LearnGrok
Prompts
PromptIntermediateGrok Bot basics

Contact map for deal coverage

Build a buying-group contact map from CRM records and account notes. For account executives and sales managers.

5 min read

Use these prompts to turn a contact list into a buying-group view of one complex opportunity. They suit account executives building deal coverage and sales managers testing whether a forecast rests on real customer access.

A contact map is only useful when it distinguishes what you know from what you assume. Run the prompts in order for a new opportunity. For an established deal, start with the prompt that matches the weak point, such as duplicate CRM records or an unfilled economic-buyer role.

Key point

Keep evidence visible

A named stakeholder without a source, date or documented interaction is a lead to validate, not confirmed deal coverage.

1. Gather the account material first

Create one working folder or document for the opportunity. Copy the source material into the placeholders without changing the prompt instructions. Include the information that can show both contact identity and deal relevance:

  • CRM contact records, including record IDs, title, employer, owner and recent activity.
  • Opportunity fields, particularly the stated customer problem, current status, next step and expected decision process.
  • Meeting notes with dates and attendees.
  • Email summaries or call notes that record a stakeholder's position, concern or commitment.
  • Public account information that your organisation has approved for this work.

Do not paste a general account summary and expect a reliable stakeholder map. The model needs the names, the evidence and the date of each interaction. If your CRM shows only activity counts, add the relevant notes. An activity count cannot show whether the contact challenged the proposal, introduced a colleague or merely accepted a calendar invitation.

Watch out

Do not fill gaps with job titles

A senior title may suggest where to investigate. It does not prove decision authority, support or internal influence.

2. Extract before you interpret

Run Extract contacts and evidence first. Its job is not to decide who matters. It creates an auditable register of every person found in the material and flags records that may refer to the same person.

This matters when a CRM contains an old title, while a meeting note uses a newer one. It also matters when two people share a surname or when an external contact moved between business units. Keep former employees in the evidence register. Their history may explain why a route has gone quiet, but they are not current buying-group coverage.

Then use Resolve duplicate contact records. Review its Do not merge section before changing CRM data. A mistaken merge can remove a valid stakeholder, assign activity to the wrong person and give your manager a false picture of account access.

Check

Check the working contact list

Every current contact should have one canonical row, a source, and a confidence level. Uncertain matches should remain separate.

3. Map roles from deal evidence

Run Assign buying roles once the contact list is stable enough to use. Give it the customer problem and proposed scope, not only names and titles. A person can be influential in one initiative and irrelevant to another.

Read the Role status column closely:

If the role is Treat it as What to do next
Confirmed Supported by explicit notes or customer statements Maintain access and test whether the role remains current
Probable A reasonable reading of incomplete evidence Ask a targeted question to confirm it
Unfilled No named person is supported by evidence Find the route to identify that person

The critical output is often the blank space. A missing procurement contact may not matter early in every opportunity. It does matter if the notes show a commercial review is approaching. Likewise, an unfilled business owner is a more serious risk than a missing technical evaluator when the customer has already validated the technical approach.

4. Test relationship and influence claims

Run Map relationships and influence after roles are assigned. This prompt makes you separate a documented interaction from an interpretation of that interaction.

For example, a meeting with an executive is a fact. Calling that executive supportive is an assumption unless the notes show support. A colleague saying that someone "has influence" is useful evidence, but it still needs a validation action. Ask who else they involve, what decision they own, or whether they would introduce the relevant owner.

Use the relationship table in manager reviews. It shows whether deal progress depends on one contact, whether the account owner has recent access, and whether a supposed champion is actually active.

5. Select one next move

Run Choose the next contact last. It turns the map into an action, without pretending that every uncertain role is a reason to send another email.

The recommended person should close the most important gap in the current opportunity. That might mean validating who owns the business problem, testing an assumed decision route, or creating a route into an unrepresented review function. It is not automatically the most senior name.

