Sales rep approving an AI quoting step in Salesforce CPQ on a laptop right after a client call

Part Three: AI Quoting in Salesforce CPQ, From Call to Quote

This is the last part of a three-part series. Part one made the case for keeping your systems of record. Part two showed the architecture: Salesforce is the spine, Claude is the cockpit. This part shows it working, live, on the job reps hate most. AI quoting in Salesforce CPQ, start to finish, without anyone rebuilding CPQ.

You know the drill. You have a call in ten minutes. You open the opportunity, then the account, then the contacts, then your notes from last time. After the call you type a summary nobody will read, update three fields, and open the quoting tool. Forty minutes later you have a quote. Maybe.

Now picture the same deal in one seat. One sentence before the call. One click after it. One approval to build the quote. Salesforce still makes every rule.

The short version

  • One seat, four jobs. Call prep, opportunity updates, the CPQ quote and the customer email all run from one Claude conversation.
  • CPQ stays in charge. Claude builds the quote through Salesforce CPQ, so every price rule, discount rule and approval still applies.
  • Draft first, always. Every change shows up as a preview. Nothing saves until the rep approves.
  • The rep’s own permissions. Claude works with the rep’s Salesforce access, through a short list of purpose-built tools, and you can narrow it further.
  • The CRM stays current. The close date, the next step, the logged call and the draft quote land in Salesforce without anyone typing a field.

Reps Spend Most of the Week Not Selling

The problem is not new, and it is getting worse. Salesforce’s own State of Sales research found that reps spend just 28% of their week actually selling, down from 34% in 2018. The rest goes to admin, deal management and data entry.

Quoting sits right in the middle of that admin. It is detailed, it is rules heavy, and it usually means bouncing between the CRM, your notes and the quoting screen.

28%of a rep’s week goes to selling
34%the same number in 2018
4jobs between the call and the quote
1seat where Claude now runs all four

Salesforce sees the same thing. Its AIforce strategy brings Salesforce to where people already work, including Claude. Our demo is what that looks like on a real sales motion.

The Demo: One Deal, Call to Quote

Gavin Franklin on our team built this in a Salesforce sandbox with a made-up customer, Maple Grove Health Partners, and a generic product catalog. The rep has a call coming up on a $162,000 platform expansion. Here is what happens next. Watch the two-minute walkthrough, then hover or tap each card below.

1. The brief

One sentence: prep me for my call with Maple Grove.

Hover for detail

1. The brief

Claude finds the two open opportunities, asks which one, and returns a call card: stage, amount, close date, the buying group, who was never contacted, and five questions to ask.

2. The update

One click after the call: update the opportunity.

Hover for detail

2. The update

Claude reads the call transcript and proposes three changes: a new close date, a firm next step and the call logged as an activity. All shown as a preview. Nothing saves until the rep clicks Approve.

3. The quote

Build the quote, through Salesforce CPQ.

Hover for detail

3. The quote

Claude asks for the start date it is missing, matches five products from the call, and flags a support line that may overlap a renewal the customer already has. CPQ calculates the final price.

4. The send

Quote PDF and customer email, drafted.

Hover for detail

4. The send

Claude drafts the PDF and the email to the buyer, saves the file where the rep wants it, and stops. Nothing goes out until the rep says send.

Back in Salesforce, the opportunity shows the new close date, the next step and the logged call. A draft quote sits on the deal with five lines, calculated by CPQ. The rep never typed a field.

Sales manager and rep reviewing a Salesforce CPQ quote built by Claude on a wall display

Why AI Quoting Has to Run Through Salesforce CPQ

There is a shortcut that looks tempting. Let the AI write quote lines straight into the database through an API and skip the quoting screen. Do not do that. A raw API write can skip the same guardrails your team relies on in the CPQ screen.

So our build does the opposite. Claude reads the CPQ setup for the products it works with, then builds the quote through the CPQ calculation itself. That means:

  • Price rules apply. List prices, billing basis and terms come from your price book, not from Claude’s guess.
  • Discount rules apply. When the buyer asks for 10% off, Claude flags it and asks. It does not hand-calculate a number.
  • Approvals apply. The quote follows your stages and approval process. Claude does not jump a stage.
  • The rep owns the quote. Once the draft exists, the rep manages it in Salesforce like any other quote.

“We didn’t rebuild CPQ. We gave it a new front door.”

Chad Anderson, CEO, Cirrius Solutions

What Keeps a Claude Salesforce CPQ Integration Safe

A cockpit that can write to your CRM needs tight controls. These are the ones we build in from day one.

  • The rep’s own credentials. Claude acts as the signed-in user. If the rep cannot do it in Salesforce, Claude cannot either.
  • Purpose-built tools only. Claude reaches Salesforce through a short list of custom MCP tools: update these fields, log this activity, build this quote. No open access to the org.
  • Narrower than the profile, if you want. A rep who can edit an account can still be blocked from doing it through Claude.
  • Draft first, approve second. Every update shows up as a preview card with Approve, Request changes and Discard.
  • Email on your terms. The email connection can be draft only or send on approval, whatever your security team prefers.

This is the same governed model Salesforce is building toward with Agentforce and its hosted MCP servers. The cockpit and the autopilot work together, and both respect your rules.

Questions We Get About AI Quoting in Salesforce

Does Claude replace Salesforce CPQ?

No. Claude is a front door to CPQ, not a copy of it. CPQ still holds the products, prices, rules and approvals, and it still calculates the quote.

Can Claude send a quote to a customer on its own?

Only if you configure it that way. In our demo Claude drafts the PDF and the email, and the rep reviews and sends. Draft only is the default we recommend.

Does it work with DocuSign or Conga?

Yes. Claude can hand the quote to the document tool you already use, or the rep can keep generating documents from Salesforce. Every step in the pipeline can be swapped out or switched off.

Where does Agentforce fit?

Agentforce runs governed, high-volume work inside Salesforce. The Claude cockpit is where a person sees across systems and approves the next step. They make each other better.

Where to Start

1

Pick one painful motion. Call to quote is a great first target because it touches the CRM, notes, CPQ and email.

2

Write down the guardrails first. Which fields Claude may change, which products it may quote, and who approves what.

3

Build narrow tools, not open access. A handful of purpose-built MCP tools beats broad API access every time.

4

Prove it in a sandbox with real reps. Run a week of real deals with made-up data, then measure the time back.

We Build This

Cirrius Solutions has delivered more than 2,500 Salesforce implementations, and we now build Claude cockpits on top of them. We keep your Salesforce investment working and give your reps their week back.

Want to see call to quote running against your own CPQ setup? Talk to our team.

Missed the start of the series? Read Part one: Do Not Replace Salesforce With AI and Part two: Salesforce Is the Spine, Claude Is the Cockpit.