# AI Ready Reps: Sales AI Enablement Guide (Company Edition)

> **Status:** TEMPLATE. Replace everything in [BRACKETS] and have your Legal, IT, and Security teams review this document before you publish it internally.
>
> **Version:** 1.0  |  **Last updated:** October 14, 2026  |  **Vendor facts last checked:** October 14, 2026
>
> **Source:** Adapted from the public AI Ready Reps site at https://aireadyreps.com by Jared Chapman. Released under the MIT License (see the repository). This is general information, not legal advice.

## How to use this template

1. Copy this file and replace each [BRACKETED] item with your company's details.
2. Delete any section you do not need, and add company-specific rules.
3. Have Legal, IT, and Security review it before it is shared with your team.
4. Review it at least every quarter, because vendor terms, laws, and tools change often.

## 1. Purpose and scope

This guide explains how [COMPANY NAME] employees in sales and go-to-market roles may use AI tools safely. It applies to [WHO IT APPLIES TO, for example: sales, marketing, customer success, sales engineering, and revenue operations] and to every AI tool used for work, whether company-provided or not.

**Core principle:** AI is a draft partner. You are accountable for everything you send, say, or record.

## 2. Who to ask

| Topic | Contact |
|---|---|
| Which AI tools are approved | [IT CONTACT OR CHANNEL] |
| Customer data, contracts, privacy, and legal questions | [LEGAL CONTACT OR CHANNEL] |
| Security incidents and mistakes | [SECURITY CONTACT OR CHANNEL] |
| Questions about this policy | [POLICY OWNER] |

## 3. Approved tools

| Tool | Plan or license | Approved for | Not approved for | Owner |
|---|---|---|---|---|
| [TOOL] | [PLAN] | [DATA TYPES AND USES] | [DATA TYPES AND USES] | [OWNER] |
| [TOOL] | [PLAN] | [DATA TYPES AND USES] | [DATA TYPES AND USES] | [OWNER] |

Anything not listed above is **not approved** until IT and Legal approve it.

## 4. The traffic-light rule

| Light | Meaning | Example |
|---|---|---|
| Green (safe) | Your own writing or public information, in an approved tool, with your review of the result | Tighten an email you wrote |
| Amber (guardrails) | Touches customer or company data. Allowed only with the listed control and sign-off | Summarize a recorded call with consent |
| Red (not safe) | Breaks confidentiality, privacy, consent, or honesty. Do not do it | Paste a customer contract into an unapproved tool |

## 5. Three questions before any prompt

1. **What data goes in?** Public, internal, or customer/regulated. Customer and regulated data needs written approval.
2. **Who checks the output?** A named person reads it and verifies every fact, number, name, and claim before it reaches a customer.
3. **What if it is wrong?** Low impact is a clumsy draft. Medium is a wrong fact in a customer email. High is a pricing, legal, security, or compliance statement and needs expert sign-off.

## 6. Data classification

| Class | Examples | Rule |
|---|---|---|
| Public | Websites, press releases, filings | May be used in approved tools |
| Internal | Playbooks, non-public pricing, forecasts | Approved enterprise tools only |
| Customer or confidential | Contracts, personal data, customer lists | Only with written approval from [LEGAL/IT] |
| Regulated | Health, payment, government ID, financial data | Never, unless explicitly approved for that tool and use |

## 7. Rules for everyone

- Use only company-approved accounts. Never use personal accounts for work data.
- Share the minimum needed, and replace names with placeholders when you can.
- Never put passwords, keys, payment details, or health data into a prompt.
- Verify every fact, number, quote, and claim against a trusted source before use.
- Do not invent references, testimonials, or statistics.
- Do not let AI set prices, discounts, or contract terms.
- Follow call-recording and consent rules for every region you sell in.
- Disclose AI use where law or contract requires it, and never deny being an AI when someone sincerely asks.
- Keep a person approving anything an AI agent sends or changes outside your account.
- Report mistakes fast. Early reports are valued.

## 8. Use case catalog

This catalog contains **287** use cases: 109 green, 98 amber, and 80 red. Adjust the rating for each to match your policy.

### Prospecting (48)

| Light | Use case | Type | Guardrail or reason |
|---|---|---|---|
| Green (safe) | Brainstorm subject lines for a cold email you wrote | Augmentation | Uses your own copy and no prospect data. |
| Green (safe) | Tighten your own outreach draft so it is shorter and clearer | Augmentation | Review the result before sending. |
| Green (safe) | Summarize a public annual report or earnings call into five bullets | Analysis | Public source. Check figures against the original. |
| Green (safe) | Draft an account brief from a company website and press releases | Augmentation | Public information only. |
| Green (safe) | Generate first-call questions for an industry you do not know | Augmentation | General knowledge, no customer data. |
| Green (safe) | Explain unfamiliar industry jargon before a call | Augmentation | Low impact if slightly off. Confirm in conversation. |
| Green (safe) | List likely buyer personas for a product category | Augmentation | Brainstorming aid. Validate with real customers. |
| Green (safe) | Write connection-note templates with placeholders | Augmentation | No real names go into the tool. |
| Green (safe) | Turn your own win patterns, described in general terms, into an ideal-customer checklist | Analysis | Keep customer names out. |
| Green (safe) | Fix grammar and tone in a message you wrote | Augmentation | Your own text. No confidential content. |
| Green (safe) | Brainstorm angles for a first email to a role you have never sold to | Augmentation | General brainstorming with no prospect data. |
| Green (safe) | Draft a short voicemail script for a cold call | Augmentation | Your own script. Read it aloud and edit it. |
| Green (safe) | Write three versions of an opening line for a call | Augmentation | Practice material you can adapt. |
| Green (safe) | List common reasons a prospect in a given industry ignores outreach | Augmentation | General reasoning. Test it against real replies. |
| Green (safe) | Draft a polite message asking a colleague for an introduction | Augmentation | Your own words and no confidential data. |
| Green (safe) | Learn a few polite phrases in a customer's language | Augmentation | Learning aid. Ask a native speaker to confirm tone. |
| Green (safe) | Turn a feature list into one-sentence benefits using approved public copy | Augmentation | Stay inside approved messaging. |
| Green (safe) | Create a checklist for researching a new account before a first call | Augmentation | A reusable process template. |
| Green (safe) | Turn a long public blog post into talking points | Augmentation | Public source. Check anything you will quote. |
| Amber (guardrails) | Write outreach in a customer's local language and have a native speaker review it | Augmentation | native-speaker review before anything is sent. |
| Amber (guardrails) | Infer likely priorities from a company's public job postings | Analysis | public postings only, and treat conclusions as hypotheses. |
| Amber (guardrails) | Draft personalized messages that reference a prospect's public professional post | Augmentation | professional context only, nothing personal or sensitive. |
| Amber (guardrails) | Draft answers to common prospect questions in a website chat | Augmentation | approved answers only, and escalate anything unknown to a person. |
| Amber (guardrails) | Use an AI research assistant that browses the web for account news | Augmentation | read-only access, open every cited source, and keep confidential data out of the query. |
| Amber (guardrails) | Combine a customer's public earnings commentary with your CRM notes | Augmentation | approved enterprise tool, and verify every figure. |
| Amber (guardrails) | Generate personalized video scripts for outreach | Augmentation | review for accuracy and tone, and disclose AI where required. |
| Amber (guardrails) | Personalize outreach at scale from CRM fields | Automation | human spot-checks every batch, no sensitive fields, and compliance with anti-spam and privacy rules. |
| Amber (guardrails) | Enrich account lists with AI research | Analysis | verify each fact and keep the source link. |
| Amber (guardrails) | Score or prioritize leads with AI | Analysis | documented criteria, bias check, and a human decides. |
| Amber (guardrails) | Summarize a prospect's public professional posts for context | Analysis | professional information only, and follow each platform's terms. |
| Amber (guardrails) | Draft multi-step outreach sequences | Automation | a manager approves every template before launch. |
| Amber (guardrails) | Generate account plans from internal CRM data | Augmentation | approved enterprise tool only. |
| Amber (guardrails) | Run an AI assistant that books meetings | Automation | disclose AI where required, honor opt-outs, limit what it may claim. |
| Red (not safe) | Use AI to infer a prospect's health, finances, or family situation from their activity | Analysis | Sensitive inferences are invasive and often unlawful. Stick to professional, public context. |
| Red (not safe) | Ask AI to write a message pretending to come from a colleague or executive | Augmentation | Impersonation is deceptive and can be fraud. |
| Red (not safe) | Create deepfake video or voice of a customer or executive | Augmentation | Impersonation and likely illegal. |
| Red (not safe) | Have AI scrape a competitor's customer list from a gated portal | Automation | Unauthorized access and likely illegal. |
| Red (not safe) | Ask AI to dig up private information about a person to use as pressure | Augmentation | Harassment and privacy risk. |
| Red (not safe) | Use AI to run fake accounts that boost social engagement | Automation | Deceptive and against platform rules. |
| Red (not safe) | Use a real customer's name or data as an example in a public forum prompt | Augmentation | Confidentiality breach. Use fictional examples. |
| Red (not safe) | Ask AI to guess a prospect's personal email address or phone number | Augmentation | Guesses are often wrong and may break privacy and anti-spam rules. Use approved data sources. |
| Red (not safe) | Ask AI to evade spam filters or hide who is really sending | Augmentation | Deceptive, and it breaks sending rules. |
| Red (not safe) | Auto-send unreviewed AI emails at scale | Automation | One error replicates thousands of times. Instead: review a sample of every batch. |
| Red (not safe) | Scrape personal data from social profiles to build individual profiles | Automation | Privacy-law risk and platform-terms violation. |
| Red (not safe) | Build lead lists from scraped personal emails and mass-mail them | Automation | Anti-spam and privacy violations. |
| Red (not safe) | Clone an executive's or customer's voice or likeness without consent | Augmentation | Impersonation and fraud risk. |
| Red (not safe) | Target or score people by protected traits such as age, race, or religion | Analysis | Discrimination risk. |
| Red (not safe) | Upload a competitor's confidential material you were not meant to have | Augmentation | Legal exposure. Report it to Legal instead. |