Check

Check the final recommendation

You should be able to point to the evidence for the chosen contact, name the unknown you will test, and explain why the alternative route is genuinely different.

6. Know when the output is wrong

Treat these outputs as wrong or incomplete when they show any of the following:

  • A contact is described as supportive, influential or a champion without a quoted note, dated activity or stated source.
  • A role is marked confirmed because of title alone.
  • A former employee appears as an active decision-maker.
  • The same person appears twice with no explanation of the difference.
  • The next contact recommendation does not link to a specific coverage gap.
  • The result names a decision process, budget owner or reporting line that was not in your source material.

Correct the source record or add the missing meeting evidence, then rerun only the affected prompt. Keep the prior output until you can see what changed.

If a prompt gives shallow or conflicting results, reduce the input to one account and one opportunity. Add dated notes and the exact CRM fields rather than more general background. If you need to confirm what inputs or capabilities are available in the environment you are using, check the xAI documentation.

Copy-ready prompts

5 prompts. Open one to read it, or take the whole pack.

1Extract contacts and evidenceStart here when your CRM export and meeting notes are scattered, duplicated or incomplete.
Create a contact evidence register for the account named [account name]. Use only the material below: do not add facts from memory or make up job titles, reporting lines or relationships.

CRM records:
[paste CRM contact records, opportunity fields and activity history]

Meeting notes:
[paste meeting notes]

Public account information:
[paste approved public information, including source URL or source name where available]

Return one Markdown table with these columns: Person name | Current title | Organisation or business unit | Contact details present | Source document | Date of source | Exact supporting evidence | Possible duplicate or identity issue | Confidence (high, medium, low).

Use a separate row for each possible person record, including suspected duplicates. Preserve names exactly as written in the source. Where a field is missing, write `Not provided`. Where sources conflict, record both values in the evidence column and set confidence to low. Do not resolve conflicts yet.

After the table, add:
1. `Records needing resolution`: a numbered list of duplicates, ambiguous names, former employees, and contacts with unclear employer or title.
2. `Source gaps`: a bullet list of documents or fields that would materially improve the contact map.
3. `Known contacts count`: the number of distinct people, followed by the number of unresolved records.
2Resolve duplicate contact recordsUse this after extraction when the same person appears under different titles, email addresses or CRM entries.
Reconcile the contact records for [account name] into a working contact list. Treat the source material as evidence, not as proof. Do not merge two records unless the evidence supports it.

Contact evidence register:
[paste the output from the contact evidence register]

Additional CRM records or corrections:
[paste any additional records]

Return:

1. A Markdown table with these columns: Canonical name | Alternate names found | Current title | Organisation or business unit | Status (current, former, unclear) | CRM record IDs | Evidence for identity match | Confidence.

2. A section called `Do not merge` listing each pair of records that could be different people. For each pair, state the exact missing fact needed to decide.

3. A section called `CRM clean-up list` with numbered actions. Each action must name the record ID, field to change, proposed value, evidence source and whether manager approval is needed.

Rules:
- If title, employer or employment status conflicts, keep the conflict visible.
- If a person has left the account, retain them in the table with status `former`; do not treat them as an active route into the deal.
- Do not guess email addresses, phone numbers, reporting lines or identities.
- Use `Unknown` where the available material cannot support a conclusion.
3Assign buying rolesUse this when you need a first view of the buying group, rather than a flat list of contacts.
Build a buying-role map for the opportunity named [opportunity name] at [account name]. Use only the contact list and deal evidence below. Identify roles from evidence, not from seniority alone.

Working contact list:
[paste reconciled contact list]

Opportunity details:
- Problem or initiative: [paste]
- Proposed scope: [paste]
- Known decision process: [paste]
- Commercial timeline: [paste]

Meeting notes and CRM activity:
[paste relevant notes and activities]

Return a Markdown table with these columns: Buying role | Named person | Title and business unit | Evidence for role | Role status (confirmed, probable, unfilled) | Deal stance (supportive, neutral, resistant, unknown) | Confidence | What would confirm this role.

