Julien Bek at Sequoia wrote that the next trillion-dollar company will be a software company masquerading as a services firm. For every dollar spent on software, six are spent on services. The real money is in capturing the labor budget, not the software budget.
That article has been living rent-free in my head because nowhere is its point more obvious than in CPQ.
The CPQ software market is roughly $3–6B depending on who you ask. But the labor market around quoting is enormous. And that labor market is exactly where the opportunity sits.
The CPQ Tax: $10 in Labor for Every $1 in Software
Every B2B company with complex pricing has some version of a deal desk. At a typical public SaaS company with 1,000+ reps, that means 10–15 deal desk analysts plus a 20–30 person Salesforce and GTM systems team keeping the machine running. Forty people whose jobs exist because CPQ software doesn't actually do the work.
Multiply that across every mid-market and enterprise SaaS company and the numbers get serious. Deal desk analysts. RevOps managers. Sales ops engineers. Salesforce admins maintaining CPQ configurations. External consultants charging $200–400/hr for 6–12 month implementations. Entire practice areas at Deloitte and Accenture that exist solely to implement and maintain Salesforce CPQ and Revenue Cloud Advanced.
According to Salesforce's own data, sales reps spend only 28% of their time actually selling. The rest goes to data entry, configuration, approvals, and administrative work. Gartner estimates that the average B2B sales cycle lengthened by 25% between 2019 and 2024, with quoting bottlenecks as a leading contributor.
For every dollar spent on CPQ software, companies spend ten or more on the people required to make it functional. That ratio is the clearest signal of where AI should be applied.
Why Traditional CPQ Failed
Traditional CPQ is a tool. It gives you a rules engine, a product catalog, and a quoting interface. Then it says: good luck.
You still need a team to configure and maintain the rules and pricing logic as your packaging evolves. Sales enablement still needs to train every rep on how to use the system. UAT is still required in your sandbox to validate that everything works when a change is implemented. And when the rules get complex enough, which happens quickly in enterprise SaaS, reps bypass the CPQ entirely and go back to spreadsheets.
The software doesn't build the quote. It builds the environment in which a human can build the quote.
That's why implementations take 6–12 months. That's why the average CPQ deployment requires specialized consultants. That's why most mid-market companies never adopt CPQ at all. A 2023 Gartner survey found that fewer than 30% of mid-market B2B companies have deployed a CPQ solution. The rest rely on spreadsheets, slide decks, and email.
The tool is too expensive and too complex for the outcome it delivers. And the outcome it delivers is, at best, a slightly faster way for a human to still do all the work.
Intelligence vs. Judgment in Quoting
Bek draws a useful line between intelligence and judgment. Intelligence is rule-following, even when the rules are complex. Judgment requires experience, taste, instinct. Both are required in quoting, but they are not equally distributed.
Intelligence is the bulk of the work. Roughly 80% of the quoting process is configuring products, applying pricing rules, checking discount guardrails, routing approvals, and validating compliance. It is all rules and math. There is a correct answer. The system either applies the right pricing or it doesn't.
Judgment is the human layer. The remaining 20% is where reps and deal desk leaders earn their keep. Offering a non-standard discount to land a strategic logo. Structuring a creative deal around a tight buyer budget. Knowing when to push back on a discount request versus when to escalate to leadership. Reading the room on timing, leverage, and competitive dynamics.
The problem with traditional CPQ is that it forces humans to do both. The rep who should be exercising judgment on deal strategy is instead spending their time on the intelligence work: looking up product codes, checking pricing tiers, submitting approval requests, and formatting documents.
Agentic CPQ flips that ratio. The AI handles the 80% that is intelligence. The human focuses on the 20% that is judgment. The quote is proof that the intelligence work is done.
Don't Sell Better Quoting Software. Build the Quote.
Every CPQ vendor is racing to add AI features. Copilots that suggest pricing. Assistants that help configure products. Chat interfaces bolted onto the same old rules engine.
This is the wrong move. They're selling better hammers.
The right move is to build the house.
A company doesn't want CPQ software. It wants a signed contract. It wants the right products configured, the right pricing applied, the right terms included, the right approvals routed, and the right document generated. The quote is the deliverable. Everything else is overhead.
This distinction has a compounding effect on competitive dynamics:
If you sell the tool, every improvement in the underlying AI models makes your product more vulnerable. Someone else will wrap Claude or GPT in a better UI and undercut you. Your moat is your interface, and interfaces are commodities.
If you sell the work, every improvement in the underlying models makes your service faster, cheaper, and more accurate. The quote lands better. Margins improve. Your moat is the accumulated context of every deal you've ever processed.
The Outsourcing Test
Bek says the right place to start is where outsourcing already exists. If a task is already outsourced, it means three things: the company accepts external delivery, there is an existing budget, and the quality bar is defined.
CPQ passes this test cleanly.
Companies already outsource most of the quoting stack. They hire consultants for the CPQ implementation. They contract Salesforce administration to managed service providers. They staff deal desk analysts who are, functionally, internal outsourcers of a process the software was supposed to handle. They pay RevOps agencies to build and maintain their pricing logic.
The entire infrastructure around quoting is already partially outsourced. The labor budget is already allocated. The only thing missing is a service that replaces all of it at once, not with more people but with an AI agent that does the work and delivers the output.
What This Looks Like in Practice
A rep gets off a discovery call. The conversation covered the prospect's requirements, budget, timeline, and decision criteria.
Today: The Tool Approach
The rep opens Salesforce. Opens the CPQ. Manually configures the products. Looks up the pricing for this customer's segment. Checks the discount policy document. Submits for approval. Waits. Gets feedback from the deal desk. Revises. Resubmits. Waits again. Eventually generates a PDF. Sends it.