### Discovery (26)

| Light | Use case | Type | Guardrail or reason |
|---|---|---|---|
| Green (safe) | Role-play a discovery call with AI as a skeptical, fictional buyer | Augmentation | Fictional scenario. Great practice. |
| Green (safe) | Build a discovery question bank organized by your qualification framework | Augmentation | Generic and reusable. |
| Green (safe) | Turn non-confidential meeting notes into a clean agenda | Augmentation | Strip customer names and numbers first. |
| Green (safe) | Draft a pre-call agenda template | Augmentation | No customer data needed. |
| Green (safe) | Practice answers to common objections | Augmentation | Fictional scenarios only. |
| Green (safe) | List what could go wrong in a typical deal of your type | Augmentation | Generic risk brainstorm. |
| Green (safe) | List follow-up questions for common buyer answers such as "we are happy with our vendor" | Augmentation | General technique with no customer data. |
| Green (safe) | Draft a script for opening a call and setting the agenda | Augmentation | A template you can adjust. |
| Green (safe) | Practice restating a buyer's problem in two sentences | Augmentation | Fictional practice. Great for listening skills. |
| Green (safe) | Brainstorm ways to handle silence on a call | Augmentation | General technique. |
| Green (safe) | Draft a meeting-request message that offers two time options | Augmentation | Template with placeholders. |
| Green (safe) | Build a one-page glossary of terms for a new industry | Augmentation | Public knowledge. Confirm important terms with an expert. |
| Amber (guardrails) | Prepare a stakeholder map from CRM contact notes | Analysis | approved enterprise tool and minimal personal data. |
| Amber (guardrails) | Generate follow-up emails from call transcripts | Automation | the rep edits and confirms every fact before sending. |
| Amber (guardrails) | Create a customer-specific demo script from discovery notes | Augmentation | remove confidential details and have a sales engineer review it. |
| Amber (guardrails) | Transcribe and summarize customer calls | Automation | recording consent, vendor retention terms approved by Legal. |
| Amber (guardrails) | Extract action items from call transcripts | Automation | rep confirms before anything is shared. |
| Amber (guardrails) | Auto-fill qualification fields in the CRM from calls | Automation | rep reviews and accepts each field. |
| Amber (guardrails) | Suggest talk tracks during live calls | Augmentation | disclose to participants where required, and never read suggestions blindly. |
| Amber (guardrails) | Analyze your call library for objection patterns | Analysis | de-identify, and restrict access. |
| Amber (guardrails) | Summarize a long customer email thread | Analysis | approved tool, and check the summary against the thread. |
| Red (not safe) | Use AI to screen out prospects based on their accent, name, or appearance | Analysis | Discriminatory. Never use protected traits. |
| Red (not safe) | Upload customer call recordings to a free transcription site | Automation | Confidential and possibly in breach of consent rules. |
| Red (not safe) | Paste a prospect's private messages into a tool for tone analysis | Analysis | Private communication shared with a third party. |
| Red (not safe) | Record or transcribe calls with an AI note-taker without consent | Automation | Violates recording laws in many places. |
| Red (not safe) | Pass an AI off as a human when disclosure is required or someone sincerely asks | Augmentation | Deceptive, and illegal in some places. |