Use these buying roles where evidence supports them: economic buyer, executive sponsor, business owner, technical evaluator, operational user, procurement, legal or risk reviewer, finance approver, champion, blocker, implementation owner. You may add another role only if it is clearly evidenced.

A person may hold more than one role. If no named person is supported, enter `Unfilled` in the named-person field. Do not force every role into the map.

After the table, write `Role coverage summary` with three bullets: confirmed roles, probable roles, and critical unfilled roles. Then write `Decision-process unknowns` as a numbered list of questions that the current evidence does not answer.
4Map relationships and influenceUse this when you know the names but do not know who can move the opportunity forward.
Assess relationship coverage and influence for [opportunity name] at [account name]. Separate documented facts from working assumptions.

Buying-role map:
[paste buying-role map]

CRM activity history:
[paste calls, emails, meetings and tasks with dates]

Meeting notes:
[paste notes that mention relationships, internal support, concerns or decision-making]

Return two Markdown tables.

Table 1, `Relationship coverage`, must use these columns: Person | Buying role | Last meaningful interaction date | Interaction owner | Relationship strength (strong, active, weak, none, unknown) | Evidence | Next relationship need.

Table 2, `Influence assumptions`, must use these columns: Person | Likely influence area | Influence direction (positive, negative, mixed, unknown) | Evidence | Assumption or fact | Confidence | Validation action.

Classify a statement as a fact only when a source explicitly supports it. Label all other interpretations as assumptions. Do not infer influence from job title alone. Do not say someone is a champion unless the evidence shows they are actively helping the opportunity.

Finish with `Coverage risks`: a numbered list ordered by deal risk. Each item must name the missing relationship or untested assumption, explain why it matters to the stated decision process, and state the evidence gap.
5Choose the next contactUse this before account planning, a manager review or the next sequence of outreach.
Create a decision-ready contact map and next-contact plan for [opportunity name] at [account name]. Use the material below. Do not present assumptions as confirmed facts.

Buying-role map:
[paste buying-role map]

Relationship coverage and influence assumptions:
[paste relationship and influence output]

Current opportunity position:
- Stage or internal status: [paste]
- Customer problem: [paste]
- Agreed next step, if any: [paste]
- Date of next customer interaction, if scheduled: [paste]

Return these sections in this exact order:

1. `Contact map`: a Markdown table with columns: Buying role | Named person | Title | Status of role | Relationship strength | Influence assessment | Evidence or assumption | Coverage gap.

2. `Priority gaps`: no more than five numbered items. Each item must name the missing role, relationship or validation needed, plus the consequence of leaving it unresolved.

3. `Next contact to approach`: name one person. If no named person is appropriate, write `No safe contact identified` and name the role to find instead. Include: why this person is the priority, the evidence supporting that choice, the specific unknown to test, the proposed channel, and one concise outreach objective.

4. `Alternative route`: name one different person or role to approach if the first route fails. State why this is a distinct route, not merely another contact with the same function.

5. `Manager review`: list the two most important assumptions that need challenge before the forecast or account plan is updated.

Rules:
- Do not recommend contacting a person solely because they are senior.
- If relationship data is absent, say so and select a validation action rather than inventing strength.
- Keep the final recommendation focused on the next move, not a generic account strategy.

Last checked against xAI’s own pages on 2026-08-21. Grok changes quickly; anything version-specific should be confirmed upstream before you rely on it.

More in Grok Bot basics

Found something out of date?

Grok changes quickly and this page is a snapshot. If something here is wrong, or you know a better resource, send it over.

Suggest a link →

Advertise on LearnGrok

$420.69one-time, for a 30-day run

Square works best. PNG, JPEG or WebP, up to 2 MB.

Stripe on the next step. Live once approved.