The short answer: To get started with AI Quoting, connect your CPQ (Configure, Price, Quote) tool to Claude or ChatGPT through MCP, then give the AI three kinds of context: how to use your CPQ's tools, your pricing and packaging rules, and data from your CRM. Start with one quote type, test it on real deals and fix every miss.
Key takeaways
- AI Quoting lets a rep ask for a quote in plain language and get back a draft quote in the CPQ, which a human reviews and sends.
- Connecting the tools is quick. Building the context is the real work, and it is what makes the AI quote the way your company sells.
- Before you build, check four things: leadership accepts a slower start, AI tools are approved, AI spend is approved and someone owns the build.
- Skills (saved, step-by-step instructions the AI follows) turn a different answer every time into the same answer every time.
- Start with one transaction type, such as renewals, and expand once it works.
Every RevOps leader I talk to wants AI Quoting. Very few know where to start.
So we asked someone who has already done it. Andrew Linford, VP of Operations at Mosaic, joined me for a live session on getting started with AI Quoting. His reps now build new quotes, renewals and amendments just by asking Claude. Here is the playbook he shared, in the order you would actually do it.
What is AI Quoting?
AI Quoting is a way of creating sales quotes where a rep describes the deal in plain language and an AI assistant builds the draft quote in the CPQ. A rep types something like "renew Globex at 40 users." The AI gathers the context (the CRM account, the current contract, product usage, call notes), applies your pricing and approval rules, shows the rep the proposed changes and writes a draft quote. Nothing goes to the customer until a human reviews it.
The connection between the AI and the CPQ usually runs through MCP (Model Context Protocol), an open standard that lets AI assistants like Claude and ChatGPT read from and act in other software. AI Quoting is also called conversational quoting, because the quote starts as a conversation instead of a form.
Why use AI for quoting instead of training reps on the CPQ?
AI Quoting removes the need to train every rep on the CPQ, because the knowledge lives in the AI's instructions instead of in each rep's head. Anyone who has run a CPQ knows the three mistakes that keep showing up on manual quotes:
- Inconsistent pricing and terms: different discounts, price increases or payment terms on similar deals.
- The wrong product mix: products combined that shouldn't be, or a required add-on left off.
- Missed details: an auto-renewal term or a promised price cap that never makes it onto the quote.
Training doesn't fix these, and I can tell you why from the sales seat. When RevOps puts CPQ training on my calendar, I'm either on a customer call or half listening. Then, months later, I finally need to build a complicated quote and I have no idea how. That's when the deal desk tickets start piling up.
Andrew put it better than I can:
"Sales reps are never getting hired for their detail orientation to quotes."
So stop asking them to be. With AI Quoting, a new rep can quote the right way on day one, and every rep quotes like your best rep.
Is your organization ready for AI Quoting?
An organization is ready for AI Quoting when four things are true. None of them is technical, and all of them are easier to settle before you build than after:
- Leadership accepts a slower start. At first, AI Quoting is slower than doing it by hand. It gets faster as you train it. If leaders expect a day-one win, the project will look like a failure in week two.
- AI tools are approved, including write access. Reading data through MCP is one thing. Letting the AI write draft quotes into your CPQ is another, and your IT team needs to be comfortable with it. At Mosaic, the AI writes drafts only, and only a rep ever sends a quote.
- AI spend is approved. Mosaic gives every rep a Claude Pro plan on a team subscription, which covers what reps use. If your company pays for AI by usage, expect finance to want a business case up front.
- Someone owns the build. One person has to build the first version and keep improving it after launch. At Mosaic, that's the operations team.
Andrew didn't hit any of these walls at Mosaic. He hears about them constantly from peers at larger companies, where one security review can cost weeks.
How does AI Quoting work behind the scenes?
AI Quoting works by chaining a rep's request through an AI assistant, a skill, the CPQ's tools and a human review. Here is the flow Andrew uses:
- A rep asks for a quote in Claude or ChatGPT, or pastes in a screenshot or email with the request.
- The AI picks the right skill for the job, such as a new quote, a renewal or an amendment.
- The skill follows its steps and reads the context: your products, rules and what needs approval.
- The AI calls the CPQ's tools through MCP, and pulls in other data if it's connected (CRM, call recordings, signed contracts).
- The AI shows the rep the before, the after and the changes, and asks for confirmation.
- The AI writes the draft quote in the CPQ. The rep reviews it, any approval rules run and the rep sends it.
Skills are what make this reliable. A Claude skill is a packaged set of instructions the AI loads when a task calls for it. You don't technically need one. But if you ask the AI the same quoting question in five different chats, you'll get five different answers. A skill tells it exactly what to do, in what order, every time.
Reps don't need to call skills by name, either. Andrew writes his skills so the AI recognizes when a request is about building, editing or renewing a quote and picks the right skill on its own.