### Proposal (39)

| Light | Use case | Type | Guardrail or reason |
|---|---|---|---|
| Green (safe) | Create a proposal outline with placeholders | Augmentation | Fill in real details yourself. |
| Green (safe) | Rewrite already-public product copy in plain English | Augmentation | Source text is already approved and public. |
| Green (safe) | Draft an executive-summary template | Analysis | Structure only. |
| Green (safe) | Write alternate phrasings of approved value propositions | Augmentation | Stay inside approved claims. |
| Green (safe) | Build an FAQ from your public product documentation | Augmentation | Public source. Product team spot-checks. |
| Green (safe) | Outline a presentation from a topic you describe | Augmentation | You supply the facts. |
| Green (safe) | Draft a proposal cover letter with placeholders | Augmentation | Fill in real details yourself. |
| Green (safe) | Explain your pricing model in plain language from your public pricing page | Augmentation | Public source only. Confirm accuracy. |
| Green (safe) | Write a "what happens next" section for a proposal template | Augmentation | Structure only. |
| Green (safe) | Draft a thank-you note after a presentation | Augmentation | Template. Add specifics yourself. |
| Green (safe) | Suggest a clearer title and structure for a slide deck you wrote | Augmentation | Your own work. Remove confidential content first. |
| Green (safe) | Rewrite long approved product paragraphs as short bullets | Augmentation | Approved copy only. Check the meaning did not change. |
| Amber (guardrails) | Draft a business case document from discovery data | Analysis | numbers verified and inputs confirmed by the customer. |
| Amber (guardrails) | Build competitor comparison tables for internal use from public sources | Analysis | cite sources, and have product marketing review before external use. |
| Amber (guardrails) | Draft a summary of contract clauses for your own understanding | Analysis | Legal interprets the contract. Never rely on the AI summary. |
| Amber (guardrails) | Prepare an executive briefing for a strategic account | Augmentation | approved tool, verified facts, and limited sensitive data. |
| Amber (guardrails) | Create tailored value messaging by industry from approved positioning | Augmentation | product marketing approval. |
| Amber (guardrails) | Draft a mutual success plan from stakeholder goals | Augmentation | the customer reviews and agrees to it. |
| Amber (guardrails) | Draft talking points for an executive sponsor meeting | Augmentation | verify facts and align with the account lead. |
| Amber (guardrails) | Draft answers to RFPs and security questionnaires from an approved answer library | Augmentation | subject-matter expert and Legal review every answer. |
| Amber (guardrails) | Draft proposals from templates plus customer details | Augmentation | a person checks pricing, scope, and terms line by line. |
| Amber (guardrails) | Assemble slides from approved case studies | Augmentation | only approved facts and quotes. |
| Amber (guardrails) | Write an ROI narrative | Augmentation | calculate numbers outside the AI and verify them. |
| Amber (guardrails) | Translate proposals into other languages | Augmentation | native-speaker review. |
| Amber (guardrails) | Summarize a customer's contract redlines for yourself | Analysis | Legal makes every decision. Approved tool only. |
| Amber (guardrails) | Draft a mutual action plan | Augmentation | both sides review dates and owners. |
| Red (not safe) | Have AI generate fake case study numbers to hit a deadline | Augmentation | Fabrication. Use real, approved results only. |
| Red (not safe) | Ask AI to write a negative review about a competitor | Augmentation | Deceptive and potentially defamatory. |
| Red (not safe) | Feed a customer's proprietary source code or designs into a public tool | Augmentation | Intellectual property and confidentiality breach. |
| Red (not safe) | Insert AI-written legal terms into a customer contract without Legal review | Augmentation | Contract language needs qualified review. |
| Red (not safe) | Publish customer logos or names generated by AI without permission | Augmentation | Needs written customer approval. |
| Red (not safe) | Use AI to fabricate a testimonial or review | Augmentation | Fake endorsements are deceptive and often illegal. Use real, approved customer quotes. |
| Red (not safe) | Paste another company's internal documents you received by mistake | Augmentation | Do not open or share them. Tell Legal. |
| Red (not safe) | Paste unreleased roadmap or NDA material | Augmentation | Breaks your NDA obligations. |
| Red (not safe) | Invent customer references, quotes, case studies, or statistics | Augmentation | Fabrication. AI can produce convincing fakes, so use only verified material. |
| Red (not safe) | Give customers legal, tax, or regulatory advice from AI output | Augmentation | Needs a qualified professional. |
| Red (not safe) | State security or compliance certifications from AI memory | Augmentation | Wrong claims create liability. Use the vendor's current documents. |
| Red (not safe) | Trust AI arithmetic in a quote or ROI without checking | Analysis | Models can miscalculate. Use a spreadsheet and verify. |
| Red (not safe) | Submit AI-written security questionnaire answers without verification | Augmentation | These are formal attestations. |

### Negotiation (19)

| Light | Use case | Type | Guardrail or reason |
|---|---|---|---|
| Green (safe) | Role-play a negotiation with a fictional buyer | Augmentation | No real terms or customers. |
| Green (safe) | Brainstorm a trade-off framework in generic terms | Augmentation | No real pricing goes in. |
| Green (safe) | Draft polite follow-up templates for silent prospects | Augmentation | Templates with placeholders. |
| Green (safe) | Practice staying calm and polite when a buyer pushes back hard | Augmentation | Fictional role-play. |
| Green (safe) | Draft a respectful way to say no to a request outside your authority | Augmentation | Template wording. |
| Green (safe) | Brainstorm questions that uncover a buyer's real priorities before discussing terms | Analysis | General technique. |
| Amber (guardrails) | Draft a response to a quote request using approved price lists | Augmentation | pricing and terms come only from approved sources. |
| Amber (guardrails) | Prepare negotiation strategy from deal context | Augmentation | approved enterprise tool and minimal data. |
| Amber (guardrails) | Compare differences between competing quotes | Analysis | verify each figure. |
| Amber (guardrails) | Draft concession-request emails | Augmentation | review tone and commitments before sending. |
| Red (not safe) | Paste a competitor's leaked confidential pricing sheet into a tool | Augmentation | Using it creates legal exposure. Tell Legal instead. |
| Red (not safe) | Let AI approve discounts or credits in a billing system without review | Automation | Financial commitments need an authorized human. |
| Red (not safe) | Ask AI to find loopholes in a customer's contract to your advantage without Legal | Augmentation | Contract interpretation belongs with Legal, and bad-faith tactics harm trust. |
| Red (not safe) | Let AI accept or sign online terms on a customer's behalf | Automation | Binding acts need an authorized person. |
| Red (not safe) | Use AI answers about export controls or sanctions as a final decision | Augmentation | Compliance questions need qualified experts. |
| Red (not safe) | Paste a customer contract or MSA into a consumer or unapproved AI tool | Augmentation | Breaches confidentiality. Instead: an approved enterprise tool with Legal's written OK. |
| Red (not safe) | Share non-public pricing, margins, or discount matrices in an unapproved tool | Augmentation | Competitive and contractual risk. |
| Red (not safe) | Let AI set prices, discounts, or contract commitments | Augmentation | Binding commitments need a human with authority. |
| Red (not safe) | Use AI to fabricate urgency, scarcity, or false claims | Augmentation | Deceptive practice. |

### Post-sale (33)

| Light | Use case | Type | Guardrail or reason |
|---|---|---|---|
| Green (safe) | Draft a kickoff email template | Augmentation | Add customer specifics yourself. |
| Green (safe) | Create an onboarding checklist template | Augmentation | Generic structure. |
| Green (safe) | Draft a QBR agenda template | Augmentation | No customer data needed. |
| Green (safe) | Write a generic renewal-reminder template | Augmentation | Placeholders only. |
| Green (safe) | Turn public release notes into a friendly customer update | Augmentation | Public source. Product team approves. |
| Green (safe) | Draft a welcome message for new customers with placeholders | Augmentation | Template. |
| Green (safe) | Write a short "how to get help" guide from your public support pages | Augmentation | Public source. |
| Green (safe) | Draft a thank-you note for a customer referral with placeholders | Augmentation | Template. Personalize it yourself. |
| Green (safe) | Create a template for asking a happy customer for a review | Augmentation | Template. Only ask people who agreed to be contacted. |
| Amber (guardrails) | Draft a customer email about a delayed delivery or issue | Augmentation | facts confirmed with operations, and manager approval for sensitive cases. |
| Amber (guardrails) | Create a customer onboarding plan from the signed scope | Augmentation | the delivery lead reviews it. |
| Amber (guardrails) | Rewrite a support article so sales can share it with customers | Augmentation | the support owner approves. |
| Amber (guardrails) | Build a quarterly review deck from usage exports | Augmentation | de-identify data, verify figures, and keep it customer-safe. |
| Amber (guardrails) | Prepare for a contract renewal using account history | Augmentation | approved tool, and the account owner verifies facts. |
| Amber (guardrails) | Create FAQs for a new product launch from internal briefs | Augmentation | product approval and respect embargo dates. |
| Amber (guardrails) | Draft an apology and recovery plan after a customer escalation | Augmentation | manager review, and Legal review for serious issues. |
| Amber (guardrails) | Prepare questions for a customer advisory board from feedback themes | Augmentation | redact names and get approval. |
| Amber (guardrails) | Suggest an agenda and talking points for a partner business review | Augmentation | check partner data-sharing terms. |
| Amber (guardrails) | Write a customer-facing FAQ from your internal support docs | Augmentation | support owner approves each answer and removes internal-only detail. |
| Amber (guardrails) | Draft replies in a shared support or sales inbox for a person to approve | Augmentation | nothing sends until a human approves it. |
| Amber (guardrails) | Draft renewal and expansion emails from usage data | Augmentation | data use approved, and facts checked. |
| Amber (guardrails) | Summarize support tickets to spot churn risk | Analysis | remove personal data before processing. |
| Amber (guardrails) | Summarize customer feedback for a business review | Analysis | redact names, and check themes against the source. |
| Amber (guardrails) | Write customer health narratives | Analysis | account owner confirms every claim. |
| Amber (guardrails) | Write customer-facing release communications | Augmentation | product team approval. |
| Red (not safe) | Share a customer's security incident details in a public AI tool | Augmentation | Confidential and potentially regulated. |
| Red (not safe) | Let AI reply to customer complaints with nobody reviewing | Automation | Complaints need human judgment and accountability. |
| Red (not safe) | Translate a confidential customer document with a public tool | Augmentation | The text is shared with a third party. Use an approved process. |
| Red (not safe) | Ask AI to word a message that hides a known product defect or outage | Augmentation | Deceptive and can create liability. |
| Red (not safe) | Rely on AI to decide eligibility for a regulated product or service | Analysis | Regulated decisions need approved processes and accountability. |
| Red (not safe) | Let an AI answer customer questions about pricing or contracts unsupervised | Augmentation | Binding statements need a person with authority. |
| Red (not safe) | Paste patient or health information without a signed agreement | Analysis | Regulated data. Needs a business associate agreement and Legal sign-off. |
| Red (not safe) | Let AI write incident, outage, or layoff communications alone | Augmentation | Sensitive. Needs communications and Legal review. |

