Revenue's Turn

Coding agents ship PRs. Support agents deflect tickets. Both are mature, both save you money. Revenue agents? There are dozens of companies calling themselves that. Enrichment tools, AI SDRs, AI CRMs, sequencing platforms. They each do one slice well. But nobody has stitched the full thing together. Nobody is running pipeline generation, inbound routing, deal management, and renewals as one continuous system. The revenue lifecycle is still a patchwork of point solutions with a human gluing them together.

That's what we've been building at Rox. Not another point solution. The full horizontal stack, on autopilot.

Rox Autopilot. Agents that run end-to-end workflows across the entire revenue lifecycle, from prospecting to renewals, without a human touching them.

But before I get into what we're launching, I want to talk about why nobody has done this yet. It's not a market gap. It's a genuinely hard technical problem. And the reason every other company in this space stopped at one slice is because solving the whole thing requires infrastructure that didn't exist.

When we started Rox in 2024, we saw the same thing everyone else saw. Data was moving out of business systems and into the lakehouse. What looked like a B2B SaaS trend turned out to be secular. Semiconductors, banks, energy companies, fintech, everyone was doing it. The interesting question wasn't whether this was happening. It was what you could build on top of it.

We wanted to build agents that could do real revenue work end to end. But there were three structural problems standing in the way, and none of them existed for coding or support.

No private context

A coding agent lives in the repo. A support agent mostly lives behind a help desk and a knowledge base. A revenue agent needs data from everywhere. CRM, conversation transcripts, product usage, billing, enrichment providers, renewal history, spread across a dozen systems that don't agree with each other. You can't just connect to them. You have to unify the data, impute missing fields, and cleanse the whole thing to make it AI-ready. We had to build a knowledge graph from scratch before a single agent could take a single action. Coding and support agents never had to solve this.

No standard UI

Coding agents built on top of editors and IDEs. The form factor existed, they just had to extend it. Support agents had the chat window. Sales had Salesforce, or one of thirty point tools bolted on top, and reps were spending most of their day alt-tabbing between them. Nobody wanted another tool. So we had to build a UI layer that felt like something reps, SDRs, RevOps, and leaders already knew. Not a new product. A familiar surface.

No feedback loop

This is the one that doesn't get enough attention. Coding agents got good because they could hill-climb. Run the tests, check if they pass, iterate. Support agents had deflection rates as clean eval signals. Revenue had nothing. No standardized evals, no automated way to know if the agent did something useful. And unlike coding where the agent can iterate until the tests pass, in revenue you often get one shot. If you put the wrong context into the window, you get garbage out, and that garbage goes to a prospect or a customer. We built these feedback loops painstakingly over the course of a year, collecting manual signal from every team that onboarded and running local maxima optimization until the system got sharp enough to trust.

We had to solve all three of these before we could build what we actually set out to build. Just the slow effort of getting the data layer right, getting the integrations right, getting the evals tight.

I'll be honest, we rewrote the UI multiple times as we onboarded more customers and more data surfaced. We rebuilt our agent orchestration layer three times over. Every time we thought the system was solid, a new customer's data would expose assumptions we didn't know we'd made. But each rewrite made the foundation stronger. And the foundation was the product. We just didn't know it yet.

So here's what Rox Autopilot actually does.

Outbound on autopilot

The agent identifies your ICP and persona, enriches contacts, and generates sequences. I want to be specific here. We spent a disproportionate amount of engineering time on sequence quality because the bar for "not AI slop" is high and almost nobody is clearing it. The agent handles auto-replies and books meetings on your calendar. One customer is generating 8 meetings per rep per week through Rox with zero manual intervention. Another is running $40M per week in pipeline through the system.

Inbound on autopilot. Same architecture, different trigger. Leads come in from any source. The agent qualifies, routes, sequences, and schedules.

Renewals on autopilot

This is where the knowledge graph really earns its keep. The agent pulls conversation history, value delivered, consumption data, and usage projections from the graph, generates a full renewal deck with an order form attached, and sends it out ahead of the renewal date. You're not going to use this for strategic enterprise accounts. But customers are running $800M of renewals through Rox right now, primarily through channel and mid-market motions. That number surprised even us.

Back-office automation

CRM updates, stage movements, the low-order tasks that eat hours every week. Handled in the background.

Here's the thing about revenue that makes it fundamentally different from every other agent domain. Humans buy from humans. Coding agents still need a human to review and merge the PR, but the goal is to minimize that touchpoint. Support agents are optimized to deflect entirely. Revenue agents are the opposite. You want a human in the loop, not as a bottleneck but as the closer. The rep handles the relationship. The agent handles everything else. And the "everything else" is massive. Generating a renewal deck, prospecting into a new account, writing a follow-up, building an org chart, enriching contacts. That's where the time goes, and that's where the agents live.

Every CIO and CRO conversation we've had in the last six months lands in the same place. They're not just looking for efficiency anymore. They want agents that impact the top line. Find net new business. Book real meetings. And they want consolidation, because the tool sprawl in revenue is out of control.

This space has been full of enrichment providers and AI CRMs that do one or two of these things. Consolidation or automation or pipeline generation, pick two. Nobody does all three.

We do. And because we do, we can price in a way nobody else can.

We give enrichment away at cost. No seat-based pricing. We charge for usage and outcomes. We can price at cost on the usage piece too, because we're not trying to monetize the data layer. We're vested in your outcomes. That's a fundamentally different alignment than paying per seat for software you might not use or per record for data that might not convert.

Coding and support agents were ahead of us. They had structural advantages we didn't have. Cleaner data, standardized UIs, built-in eval loops. We had to build all of that ourselves.

But revenue agents have caught up. And revenue is the last major domain left.

• $800M in renewals running through the system.

• $40M per week in pipeline.

• 8 meetings per rep per week, fully automated.

A year ago none of this was possible on Rox. Now it is.

If you tried us before and it wasn't ready, it's ready now.