What it's not
It's not CPQ with a chatbot. Slapping a conversational UI on top of your existing quoting tool doesn't make it agentic. If a human still has to open the app, browse a product catalog, select line items, adjust pricing, and generate a PDF, you've just added a step.
It's not AI-assisted CPQ. “AI-assisted” means the human does the work and the software suggests things. Autocomplete for quotes. That's useful, but it's not a new category. It's a feature.
It's not a copilot. Copilots watch you work and offer help. Agentic systems do the work and ask you to approve it. The difference matters.
What it is
Agentic CPQ is a system where AI is the primary operator of the configure-price-quote workflow. Humans review, approve, and send. The AI does everything else.
Here's what that looks like in practice:
1. The AI has context.
It reads the sales call transcript, the CRM record, the pricing rules, the contract history. It doesn't ask the rep to re-enter information that already exists somewhere in the revenue stack.
2. The AI has tools.
Not suggestions. Tools. It can read from and write to your CRM, pull product and pricing data, apply discount rules, and assemble a real quote. This is where MCP (Model Context Protocol) changes everything. When your CPQ exposes tools that any AI can call, the quote becomes a natural output of the sales conversation, not a separate workflow.
3. The AI produces a real deliverable.
The output isn't a summary or a recommendation. It's a quote. With line items, pricing, terms, and a structure your buyer can actually review. The quote is proof of work. If the system can't produce one from a real conversation, it's not agentic.
4. The human reviews, not rebuilds.
The rep opens a quote that's 90% done. They adjust, approve, send. The time from “verbal yes” to “quote in inbox” drops from days to minutes.
Why now
Three things changed.
Tool-use protocols matured.
MCP gives AI models a standard way to call external tools. Your CPQ, your CRM, your conversation intelligence platform can all expose capabilities that AI can compose into workflows. This didn't exist two years ago.
Every revenue tool is becoming AI-readable.
Gong, HubSpot, Salesforce, Slack. They're all building interfaces that AI agents can interact with. The data and context required for quoting are finally accessible programmatically in real time.
The last mile is still manual.
Revenue teams invested millions in tools that capture conversations, enrich contacts, forecast deals, and automate sequences. But the actual quote? Someone still opens a spreadsheet or a legacy CPQ and builds it by hand. That's the gap.
The Veles position
This is what we're building. A CPQ platform designed from day one for AI agents to operate. Not retrofitted. Not “AI-enabled.” Native.
Our MCP server exposes 22 tools. Any AI agent, whether it lives in Gong, Claude, Slack, or a custom workflow, can configure products, apply pricing rules, and generate quotes through Veles without a human touching the UI.
The quote is the proof of work. We think it's the most important artifact in the deal, and it should be generated automatically from the context already in your revenue stack.
That's Agentic CPQ. We're building the defining platform in this category. That's the bet.
Simon Ooley is the CEO and Co-Founder of Veles (YC W24), building Agentic CPQ. getveles.com