### Field sales (60)

| Light | Use case | Type | Guardrail or reason |
|---|---|---|---|
| Green (safe) | Practice planning a day of visits with made-up stops or public business addresses | Augmentation | Nothing private goes in. Compare the order it suggests with a mapping app. |
| Green (safe) | Draft a short script for introducing yourself to a receptionist or gatekeeper | Augmentation | Template wording. Adjust it to your style. |
| Green (safe) | Plan a visit day around lunch, breaks, and buffer time using fictional stops | Augmentation | Practice planning with made-up stops. |
| Green (safe) | Brainstorm ways to open a conversation at a trade show or event | Augmentation | General ideas with no customer data. |
| Green (safe) | Create a checklist for preparing your car and materials before a field day | Augmentation | Personal preparation. |
| Green (safe) | Draft a polite "I was in the area" message template | Augmentation | Template. Only send it to people you already know. |
| Green (safe) | Write a template for confirming a visit the day before | Augmentation | Fill in names and times yourself. |
| Green (safe) | Draft a "sorry I missed you" note template | Augmentation | Template with placeholders. |
| Green (safe) | Practice explaining your product in 30 seconds to someone on a job site | Augmentation | Fictional practice. |
| Green (safe) | Create a pre-visit research checklist for a local business using public information | Augmentation | Public sources only. |
| Green (safe) | Write a template for requesting a facility visit, including safety expectations | Augmentation | Template. Follow the customer's own rules. |
| Green (safe) | Practice handling a walk-in conversation when the buyer is busy | Augmentation | Fictional role-play. |
| Green (safe) | Brainstorm what to do with ten unexpected minutes between visits | Augmentation | Personal productivity. |
| Green (safe) | Draft a weekly territory review template | Augmentation | Structure only, no customer data. |
| Green (safe) | Learn general business etiquette for visiting a new region | Augmentation | Learning aid. Confirm with locals or colleagues. |
| Green (safe) | Create a simple end-of-day routine for logging your visits | Augmentation | Personal process. |
| Green (safe) | Draft a template for a short post-visit recap you send to yourself | Augmentation | Template with placeholders. |
| Green (safe) | Make a cheat sheet of common field objections with approved responses | Augmentation | Use only approved messaging. |
| Green (safe) | Brainstorm professional ways to follow up after a chance meeting | Augmentation | General ideas. |
| Green (safe) | Plan what to pack and prepare for a day of in-person visits | Augmentation | Personal planning. |
| Green (safe) | Draft an outline for a leave-behind after a site visit | Augmentation | Template using approved messaging. |
| Green (safe) | Brainstorm good questions to ask during a facility walk-through | Augmentation | General ideas. |
| Green (safe) | Brainstorm conversation starters for a trade show booth | Augmentation | Generic ideas, no customer data. |
| Amber (guardrails) | Map out the stops on a route and rank the most logical order | Analysis | use AI to group nearby stops and place priority meetings, then confirm drive times, traffic, and opening hours in a mapping app. Share only the customer details your policy allows. |
| Amber (guardrails) | Rank the stops on a route by deal value and urgency | Analysis | you supply the priorities, check them against the CRM, and use an approved tool only. |
| Amber (guardrails) | Cluster a list of customer business addresses into daily visit zones | Augmentation | business addresses only, approved tool, and confirm the result in a mapping app. |
| Amber (guardrails) | Build a weekly territory plan from account priorities and locations | Analysis | approved tool, and a manager reviews fairness and coverage. |
| Amber (guardrails) | Reorder the day's route after a cancellation | Augmentation | confirm drive times in a mapping app and tell affected customers. |
| Amber (guardrails) | Suggest visit windows based on typical business hours for each stop | Augmentation | verify real opening hours and call ahead. |
| Amber (guardrails) | Estimate how much time each visit needs from past visit notes | Augmentation | use as a rough guide and add buffer. |
| Amber (guardrails) | Plan a multi-day trip with hotels and meals | Augmentation | follow your travel and expense policy and verify every booking. |
| Amber (guardrails) | Draft a visit brief for each stop from CRM notes | Augmentation | approved enterprise tool and minimal personal data. |
| Amber (guardrails) | Build a route that balances priority accounts with prospecting stops | Augmentation | manager review, and confirm in a mapping app. |
| Amber (guardrails) | Generate a safety checklist for visiting industrial or construction sites | Augmentation | follow the site's own rules and confirm requirements with the customer. |
| Amber (guardrails) | Draft visit confirmations to several customers at once | Augmentation | review each message, name, and time before sending. |
| Amber (guardrails) | Suggest which customers to visit near a conference or event | Augmentation | check against the CRM and any do-not-contact list. |
| Amber (guardrails) | Summarize which customers in a region have open issues before you visit | Analysis | use only data you are permitted to access, for internal use. |
| Amber (guardrails) | Estimate mileage and cost for a planned route | Augmentation | verify in a mapping or expense tool, and never submit AI figures as is. |
| Amber (guardrails) | Draft follow-up tasks from your visit notes | Augmentation | confirm each task before it goes into the CRM. |
| Amber (guardrails) | Prepare a local-market brief for a new territory from public sources | Augmentation | public sources only, and verify key facts. |
| Amber (guardrails) | Plan visits around local holidays and regional events | Augmentation | confirm dates on an official calendar. |
| Amber (guardrails) | Create bilingual visit scripts or leave-behinds | Augmentation | native-speaker review. |
| Amber (guardrails) | Capture field visit notes by voice and have AI write them up | Automation | approved tool, consent where required, and no sensitive spoken detail. |
| Amber (guardrails) | Convert handwritten visit notes into typed notes | Automation | approved tool and no sensitive personal data. |
| Red (not safe) | Post a customer's exact location or schedule publicly | Augmentation | Privacy and safety risk. |
| Red (not safe) | Use AI to track a customer's or coworker's real-time location without consent | Augmentation | Surveillance without consent is unlawful in many places. |
| Red (not safe) | Paste a list of customers' home addresses into a public AI tool | Augmentation | Personal data shared with a third party. |
| Red (not safe) | Follow AI directions onto a site without checking access and safety rules | Augmentation | Sites have their own entry rules and hazards. Confirm with the customer. |
| Red (not safe) | Let AI decide you can skip a required safety procedure | Augmentation | Safety rules are not optional. |
| Red (not safe) | Type prompts into an AI assistant by hand while driving | Augmentation | Distracted driving. Plan before you drive, or use hands-free approved tools. |
| Red (not safe) | Ask AI to invent a reason for being on a customer's site to get access | Augmentation | Deceptive and may be trespassing. |
| Red (not safe) | Use AI to fake visit logs or mileage claims | Augmentation | Falsifying records is fraud. |
| Red (not safe) | Upload photos taken inside a customer facility to a public AI tool without permission | Augmentation | Confidential, and possibly restricted by site policy. |
| Red (not safe) | Rely on AI to know a site's access rules, permits, or restrictions | Augmentation | Only the customer and site owner can confirm these. |
| Red (not safe) | Use AI-generated directions in remote areas without checking offline maps | Augmentation | Bad directions can strand you. Verify with a reliable map. |
| Red (not safe) | Ask AI to identify people in photos taken at a customer site | Augmentation | Privacy and biometric data risk. |
| Red (not safe) | Use AI to plan gifts or entertainment that break your gift or anti-bribery policy | Augmentation | Compliance and legal risk. Follow your policy. |
| Red (not safe) | Record an in-person conversation with an AI device without consent | Augmentation | Consent laws apply to in-person recording too. |
| Red (not safe) | Use AI to profile a customer's employees' personal routines | Augmentation | Invasive and potentially harassment. |
| Red (not safe) | Trust AI-estimated drive times or distances without checking a map | Analysis | AI does not see live traffic or closures, and a missed appointment costs trust. Instead: confirm the plan in a mapping app. |