What context does an AI Quoting agent need?
An AI Quoting agent needs three kinds of context: how to use the CPQ's tools, how your company prices and packages, and the outside data that shapes each deal. This is where most of the work goes. As Andrew said:
"You can hook up the tools fairly quickly, but having all the right context, that takes time and effort to build out."
- Tool instructions: which CPQ action to use in which situation. An MCP connection only carries a short description of each tool, so the AI needs more detail to stop guessing. Many vendors provide this in their API docs or a ready-made plugin.
- Your pricing and packaging rules: quote defaults, service hours, legal terms (and how you phrase them), per-user versus volume pricing, and rules like "this package always includes this onboarding."
- Outside data: the CRM fields to check, product usage, call recordings and the signed contract on file.
A renewal shows why the third kind matters. Say a customer has 50 users under contract but only 40 active. A good AI Quoting setup checks the CRM account, the usage data, the latest call recording (did they ask about an add-on?), the current contract in the CPQ, the signed agreement in your document storage and the account manager's input. Then it proposes 40 users and flags anything that doesn't line up.
The good news: you don't write this context by hand. Andrew says he hasn't written a line of his. The fastest start is to load your past contracts, quotes and invoices into the AI and have it summarize your pricing patterns. Existing quote training documents work too. That gets you roughly the first half of your quotes. The corner cases come next.
How do you build an AI Quoting workflow?
You build an AI Quoting workflow in four steps, and the last one never really ends:
- Build the context. Cover the three kinds above, starting from your vendor's documentation and your own historical quotes.
- Write prescriptive steps. AI is non-deterministic, which means it can answer differently each time. Numbered steps (check this, use this tool, confirm with the rep, then write the draft) make it consistent.
- Test on real deals. Run actual quotes through it. It will get some wrong.
- Fix every miss. When a quote fails, have the AI summarize what went wrong and add the fix to the skill. Every failure makes the next quote better.
Then roll it out as a plugin. A plugin bundles a set of skills so the whole team gets the same version, and every update reaches everyone at once. Mosaic's go-to-market plugin holds 44 skills. Turning it on is part of every new rep's onboarding, along with a setup check that confirms all the connections work.
The payoff: Andrew says quotes now take at least half the time in most cases, and accuracy is "way, way higher." That's about two to three months in, and it keeps improving with every fix.
Where should you start with AI Quoting?
Start AI Quoting with a single transaction type and a single selling motion, not your whole quoting process. Don't boil the ocean. Pick renewals, or new business for one product line, and get that working on real deals first.
You'll still need a baseline of shared context, like your products and pricing rules, before anything works. But a new quote and a renewal quote are built differently, and solving both at once is how projects stall. Get one working, test it yourself, bring in a few reps, then expand.
I spent a decade implementing NetSuite, and the projects that went live on time were never the ones that tried to automate everything in phase one. AI Quoting is no different.
Frequently asked questions
Does AI Quoting work with ChatGPT or only Claude?
AI Quoting works with any AI assistant that supports MCP, including Claude and ChatGPT, as long as your CPQ offers an MCP connection. Mosaic built its workflow in Claude because its team already used Claude every day, with its CRM and call recordings already connected. Pick the assistant your team already lives in, so the quoting happens where the reps already work.
Does AI Quoting replace quote approvals?
No. AI Quoting does not replace quote approvals. The CPQ keeps its hard approval rules, and the AI skill adds soft checks on top, such as written rules about what the AI should never do on a quote. The AI also proposes every change for the rep to confirm before it writes the draft, and only a person sends the quote.
How much time does AI Quoting save?
At Mosaic, AI Quoting cuts quote time by at least half in most cases, with much higher accuracy, according to Andrew Linford, VP of Operations. It wasn't faster at first, because reps were used to the old process and the AI hadn't learned the edge cases yet. The time savings grow as each failure gets fixed in the instructions.
Do I need to write the AI's instructions myself?
No. You don't need to write AI Quoting instructions by hand. Ask the AI to draft them from your vendor's documentation, your historical quotes and contracts, and your existing training material. Then refine them by testing on real deals. Andrew says he no longer writes skills by hand at all. He has Claude write them.
Get AI Quoting right from the start
AI Quoting only works as well as the CPQ underneath it. MonetizeNow, a quote-to-revenue solution for B2B SaaS covering CPQ, Billing and Revenue Recognition, connects to Claude and ChatGPT through MCP. Your reps quote in plain language while your pricing, approvals and rules stay governed in MonetizeNow Quoting. If you're weighing whether to build quoting yourself, read 10 reasons not to vibe code your CPQ first. And if your team is still building quotes by hand, start with the 7 problems with manually created order forms or our primer on what CPQ is.
