A Claude Skill that reviews a proposed deal against your pricing and discount policy and returns a structured response: which terms are within policy, which need approval, what the recommended counter is, and which historical analogs support the proposed structure. Cuts deal-desk turnaround from days to minutes for routine asks.

What you’ll need

Salesforce with Quote and Opportunity Line Item objects
Claude Code with the Skill installed
Your pricing and discount policy as a Markdown doc
A reference set of approved past deals for analog matching
A standard quote template (PDF or DOCX)

Setup

Codify the policy. Edit references/pricing-policy.md with your discount tiers, multi-year incentive math, payment terms boundaries, and approval thresholds. Be explicit: which discount levels need which approval (manager, VP, CRO).
Install the Skill. Drop deal-desk-pricing.skill into ~/.claude/skills/. Set SFDC_TOKEN.
Load analogs. Place a CSV of the last 50 approved non-standard deals at references/analogs.csv with columns for ACV, discount, term length, payment terms, and approver.
Run. AE submits a deal via a Slack command or a form that hits the Skill: review_deal(opp_id="..."). Skill returns a one-page review.
Approval routing. If everything is in policy, the Skill auto-approves. If something needs review, it routes to the right approver with the gap surfaced.

How it works

The Skill pulls the proposed quote line items and contract terms from Salesforce, then evaluates each against the policy doc. It computes effective discount, term-adjusted ACV, payment terms cost, and any custom-terms drift. Each evaluation returns one of: in-policy, exception-needed, out-of-policy.

For exceptions, it pulls the closest five historical analogs from the CSV and shows whether similar asks were approved — and if so, by whom and under what conditions. This converts deal desk from a gatekeeper into a context provider; the approver sees precedent at a glance.

The recommended counter is a small but high-value output. The Skill suggests a structure that hits the buyer’s effective price target while respecting policy: trade higher discount for longer term, trade upfront discount for ramp.

Watch-outs

Policy drift. Pricing policy is updated quarterly. The Skill is only as fresh as the doc. Set a reminder to refresh references on a schedule.
Analog selection bias. If your CSV is full of “we approved everything” deals, the Skill will rubber-stamp future asks. Curate the analogs.
Auto-approval scope. Start narrow. Only auto-approve the truly trivial — single-year, in-policy, list-price deals. Expand the scope after watching a few weeks of decisions.

Stack

Salesforce — quote and opportunity data
Claude — policy evaluation, counter recommendation, analog matching