### Sales ops (45)

| Light | Use case | Type | Guardrail or reason |
|---|---|---|---|
| Green (safe) | Explain or build a spreadsheet formula using sample data | Augmentation | Use made-up data. |
| Green (safe) | Plan and prioritize your own week | Analysis | Personal productivity. |
| Green (safe) | Summarize a long public article or industry report | Analysis | Public source. |
| Green (safe) | Convert bullet points into an internal recap you will edit | Automation | Only non-confidential internal content. |
| Green (safe) | Learn a new product feature by asking for plain explanations from public docs | Augmentation | Learning aid. |
| Green (safe) | Prepare likely questions for a webinar | Augmentation | Topic-level only. |
| Green (safe) | Create a battlecard format from public competitor information | Augmentation | Public sources only. Cite them. |
| Green (safe) | Suggest questions that expose competitor weaknesses, from public reviews | Augmentation | Public sources. |
| Green (safe) | Write a plain-English description of your sales stages for a new hire | Augmentation | Generic process with no confidential detail. |
| Green (safe) | Draft a weekly pipeline update template with placeholders | Augmentation | Structure only. |
| Green (safe) | Explain a report concept such as conversion rate or win rate in simple terms | Augmentation | Learning aid. |
| Green (safe) | Draft a checklist for cleaning your own contact list | Augmentation | Process template. |
| Green (safe) | Turn a rough process into a clear step-by-step guide | Augmentation | Your own process, with confidential detail removed. |
| Amber (guardrails) | Condense a long email thread before a handoff to a colleague | Analysis | check accuracy and leave out sensitive attachments. |
| Amber (guardrails) | Classify inbound inquiries by topic and urgency | Automation | sample audits, and no sensitive categories acted on automatically. |
| Amber (guardrails) | Propose meeting times across time zones | Augmentation | you confirm in your own calendar. |
| Amber (guardrails) | Adapt internal enablement docs into partner job-aids | Augmentation | check the partner agreement and confidentiality first. |
| Amber (guardrails) | Suggest next best actions for open deals from CRM fields | Analysis | managers check the logic and the rep decides. |
| Amber (guardrails) | Write a competitive win/loss narrative from deal notes | Analysis | de-identify the notes and verify claims. |
| Amber (guardrails) | Create localized versions of marketing copy | Augmentation | native review and Legal review of claims. |
| Amber (guardrails) | Propose territory or account assignments from account data | Augmentation | leaders review fairness and logic before adopting anything. |
| Amber (guardrails) | Write forecast commentary from pipeline data | Analysis | a leader verifies the numbers. |
| Amber (guardrails) | Clean and deduplicate CRM records | Automation | test on a sample and keep a backup. |
| Amber (guardrails) | Categorize and route inbound leads | Automation | audit results for bias and errors. |
| Amber (guardrails) | Analyze win/loss notes | Analysis | de-identify and verify patterns. |
| Amber (guardrails) | Build sales enablement content from internal documents | Augmentation | answer only from approved sources, with citations. |
| Amber (guardrails) | Stand up a product-question assistant on internal docs | Augmentation | respects document permissions and shows sources. |
| Amber (guardrails) | Create shared custom assistants for the team | Augmentation | IT approves, and no confidential files uploaded without sign-off. |
| Amber (guardrails) | Generate marketing images or video | Augmentation | brand and IP review, and disclose where required. |
| Amber (guardrails) | Use synthetic voice or avatars for outreach video | Augmentation | consent, disclosure, and Legal approval. |
| Amber (guardrails) | Summarize third-party competitive reports | Analysis | check the license allows internal summarization. |
| Red (not safe) | Upload a spreadsheet containing national ID or social security numbers | Augmentation | Highly sensitive data never goes into AI tools. |
| Red (not safe) | Paste internal HR documents or performance reviews into a tool | Augmentation | Confidential personal information. |
| Red (not safe) | Use AI to monitor coworkers' personal social media | Augmentation | Privacy violation and a workplace policy issue. |
| Red (not safe) | Leave an AI agent with admin rights running on your CRM unattended | Automation | One bad instruction can change everything. Restrict permissions and supervise. |
| Red (not safe) | Use AI to generate meeting notes for activity that never happened | Automation | Falsifying records is misconduct. |
| Red (not safe) | Connect an AI agent to your inbox with permission to send and delete | Automation | Hidden instructions or mistakes can cause real damage. Start read-only with approval for outside actions. |
| Red (not safe) | Upload a CRM export with customer personal data to a free chatbot | Augmentation | Exposes personal data. Instead: use the approved tool or anonymize first. |
| Red (not safe) | Enter payment card or bank details | Augmentation | Payment data never goes into AI tools. |
| Red (not safe) | Share passwords, API keys, or access tokens | Augmentation | Security incident waiting to happen. |
| Red (not safe) | Submit an AI forecast as your commit without checking the numbers | Analysis | You own the number, not the tool. |
| Red (not safe) | Feed customer data into a tool that trains models without contract rights | Automation | Contract and privacy breach. |
| Red (not safe) | Use personal AI accounts for work data | Augmentation | Shadow IT. The company loses control and visibility. |
| Red (not safe) | Bypass security or data-loss controls to use AI faster | Automation | Policy violation and incident risk. |
| Red (not safe) | Give an AI agent unrestricted write access to the CRM or email | Automation | One bad instruction can change or send everything. Scope permissions tightly. |

### Coaching (17)

| Light | Use case | Type | Guardrail or reason |
|---|---|---|---|
| Green (safe) | Ask for feedback on your own pitch script | Augmentation | Remove customer names. |
| Green (safe) | Make flashcards for product knowledge from public docs | Augmentation | Study aid. |
| Green (safe) | Generate variations of a 30-second pitch | Augmentation | Practice material. |
| Green (safe) | Describe a call you ran in general terms and get coaching tips | Augmentation | No names, no figures. |
| Green (safe) | Draft a personal development plan | Augmentation | Your own career, not company data. |
| Green (safe) | Practice a tough conversation with your manager | Augmentation | Fictional, private. |
| Green (safe) | Ask for a plain-language explanation of a sales methodology | Augmentation | Learning aid. |
| Green (safe) | Create a five-minute daily practice routine for objection handling | Augmentation | Personal skill building. |
| Green (safe) | Ask for an explanation of a sales term you keep hearing | Augmentation | Learning aid. |
| Green (safe) | Generate practice scenarios for handling a difficult customer email | Augmentation | Fictional scenarios. |
| Green (safe) | Draft quarterly goals in clear, measurable wording | Augmentation | Personal planning. |
| Green (safe) | Prepare questions for a mentor or manager one-on-one | Augmentation | Personal preparation. |
| Green (safe) | Ask AI to review your own draft for claims that need a source | Augmentation | Make it list issues instead of rewriting. You decide what to change. |
| Green (safe) | Rewrite your own message in a warmer, more empathetic tone | Augmentation | Your own text with no confidential data. |
| Green (safe) | Create a study guide for a product or certification exam from public material | Augmentation | Learning aid built from public sources. |
| Amber (guardrails) | Convert recorded training sessions into written guides | Automation | participant consent and an accuracy check. |
| Amber (guardrails) | Create onboarding quizzes for new reps from internal playbooks | Augmentation | approved sources, and Enablement verifies the answers. |

## 9. Prompt library

Replace the [brackets]. Use only public or approved content, in an approved tool.