Book a demo →
The short answer: To get started with AI Quoting, connect your CPQ (Configure, Price, Quote) tool to Claude or ChatGPT through MCP, then give the AI three kinds of context: how to use your CPQ's tools, your pricing and packaging rules, and data from your CRM. Start with one quote type, test it on real deals and fix every miss.
Key takeaways
- AI Quoting lets a rep ask for a quote in plain language and get back a draft quote in the CPQ, which a human reviews and sends.
- Connecting the tools is quick. Building the context is the real work, and it is what makes the AI quote the way your company sells.
- Before you build, check four things: leadership accepts a slower start, AI tools are approved, AI spend is approved and someone owns the build.
- Skills (saved, step-by-step instructions the AI follows) turn a different answer every time into the same answer every time.
- Start with one transaction type, such as renewals, and expand once it works.
Every RevOps leader I talk to wants AI Quoting. Very few know where to start.
So we asked someone who has already done it. Andrew Linford, VP of Operations at Mosaic, joined me for a live session on getting started with AI Quoting. His reps now build new quotes, renewals and amendments just by asking Claude. Here is the playbook he shared, in the order you would actually do it.
What is AI Quoting?
AI Quoting is a way of creating sales quotes where a rep describes the deal in plain language and an AI assistant builds the draft quote in the CPQ. A rep types something like "renew Globex at 40 users." The AI gathers the context (the CRM account, the current contract, product usage, call notes), applies your pricing and approval rules, shows the rep the proposed changes and writes a draft quote. Nothing goes to the customer until a human reviews it.
The connection between the AI and the CPQ usually runs through MCP (Model Context Protocol), an open standard that lets AI assistants like Claude and ChatGPT read from and act in other software. AI Quoting is also called conversational quoting, because the quote starts as a conversation instead of a form.
Why use AI for quoting instead of training reps on the CPQ?
AI Quoting removes the need to train every rep on the CPQ, because the knowledge lives in the AI's instructions instead of in each rep's head. Anyone who has run a CPQ knows the three mistakes that keep showing up on manual quotes:
- Inconsistent pricing and terms: different discounts, price increases or payment terms on similar deals.
- The wrong product mix: products combined that shouldn't be, or a required add-on left off.
- Missed details: an auto-renewal term or a promised price cap that never makes it onto the quote.
Training doesn't fix these, and I can tell you why from the sales seat. When RevOps puts CPQ training on my calendar, I'm either on a customer call or half listening. Then, months later, I finally need to build a complicated quote and I have no idea how. That's when the deal desk tickets start piling up.
Andrew put it better than I can:
"Sales reps are never getting hired for their detail orientation to quotes."
So stop asking them to be. With AI Quoting, a new rep can quote the right way on day one, and every rep quotes like your best rep.
Is your organization ready for AI Quoting?
An organization is ready for AI Quoting when four things are true. None of them is technical, and all of them are easier to settle before you build than after:
- Leadership accepts a slower start. At first, AI Quoting is slower than doing it by hand. It gets faster as you train it. If leaders expect a day-one win, the project will look like a failure in week two.
- AI tools are approved, including write access. Reading data through MCP is one thing. Letting the AI write draft quotes into your CPQ is another, and your IT team needs to be comfortable with it. At Mosaic, the AI writes drafts only, and only a rep ever sends a quote.
- AI spend is approved. Mosaic gives every rep a Claude Pro plan on a team subscription, which covers what reps use. If your company pays for AI by usage, expect finance to want a business case up front.
- Someone owns the build. One person has to build the first version and keep improving it after launch. At Mosaic, that's the operations team.
Andrew didn't hit any of these walls at Mosaic. He hears about them constantly from peers at larger companies, where one security review can cost weeks.
How does AI Quoting work behind the scenes?
AI Quoting works by chaining a rep's request through an AI assistant, a skill, the CPQ's tools and a human review. Here is the flow Andrew uses:
- A rep asks for a quote in Claude or ChatGPT, or pastes in a screenshot or email with the request.
- The AI picks the right skill for the job, such as a new quote, a renewal or an amendment.
- The skill follows its steps and reads the context: your products, rules and what needs approval.
- The AI calls the CPQ's tools through MCP, and pulls in other data if it's connected (CRM, call recordings, signed contracts).
- The AI shows the rep the before, the after and the changes, and asks for confirmation.
- The AI writes the draft quote in the CPQ. The rep reviews it, any approval rules run and the rep sends it.
Skills are what make this reliable. A Claude skill is a packaged set of instructions the AI loads when a task calls for it. You don't technically need one. But if you ask the AI the same quoting question in five different chats, you'll get five different answers. A skill tells it exactly what to do, in what order, every time.
Reps don't need to call skills by name, either. Andrew writes his skills so the AI recognizes when a request is about building, editing or renewing a quote and picks the right skill on its own.