This takes 2–5 days for a complex deal. Along the way, 3–4 people touch the quote. Version control becomes a problem. The pricing may have changed between the call and the final document. The buyer's momentum fades.
Tomorrow: The Work Approach
An AI agent reads the call transcript. It extracts the requirements, budget parameters, and timeline. It configures the products in the CPQ, applies the right pricing rules and discount guardrails, routes approvals to the right people (auto-approving within policy), and generates the quote document.
The rep opens a quote that is 90% complete. They review, make judgment-level adjustments, and send. Total time from verbal agreement to quote-in-inbox: minutes, not days.
The second version isn't a better CPQ tool. It's a quoting service. And the quote is the proof of work.
The Pricing Model Tells You Everything
Every legacy CPQ vendor charges per seat. Salesforce CPQ. DealHub. Conga. They are selling access to a tool. The pricing says: here's the environment, now go do the work yourself.
If you are selling the work, you don't charge per seat. You charge per outcome. Per quote generated. Per active contract managed. Per deal closed.
Seats are irrelevant when the AI does the work. What matters is how many quotes you generate, how many contracts you manage, how much revenue flows through the system. The unit of value is the deliverable, not the license.
This is Bek's difference between selling QuickBooks and closing the books. One is a tool for $30/month. The other is a service worth $2,000/month. The output is the same. The labor displacement is the value.
| Dimension | Tool Model (Legacy CPQ) | Work Model (Agentic CPQ) |
|---|---|---|
| Pricing | Per seat/user | Per active contract or quote |
| Value proposition | Access to a rules engine | Finished quotes, delivered |
| Implementation | 6–12 months | Days to weeks |
| Ongoing labor required | Deal desk + RevOps + admins | Human review and judgment only |
| Moat | UI and feature set | Accumulated deal context |
| Effect of better AI models | Increases competitive pressure | Increases margin and speed |
The Compounding Advantage
Every quote an AI-native CPQ builds teaches it something. What products get bundled together. What discount levels get approved versus rejected. How pricing structures convert at different deal sizes. Which terms cause legal friction. What approval paths are fastest.
This data compounds. After a thousand quotes, the system knows your pricing dynamics better than any individual rep. After ten thousand, it has pattern-matched across your entire deal history.
A traditional CPQ vendor can't access this loop because the human is doing the work. The software is just the container. It sees the final output but not the reasoning, the false starts, the revisions, or the judgment calls that shaped the deal.
An AI-native service captures every decision, every approval path, every outcome. The service gets better with every quote it builds. The accuracy improves. The approval routing gets smarter. The pricing recommendations sharpen. This is a compounding advantage that scales with usage and cannot be replicated by a tool vendor adding AI features retroactively.
The Real Market Size
The CPQ software market is $3–6B. But that's the wrong number to focus on.
The right number is total labor spend on quoting across B2B companies. Deal desk teams. RevOps headcount. Sales ops engineers. Implementation consultants at $200–400/hr. Managed service providers. External agencies. Contract workers maintaining pricing logic.
That number is north of $50B.
For every dollar spent on CPQ software, the labor multiplier is 10x or more. That's the real TAM. And it is available right now to the company that stops selling the tool and starts delivering the quote.
This is not a theoretical market. The labor is being spent today, right now, at every B2B company with complex pricing. The budget exists. The dissatisfaction exists. The only question is which company captures it first.
The Path Forward
The next great CPQ company won't help you build quotes faster. It will build them for you.
It won't charge you per seat. It will charge you per outcome.
It won't require a 6-month implementation. It will connect to your CRM, read your pricing rules, and start delivering quotes.
It won't need a team of consultants to maintain. It will get better automatically with every deal it processes.
The quote is the proof of work. Everything else is overhead.
Frequently Asked Questions
What does “the quote is the proof of work” mean?
It means that the value of a CPQ system should be measured by its output (finished, accurate quotes), not by the features of the software itself. If the system can take a sales conversation and produce a ready-to-send quote without manual work, it has proven its value. The quote is the deliverable that demonstrates the AI actually did the job.
How is this different from AI-assisted CPQ?
AI-assisted CPQ adds suggestions and autocomplete to a tool that humans still operate. The human does the work with help. In an agentic, work-first model, the AI builds the quote end-to-end and the human reviews it. The difference is who does the work versus who reviews it.
What is the labor multiplier in CPQ?
For every dollar spent on CPQ software licenses, companies spend roughly $10 or more on the people required to make the system functional: deal desk analysts, RevOps staff, Salesforce admins, implementation consultants, and managed service providers. This 10x labor multiplier represents the real market opportunity for AI-native CPQ.
Can AI handle complex enterprise deals?
AI handles the intelligence work: product configuration, pricing rules, discount validation, approval routing, and document generation. This represents roughly 80% of the quoting process. The remaining 20%, including strategic discounting decisions, creative deal structuring, and relationship judgment, stays with the human. The result is that reps spend their time on judgment, not data entry.
How does outcome-based pricing work for CPQ?
Instead of charging per user seat (which prices access to a tool), outcome-based CPQ pricing charges per active contract, per quote generated, or per deal managed. This aligns the vendor's revenue with the customer's results. If the system generates more quotes and manages more contracts, both parties benefit.
What is the total addressable market for AI-native CPQ?
The CPQ software market is estimated at $3–6B. But the total labor spend on quoting, including deal desk teams, RevOps, sales ops, implementation consulting, and managed services, exceeds $50B across B2B companies. AI-native CPQ that displaces labor, not just software, addresses the larger market.
Simon Ooley is the CEO and Co-Founder of Veles (YC W24), building the agentic CPQ platform. Previously, Simon held sales and sales leadership roles at Procore Technologies (PCOR), Buildr CRM, and SINAI Technologies.