### 1. Account research brief (Prospecting)

```text
Act as a sales researcher. Using only the public text I paste below about [company], write a six-bullet brief: what they do, recent news, likely priorities, and three questions I could ask. Flag anything you are unsure about. [paste public text]
```

**Check before use:** Verify every fact against the source you pasted.

### 2. First-touch email (Prospecting)

```text
Write a 90-word cold email from me to a [job title] at a [industry] company. Reason for reaching out: [public trigger]. One clear question. No hype, no jargon. Give three subject lines.
```

**Check before use:** Remove any claim you cannot back up, and confirm the trigger is public.

### 3. Connection notes (Prospecting)

```text
Write three LinkedIn connection notes under 250 characters, friendly and specific to [shared professional interest]. No flattery.
```

**Check before use:** Make sure it is relevant, not creepy.

### 4. Buyer concerns (Prospecting)

```text
List ten questions a [role] at a [industry] company is probably asking about [problem area]. Group them by urgency.
```

**Check before use:** Treat as hypotheses and test them on real calls.

### 5. Discovery question bank (Discovery)

```text
Give me twelve discovery questions for a [role] about [problem]. Mix situation, problem, impact, and decision-process questions. Avoid yes/no questions.
```

**Check before use:** Choose the five that fit your style.

### 6. Call agenda (Discovery)

```text
Draft a 30-minute call agenda for a first meeting with a [role]. Include a goal, three topics, and a clear close. Keep it to eight lines.
```

**Check before use:** Share it with the buyer ahead of time only if you are happy with it.

### 7. Skeptical buyer role-play (Discovery)

```text
Play a skeptical purchasing manager at a mid-size [industry] company. I will pitch you. Push back with realistic objections about price and timing. After five exchanges, give me feedback.
```

**Check before use:** Use a fictional scenario only.

### 8. Recap from your notes (Discovery)

```text
Turn these notes into a recap email: summary, decisions, and next steps with owners and dates. Mark anything unclear as "unclear". [notes with names replaced by roles]
```

**Check before use:** Confirm every date and owner before sending.

### 9. Executive summary (Proposal)

```text
Using only the facts below, write a 150-word executive summary for a [role]: problem, proposed approach, expected outcome. Do not add numbers or claims that are not in my facts. [facts]
```

**Check before use:** Check that nothing was added.

### 10. Plain-English rewrite (Proposal)

```text
Rewrite this approved product description for a non-technical buyer in three short sentences, with no jargon. [approved text]
```

**Check before use:** Make sure the meaning did not change.

### 11. Objection responses (Proposal)

```text
A buyer says: "[objection]". Give three ways to respond: acknowledge, ask a question, and reframe. Keep each under 40 words.
```

**Check before use:** Pick the one that sounds like you.

### 12. Mutual action plan outline (Proposal)

```text
Create a mutual action plan outline with phases, owners, and example dates for a [deal type] with [number] stakeholders. Use placeholders.
```

**Check before use:** Agree real dates with the customer.

### 13. Trade-off brainstorm (Negotiation)

```text
List types of concessions I could trade other than price, what each is worth to a buyer, and what to ask for in return. Use generic terms. I will add real numbers myself.
```

**Check before use:** Do not paste real prices or margins.

### 14. Follow-up after silence (Negotiation)

```text
Write three short follow-up emails to a buyer who has gone quiet after [stage]. One friendly, one direct, one that adds a useful resource. Each under 60 words.
```

**Check before use:** Do not invent urgency.

### 15. Kickoff email (Post-sale)

```text
Draft a kickoff email to a new customer. Include what happens in week one, who is involved, and what we need from them. Use placeholders for names and dates.
```

**Check before use:** Fill in real details yourself.

### 16. Business review agenda (Post-sale)

```text
Draft an agenda for a 45-minute quarterly business review: goals recap, results, risks, and plan. Keep it to six lines.
```

**Check before use:** Add real results from your own data.

### 17. Renewal check-in (Post-sale)

```text
Write a warm, short renewal check-in email that asks how things are going and what would make the next year better. No pressure language.
```

**Check before use:** Customize with something true and specific.

### 18. Feedback themes (Post-sale)

```text
Group these customer comments into themes, with a one-line summary of each and the number of comments per theme. [comments with names removed]
```

**Check before use:** Remove personal data first and spot-check the grouping.

### 19. Plan a route (Field sales)

```text
I am visiting these businesses tomorrow: [business names and street addresses]. Suggest the most logical order to reduce backtracking, group nearby stops, and place the most important meeting at the best time of day. Tell me what to double-check in a mapping app.
```

**Check before use:** Confirm drive times and opening hours in a map.

### 20. Plan the day (Field sales)

```text
Here are my meetings and tasks for tomorrow: [list]. Build a realistic schedule with buffer time and flag anything that conflicts.
```

**Check before use:** Check it against your calendar.

### 21. Check my work (Coaching & ops)

```text
Review the draft below. List claims that need a source, vague statements, and anything that could be misread. Do not rewrite it. [draft]
```

**Check before use:** You decide what to change.

### 22. Pitch feedback (Coaching & ops)

```text
Here is my 60-second pitch: [pitch]. Tell me what is unclear, what is strongest, and one change to make it more specific. Do not rewrite it for me.
```

**Check before use:** Remove customer names first.

### 23. Learn a topic fast (Coaching & ops)

```text
Explain [topic] to a new sales rep in plain language. Give a two-sentence summary, five key terms, and three questions buyers ask about it.
```

**Check before use:** Confirm key terms with a subject-matter expert.

### 24. Summarize a public article (Coaching & ops)

```text
Summarize this public article in five bullets, then list three questions I could ask a prospect in this industry. [article text]
```

**Check before use:** Open the original for anything you plan to quote.

### 25. Tone adjustment (Coaching & ops)

```text
Rewrite my message below in a [warmer / more direct / more formal] tone and keep it the same length. [your message]
```

**Check before use:** Make sure it still sounds like you.

## 10. If someone makes a mistake

1. **Stop.** Do not paste or send anything else.
2. **Do not delete or hide it.** Leave the evidence so IT can assess it.
3. **Write down the basics:** what was shared, which tool and account, and when.
4. **Tell** your manager, [IT CONTACT], and [LEGAL CONTACT] right away.
5. **Follow instructions,** which may include vendor deletion requests or customer notification.
6. **Learn:** share what you would do differently.

