B2B SaaS · AI-assisted sales proposals

B2B SaaS company prevented $340K in unauthorized AI-generated pricing commitments

Published April 20, 2026

The problem

The customer used an AI assistant to help their sales team draft proposals and commercial emails. Sales discovered an AI-generated proposal had quoted a custom infrastructure configuration at $2,400/mo — a price the AI derived from pattern-matching public pricing pages, but which was $800/mo below the actual cost of the configuration. The proposal had been signed by the customer before the error was caught.

The risk

The signed proposal at the wrong price created a contractual obligation the company had to honor at a $9,600 annual loss on a single deal. Analysis of all AI-assisted proposals from the prior 90 days revealed 12 additional proposals with pricing errors, including 3 already signed. Total unauthorized commitment: $340K annualized.

Implementation

Configured DataVibe with a sales compliance policy: BLOCK rules for any pricing figure outside approved price book ranges (regex on dollar amounts with validation against approved tier list), BLOCK on discount percentages exceeding 15% (sales manager pre-approval required above that), WARN on custom configuration language that requires pricing team sign-off, and WARN on SLA commitments not in the standard order form. All proposal content from the AI assistant routes through the gate before the sales rep can send.

What changed

• Zero pricing errors in AI-generated proposals in the 90 days post-deployment.
• $340K in unauthorized commitments prevented in the first quarter alone.
• Average proposal review time: 28 seconds (sales manager reviews flagged items only).
• Deal velocity unchanged — 94% of proposals cleared the gate automatically with no review required.

Sales proposal AI — pricing error prevention

// Proposal content gate — catches pricing errors before the email is sent.
// Sales reps draft with AI; gate intercepts before they hit "send."

const result = await dv.intercept({
  recipient: prospect.email,
  subject:   `${company.name} — Proposal: ${configurationName}`,
  body_html: aiGeneratedProposal,
  metadata: {
    deal_id:          deal.salesforceId,
    rep_id:           rep.id,
    configuration_id: configuration.id,
    // Gate uses this to validate against approved price book
    quoted_mrr:       parsedMrrFromProposal,
  },
});

if (result.status === "QUEUED") {
  // Pricing rule fired — sales manager review required
  // result.policy_violations[0].detail shows exact line that triggered
  await notifySalesManager(result.review_url, deal.id);
}