct comparison, which is how you know if it's actually working, and it's how you get humans to engage with it rather than quietly work around it. Getting the measurement right is only part of it though. We'll get to the compensation piece in a moment.

The QA question came up, and this is where most agentic deployments fail quietly. Who owns it?

The ownership is the same. The person who owns the outcome owns the QA. If the forecasting model hallucinates, that's the VP of Sales Ops's problem to catch. For outbound emails, Kieran's BDRs reviewed and approved every AI-generated email before it went out, a one-second read and a button click, because he couldn't trust a single engineer to QA thousands of emails, and he wasn't yet ready to trust the agent to review itself.

"You're not going to set it and forget it," he said. "The more you QA, the better your output will be."

The human ego problem is real. Here's how he handled it.

Someone in the audience asked the question everyone was thinking: what do you do when your people feel threatened?

Kieran gathered his 20 outbound reps and gave them a choice, and he didn't soften it.

"You can not get credit for what AI generates and just keep pounding the phones. Or you can get credit for what AI produces for you, be more productive, and I reserve the right to raise your quota 25% each quarter."

Every single hand went up for option two.

He also changed the compensation attribution models so there was no incentive to work around AI. Reps shouldn't have to think twice about whether to let AI help them because they're worried it might cost them credit. If you want adoption, eliminate the conflict.

Buy platforms. Stop buying point solutions.

Kieran was blunt here: the era of the point solution in AI is over.

"I'm looking at what are my systems of record and what are my systems of action. AI should compound across all of them, not create another data silo."

For his team, everything lives in Salesforce. So every AI vendor gets evaluated on how deeply it integrates there, and whether it's nimble enough to move when the business moves. Because it will move. Your Salesforce schema will change, your processes will evolve, and you don't want to rebuild your AI stack every time they do.

His buy vs. build framework is straightforward: look at time to value and total cost of ownership. Some things are obvious builds, simple, contained, done in a day with the right tools. Anything larger, anything you'd need to maintain, anything where you want someone else's engineers fixing the bugs at 2am, that's a buy.

"I love having another neck to choke," he said. "When Von gives an incorrect answer, I get to go to Sahil. Versus my CRO coming to me."

What RevOps actually looks like in 12 months

This is the part that should make every RevOps leader sit up.

Kieran's take: the ad hoc reporting function is gone. Not going. Gone. "Six months," he said. "If you're in a role where you're turning and burning reports all day and can't get to the bigger rocks, it's there."

What doesn't go away is the infrastructure layer. The people who make sure the agents are set up correctly, who coach the system on what a good answer looks like for a sales leader asking about pipeline risk versus a CRO asking about forecast confidence. That work becomes more important, not less.

RevOps becomes the team that makes the AI trustworthy. That's a different job, and honestly, a better one.

The vision Zimmerman described two years ago, running a revenue business from a single chat interface, on demand, without a queue, without waiting, isn't theoretical anymore.

We built Von to be additional headcount for your revenue team, the kind that takes on the tasks that pile up, the ad hoc requests, the Friday afternoon asks, the analysis that's been sitting in the backlog for a month, and delivers finished work without you having to manage it. Every member of your revenue team gets the analyst, the data scientist, and the ops resource they've been asking for.

We're currently offering a free trial for teams who want to see it before they commit. Minimal setup, no lengthy sales process. Log in and give it something from your queue.

‍

Start your free trial at vonlabs.ai

Sara Kinsey is VP of Marketing at Von. Von is an AI data scientist for revenue teams, built by the team at Rattle.