## 11. Laws and rules to be aware of

A plain-language overview, not legal advice. Rules differ by location and change often. Ask Legal what applies to you.

| Area | What to know | Source |
|---|---|---|
| US anti-spam (CAN-SPAM) | Accurate headers and subject lines, a valid physical address, a clear opt-out honored within 10 business days, and penalties of up to $53,088 per violating email (FTC figure) | FTC CAN-SPAM compliance guide: https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business |
| US automated calls and texts (TCPA) | In February 2024 the FCC ruled AI-generated voices are "artificial" voices under the TCPA, so such calls generally need prior express consent | Cooley summary: https://www.cooley.com/news/insight/2024/2024-02-15-fcc-ai-generated-robocalls-illegal-under-the-tcpa |
| EU AI transparency | Article 50 requires people to be told they are interacting with an AI system and requires deepfake disclosure. The source lists the duties as applying from 2 August 2026, so confirm the current status | https://artificialintelligenceact.eu/article/50/ |
| Privacy | GDPR, UK GDPR, CCPA/CPRA, and other laws limit how personal data is collected and used | Ask Legal |
| Call recording | Some places require all-party consent. AI note-takers count as recording | Ask Legal |
| Confidentiality | NDAs and customer agreements may restrict sharing information with AI vendors | Review the agreement |
| Regulated industries | HIPAA, financial rules, PCI, and public-sector rules add requirements | Ask Legal |

## 12. Vendor notes (with sources and check dates)

Vendor terms change often. These notes were last checked on **October 14, 2026**. "Checked at source" means the vendor's page was opened and read. "Excerpt only" means the page blocked automated access and the statement was read from a search excerpt. Confirm everything in your contract.

### Anthropic Claude

- **Check status:** Checked at source on October 14, 2026. Anthropic privacy page dated August 18, 2026.
- **Summary:** Claude is available through Claude apps, an API, and cloud marketplaces. Data handling depends on whether you are on a commercial plan or a consumer plan.
- Anthropic states that by default it does not use inputs or outputs from its commercial products to train its models. [1]
- Feedback you submit, or an explicit opt-in, can be used for training. Anthropic says feedback is kept for up to five years and de-linked from user and customer IDs before use. [1]
- Free, Pro, and Max users choose whether their data is used to improve Claude. Opting in extends retention to five years. Otherwise the existing 30-day period applies, and the change covers new or resumed chats and coding sessions. [2]
- Commercial Terms services, including Claude for Work (Team and Enterprise) and API use through Amazon Bedrock and Google Cloud Vertex AI, are excluded from the consumer changes. [2]
- **Watch out:** Confirm which plan each rep is actually on. A personal account is a consumer plan.
- **Watch out:** Neither page covers data residency. Ask for it in writing if it matters to you.

Sources:

1. Anthropic, [Is my data used for model training? (Anthropic Privacy Center)](https://privacy.claude.com/en/articles/7996868-is-my-data-used-for-model-training). Page dated August 18, 2026. Opened and read October 14, 2026.
2. Anthropic, [Updates to our consumer terms](https://www.anthropic.com/news/updates-to-our-consumer-terms). Announced August 28, 2025. Opened and read October 14, 2026.

Questions to ask: Is our organization on Commercial Terms? What retention and zero-data-retention options does our contract offer? Where is our data processed?

### OpenAI GPT

- **Check status:** Excerpt only on October 14, 2026. Vendor page date not captured.
- **Summary:** OpenAI offers ChatGPT plans, an API, and cloud availability through partners. Business products have different terms from personal accounts.
- OpenAI states that by default it does not use data from ChatGPT Enterprise, ChatGPT Business, ChatGPT Edu, or its API platform, including inputs or outputs, for training or improving its models. [1]
- OpenAI states that your organization's data remains confidential and owned by you, and that API customers can opt in to help improve models. [1]
- **Watch out:** The official page blocked automated access, so confirm the exact wording on openai.com.
- **Watch out:** Personal ChatGPT accounts follow different terms than business plans.

Sources:

1. OpenAI, [Business data privacy, security, and compliance](https://openai.com/business-data/). The page blocked automated access. Wording read from a search excerpt of the page. Excerpt read October 14, 2026.

Questions to ask: Which plan covers our team, and does it include a data processing addendum? What are the retention controls for our workspace? Which compliance reports can you provide?

### Google Gemini

- **Check status:** Checked at source on October 14, 2026. Google privacy hub dated October 6, 2026 (footer shows an update on October 13, 2026).
- **Summary:** Gemini is built into Google Workspace and also available through the Gemini app and Google Cloud.
- Google states that Workspace does not use customer data for training models without the customer's prior permission or instruction. [1]
- Google states that chats and uploaded files in the Gemini app are not reviewed by human reviewers. [1]
- Google lists SOC 1, 2, and 3 and ISO certifications for Gemini, including ISO 42001, FedRAMP High for Gemini in Workspace apps and the Gemini app, and support for HIPAA workloads. [1]
- Gemini Notebook and Gemini in Chrome are listed as currently lacking ISO, SOC, and FedRAMP certifications and as not covered by the HIPAA business associate agreement. [1]
- **Watch out:** Some Gemini features have different certification coverage than the main Workspace apps.
- **Watch out:** Organization data-region settings do not apply to Gemini Notebook data, according to the page.

Sources:

1. Google, [Generative AI in Google Workspace Privacy Hub](https://knowledge.workspace.google.com/admin/gemini/generative-ai-in-google-workspace-privacy-hub). Page dated October 6, 2026. Opened and read October 14, 2026.

Questions to ask: Which Gemini features are enabled for our edition? Which of those are covered by our compliance agreements? Can we restrict features we have not reviewed?

### xAI Grok

- **Check status:** Excerpt only on October 14, 2026. The FAQ excerpt references September 2024 for HIPAA, which may be out of date.
- **Summary:** Grok is available in a consumer app, through X, via an API, and through Business and Enterprise plans.
- xAI's enterprise FAQ states that it does not use enterprise or API data, inputs, or outputs for training its models. [1]
- The FAQ states that the customer retains rights to the inputs they provide and the outputs they receive, and that conversations are deleted after 30 days, subject to exceptions for legal retention or flagged potential violations. [1]
- The FAQ states that xAI signs a data processing addendum with enterprise customers, and an excerpt says Grok for enterprise could not support HIPAA workloads. That excerpt appears dated, so confirm the current position. [1]
- **Watch out:** The page blocked automated access, so every statement here should be confirmed on x.ai.
- **Watch out:** Consumer app terms differ from enterprise terms. Do not use personal accounts for work data.

Sources:

1. xAI, [Frequently Asked Questions (Enterprise)](https://x.ai/legal/faq-enterprise). The page blocked automated access. Wording read from a search excerpt of the page. Excerpt read October 14, 2026.

Questions to ask: Which terms apply to our plan? What is the current position on regulated data such as health data? Who is the contracting entity?

### Amazon Nova

- **Check status:** Checked at source on October 14, 2026. AWS documentation pages carry no date. Lifecycle page covers models launched before September 7, 2026.
- **Summary:** Nova models are used through Amazon Bedrock inside your AWS account. They are typically set up by developers and IT, not by individual reps.
- AWS states that Bedrock gives you explicit control over whether prompts and outputs from inference requests are retained, using modes such as none, default, and aws_review. [1]
- In none mode, AWS says no request or response data is written to durable storage by AWS or shared with the model provider. In default mode the model's policy applies and AWS may retain data for safety and abuse prevention. [1]
- AWS states that model providers do not have access to Bedrock logs or to customer prompts and completions, because models run in AWS-operated deployment accounts. [2]
- AWS lists Nova Canvas, Nova Reel, and Nova Sonic v1 as having reached end-of-life, with end-of-life dates of September 14 and September 30, 2026. [3]
- Models in the Legacy state cannot be adopted by new customers, and existing customers may lose access after 15 days of inactivity. [3]
- **Watch out:** The model lineup changes. Check the lifecycle page before building on a model.
- **Watch out:** Retention depends on the mode and the model, so check the model card for each model you use.

