Sales AI Agents,
Monitored for Accuracy

1. Target

   Quote generation, proposal terms, SLA commitments and discount fields in CPQ and CRM

2. Trigger

   Draft quote or reply exceeding the rep’s delegation-of-authority matrix for discount, term or SLA

3. Action — enforced

   Platform holds the quote unsent and strips the offending term; agent may request approval or offer in-matrix alternatives.

4. Human override

   Sales VP via deal-desk approval recorded against the opportunity id

Logged evidenceopportunity id · requested vs authorized discount · matrix version · approver identity · UTC timestamp

1. Target

   CRM records, quotes, battlecards and negotiation notes partitioned per account and opportunity

2. Trigger

   Retrieval or generation referencing one account’s pricing, terms or strategy inside another account’s session

3. Action — enforced

   Platform blocks the fetch and scrubs session memory of foreign-account content; agent may use only current-account records.

4. Human override

   None — cannot be overridden in session

Logged evidencesession id · blocked record ids · account partition labels · memory-scrub hash · block timestamp

1. Target

   Inbound prospect emails, uploaded RFP documents, calendar invites and attachments the agent parses

2. Trigger

   Imperative or tool-directing text detected inside any prospect-supplied message, document or attachment

3. Action — enforced

   Platform neutralizes the content to quoted data and refuses execution; agent may summarize it and continue the deal.

4. Human override

   None — cannot be overridden in session

Logged evidencemessage id · detected instruction span · sender domain · sanitized copy hash · detection timestamp

1. Target

   Order forms, e-signature workflows, acceptance language in emails and chat across all deals

2. Trigger

   Draft output constituting acceptance, signature or agreement to terms on the company’s behalf

3. Action — enforced

   Platform blocks the send and watermarks drafts non-binding; agent may prepare the contract for a human signatory.

4. Human override

   Commercial director via the e-signature platform’s recorded signing ceremony

Logged evidencedraft hash · blocked clause list · opportunity id · signatory of record · block timestamp

1. Target

   Proposals, outreach copy, battlecards and demo scripts making product capability or customer-proof claims

2. Trigger

   Claim, testimonial or logo lacking a match in the approved substantiation library

3. Action — enforced

   Platform blocks the send and flags each unsupported claim; agent may substitute approved claims or request substantiation.

4. Human override

   Product marketing lead via substantiation-library entry citing verifiable evidence

Logged evidencedocument id · flagged claim list · substantiation record ids · reviewer identity · review timestamp

1. Target

   Email sequences, dialers and SMS campaigns governed by CAN-SPAM, TCPA and Spam Act rules

2. Trigger

   Send or dial targeting a suppressed contact, missing lawful basis, or outside permitted hours

3. Action — enforced

   Platform blocks the send and quarantines the sequence step; agent may target only consented contacts within permitted hours.

4. Human override

   Marketing-compliance officer via suppression-list correction with documented consent proof

Logged evidencecontact id · suppression-list version · consent record · attempted channel and time · block timestamp

1. Target

   Chat, email, voice and SMS conversations where the counterparty could assume a human rep

2. Trigger

   Conversation start, or direct question about identity, without an AI disclosure on record

3. Action — enforced

   Platform injects the AI disclosure and blocks denial of bot status; agent may continue once disclosure is delivered.

4. Human override

   Legal counsel via approved disclosure-script change logged in the policy register

Logged evidenceconversation id · disclosure text version · delivery timestamp · channel · counterparty query span

1. Target

   AI note-takers and call-recording bots joining discovery calls, demos and negotiations

2. Trigger

   Recording or transcription starting before every participant’s consent is captured for that jurisdiction

3. Action — enforced

   Platform blocks capture and discards buffered audio; agent may restart recording only after every participant consents.

4. Human override

   Sales-operations manager via jurisdiction-consent checklist completed before the call

Logged evidencemeeting id · participant consent states · jurisdiction rule applied · audio-discard hash · stop timestamp

1. Target

   Pricing recommendations, discount guidance and win-loss models fed by external market data

2. Trigger

   Recommendation input containing non-public competitor pricing or shared-vendor pooled market feeds

3. Action — enforced

   Platform blocks the recommendation and purges the tainted feed; agent may price from public lists and internal costs.

4. Human override

   General counsel via antitrust review clearing the data source in writing

Logged evidencefeed source id · data provenance record · blocked recommendation hash · counsel review id · purge timestamp

1. Target

   Deal approvals, payment-detail changes and executive requests arriving by voice, video or chat

2. Trigger

   High-value instruction authenticated only by claimed identity, voice or video resemblance

3. Action — enforced

   Platform holds the action pending out-of-band verification on a directory-listed channel; agent may prepare but not execute.

4. Human override

   Finance controller via callback verification to the directory number of record

Logged evidencerequest id · claimed identity · verification channel and outcome · hold duration · release timestamp