What context does an AI Quoting agent need?
An AI Quoting agent needs three kinds of context: how to use the CPQ's tools, how your company prices and packages, and the outside data that shapes each deal. This is where most of the work goes. As Andrew said:
"You can hook up the tools fairly quickly, but having all the right context, that takes time and effort to build out."
- Tool instructions: which CPQ action to use in which situation. An MCP connection only carries a short description of each tool, so the AI needs more detail to stop guessing. Many vendors provide this in their API docs or a ready-made plugin.
- Your pricing and packaging rules: quote defaults, service hours, legal terms (and how you phrase them), per-user versus volume pricing, and rules like "this package always includes this onboarding."
- Outside data: the CRM fields to check, product usage, call recordings and the signed contract on file.
A renewal shows why the third kind matters. Say a customer has 50 users under contract but only 40 active. A good AI Quoting setup checks the CRM account, the usage data, the latest call recording (did they ask about an add-on?), the current contract in the CPQ, the signed agreement in your document storage and the account manager's input. Then it proposes 40 users and flags anything that doesn't line up.
The good news: you don't write this context by hand. Andrew says he hasn't written a line of his. The fastest start is to load your past contracts, quotes and invoices into the AI and have it summarize your pricing patterns. Existing quote training documents work too. That gets you roughly the first half of your quotes. The corner cases come next.
How do you build an AI Quoting workflow?
You build an AI Quoting workflow in four steps, and the last one never really ends:
- Build the context. Cover the three kinds above, starting from your vendor's documentation and your own historical quotes.
- Write prescriptive steps. AI is non-deterministic, which means it can answer differently each time. Numbered steps (check this, use this tool, confirm with the rep, then write the draft) make it consistent.
- Test on real deals. Run actual quotes through it. It will get some wrong.
- Fix every miss. When a quote fails, have the AI summarize what went wrong and add the fix to the skill. Every failure makes the next quote better.
Then roll it out as a plugin. A plugin bundles a set of skills so the whole team gets the same version, and every update reaches everyone at once. Mosaic's go-to-market plugin holds 44 skills. Turning it on is part of every new rep's onboarding, along with a setup check that confirms all the connections work.
The payoff: Andrew says quotes now take at least half the time in most cases, and accuracy is "way, way higher." That's about two to three months in, and it keeps improving with every fix.
Where should you start with AI Quoting?
Start AI Quoting with a single transaction type and a single selling motion, not your whole quoting process. Don't boil the ocean. Pick renewals, or new business for one product line, and get that working on real deals first.
You'll still need a baseline of shared context, like your products and pricing rules, before anything works. But a new quote and a renewal quote are built differently, and solving both at once is how projects stall. Get one working, test it yourself, bring in a few reps, then expand.
I spent a decade implementing NetSuite, and the projects that went live on time were never the ones that tried to automate everything in phase one. AI Quoting is no different.
Frequently asked questions
Does AI Quoting work with ChatGPT or only Claude?
AI Quoting works with any AI assistant that supports MCP, including Claude and ChatGPT, as long as your CPQ offers an MCP connection. Mosaic built its workflow in Claude because its team already used Claude every day, with its CRM and call recordings already connected. Pick the assistant your team already lives in, so the quoting happens where the reps already work.
Does AI Quoting replace quote approvals?
No. AI Quoting does not replace quote approvals. The CPQ keeps its hard approval rules, and the AI skill adds soft checks on top, such as written rules about what the AI should never do on a quote. The AI also proposes every change for the rep to confirm before it writes the draft, and only a person sends the quote.
How much time does AI Quoting save?
At Mosaic, AI Quoting cuts quote time by at least half in most cases, with much higher accuracy, according to Andrew Linford, VP of Operations. It wasn't faster at first, because reps were used to the old process and the AI hadn't learned the edge cases yet. The time savings grow as each failure gets fixed in the instructions.
Do I need to write the AI's instructions myself?
No. You don't need to write AI Quoting instructions by hand. Ask the AI to draft them from your vendor's documentation, your historical quotes and contracts, and your existing training material. Then refine them by testing on real deals. Andrew says he no longer writes skills by hand at all. He has Claude write them.
Get AI Quoting right from the start
AI Quoting only works as well as the CPQ underneath it. MonetizeNow, a quote-to-revenue solution for B2B SaaS covering CPQ, Billing and Revenue Recognition, connects to Claude and ChatGPT through MCP. Your reps quote in plain language while your pricing, approvals and rules stay governed in MonetizeNow Quoting. If you're weighing whether to build quoting yourself, read 10 reasons not to vibe code your CPQ first. And if your team is still building quotes by hand, start with the 7 problems with manually created order forms or our primer on what CPQ is.
Book a demo →