Sources:

1. Amazon Web Services, [Amazon Bedrock: Data retention](https://docs.aws.amazon.com/bedrock/latest/userguide/data-retention.html). AWS documentation. Opened and read October 14, 2026.
2. Amazon Web Services, [Amazon Bedrock: Data protection](https://docs.aws.amazon.com/bedrock/latest/userguide/data-protection.html). AWS documentation. Opened and read October 14, 2026.
3. Amazon Web Services, [Amazon Bedrock: Model lifecycle (Legacy)](https://docs.aws.amazon.com/bedrock/latest/userguide/model-lifecycle-legacy.html). AWS documentation covering models launched before September 7, 2026. Opened and read October 14, 2026.

Questions to ask: Which Nova models are Active for our regions? What retention mode applies to our account? What is the notice period before a model we use is retired?

### Meta Llama

- **Check status:** Checked at source on October 14, 2026. Llama 4 license version effective April 5, 2025.
- **Summary:** Llama models are open-weight and can be run in your own environment or through cloud providers. Meta's consumer assistant is a separate product.
- The Llama 4 Community License grants a non-exclusive, worldwide, non-transferable, royalty-free limited license. It is a custom Meta agreement. [1]
- If the licensee's products exceeded 700 million monthly active users in the preceding calendar month on the release date, the licensee must request a license from Meta, which Meta may grant at its sole discretion. [1]
- Anyone distributing Llama materials or products containing them must prominently display "Built with Llama" and include a notice file. Models created or improved using Llama materials or outputs must have a name beginning with "Llama". [1]
- The license terminates if the licensee sues Meta over IP claims related to Llama, and California law governs disputes. [1]
- On Amazon Bedrock, AWS lists Llama 4 Scout and Maverick as Legacy since September 30, 2026, with an end-of-life date of March 30, 2027. [2]
- **Watch out:** You own the security, hosting, and safety work when you run the model yourself.
- **Watch out:** This is a summary of the license text, not legal advice. Have counsel read the full agreement.

Sources:

1. Meta, [Llama 4 Community License Agreement](https://www.llama.com/llama4/license/). Redirects to dev.meta.ai. Version effective April 5, 2025. Opened and read October 14, 2026.
2. Amazon Web Services, [Amazon Bedrock: Model lifecycle (Legacy)](https://docs.aws.amazon.com/bedrock/latest/userguide/model-lifecycle-legacy.html). AWS documentation covering models launched before September 7, 2026. Opened and read October 14, 2026.

Questions to ask: Which license version applies to the exact model we use? Do our distribution and naming plans meet the attribution terms? Who maintains and patches the deployment?

### Microsoft Copilot

- **Check status:** Checked at source on October 14, 2026. Microsoft page metadata shows an update on October 8, 2026.
- **Summary:** Microsoft Copilot works inside Microsoft 365 apps such as Outlook, Teams, Word, and Excel, and is grounded in content the user can already access.
- Microsoft states that prompts, responses, and data accessed through Microsoft Graph are not used to train foundation LLMs, including those used by Microsoft Copilot. [1]
- Copilot only surfaces organizational data to which individual users have at least view permission, so existing permissions in services such as SharePoint determine what appears. [1]
- Microsoft offers third-party models from Anthropic and OpenAI as subprocessors in Copilot. Admins can decide whether to use them, and models provided by Anthropic are currently excluded from the EU Data Boundary. [1]
- Microsoft lists compliance offerings and certifications including GDPR, ISO 27001, HIPAA, and ISO 42001, and says user interaction data is stored encrypted with admin retention controls. [1]
- **Watch out:** Because Copilot honors existing permissions, overshared files can show up in answers. Clean up access before rollout.
- **Watch out:** Agents and connected tools have their own terms, so review them before enabling.

Sources:

1. Microsoft, [Data, Privacy, and Security for Microsoft Copilot](https://learn.microsoft.com/en-us/copilot/microsoft-365/microsoft-365-copilot-privacy). Page metadata shows an update on October 8, 2026. Opened and read October 14, 2026.

Questions to ask: Which third-party models are enabled in our tenant? How are sensitive files labeled and protected? What retention policy applies to Copilot interactions?

### Perplexity

- **Check status:** Checked at source on October 14, 2026. Perplexity help article dated March 3, 2026.
- **Summary:** Perplexity is an answer engine that searches the web and cites sources. Consumer and enterprise plans have different data terms.
- Perplexity states that Enterprise data is never used for AI training, and that uploaded files on Enterprise Pro and Max plans are kept for 7 days. [1]
- On Free, Pro, and Max plans, AI data retention is on by default. Users can opt out in Preferences, but opting out applies only to data collected afterward. [1]
- If an organization downgrades from an Enterprise plan to a non-Enterprise plan, consumer defaults apply and users must opt out manually. [1]
- **Watch out:** Open the cited sources. A citation does not guarantee the page says what the answer claims.
- **Watch out:** Do not use consumer plans for confidential work data.

Sources:

1. Perplexity, [Data collection at Perplexity (Help Center)](https://intercom.help/perplexity-ai/en/articles/11564572-data-collection-at-perplexity). Article dated March 3, 2026. Opened and read October 14, 2026.

Questions to ask: Which plan covers our team? Can we set custom retention and force deletion? What are the terms for API and search products?

## 13. Roles and responsibilities

| Who | Responsible for |
|---|---|
| Executive sponsor | Sets goals, funds, and removes blockers |
| Legal | Contracts, privacy, consent, and claims rules |
| IT and security | Approved tools, access, and data controls |
| RevOps | Data quality, CRM integration, and measurement |
| Enablement | Training, prompt library, and the quiz |
| Frontline managers | Coaching, review habits, and feedback loops |
| Reps and specialists | Use approved tools, check outputs, and report issues |

## 14. Rollout plan

**Days 1 to 30**

- Talk to Legal and IT
- Choose approved tools
- Write acceptable-use rules
- Train a pilot team

**Days 31 to 60**

- Run three green use cases
- Add one amber use case with controls
- Build a shared prompt library
- Collect wins and misses

**Days 61 to 90**

- Measure results
- Review with Legal and IT
- Expand to more teams
- Decide what to automate next

### Rollout checklist

- [ ] Ask Legal what data, customers, and regions are off limits for AI tools.
- [ ] Ask IT which AI tools are approved and which plans are covered by a contract.
- [ ] Confirm in writing each vendor's training, retention, and subprocessor terms.
- [ ] Write a one-page acceptable-use policy using the safe, guardrails, and not-safe tiers.
- [ ] Confirm call-recording and transcription consent rules for every region you sell in.
- [ ] Start with a pilot team and three safe use cases.
- [ ] Name an owner who reviews outputs and tracks errors during the pilot.
- [ ] Test the same three real tasks across two or more models and compare results.
- [ ] Add verification steps for numbers, claims, and quotes before anything reaches a customer.
- [ ] Train reps on what never goes into a tool.
- [ ] Decide how AI use is disclosed to customers where required.
- [ ] Review results after 30 days with Legal and IT before expanding.

## 15. How to know it is working

**Efficiency**

- Hours of admin saved per rep
- Time to first draft
- Time to respond to leads

**Effectiveness**

- Reply and meeting rates
- Win rate and cycle length
- Forecast accuracy

**Safety and quality**

- Number of policy exceptions
- Error rate found in review
- Rep confidence and quiz scores

## 16. Acknowledgements and license

Adapted from AI Ready Reps, https://aireadyreps.com, by Jared Chapman, and released under the MIT License. Keep this attribution when you adapt it. This document was drafted with the help of an AI assistant (Claude, made by Anthropic) and reviewed by the author. It is not legal advice and is not affiliated with any AI vendor.