Sales Outreach: Account Research, Ranking, and Reviewable Drafts
💡 One-sentence summary: Use Grok Bot to build your AI sales outreach — research accounts, score and rank them by profile and intent, and draft reviewable email and LinkedIn outreach, with sending always left behind approval.
This is the third lesson on Route A (workplace), using the same official six-step method as the Daily Briefing and Expense Reconciliation. This lesson corresponds to one of the 8 roles on the official use-cases page: Sales outbound (sales outreach specialist). The official positioning for it is: account research, contact ranking, and reviewable outreach — not "writing a few cold emails for you," but a daily account list with explainable rationale, with outreach drafts stopping at the review point.
The whole lesson takes about 12 minutes. The official docs also point to a clear advancement path for this role: once the output is reliable, schedule a nightly research routine that likewise stops at a reviewable list — step 5 of this lesson covers exactly that.
What you'll be able to do after this lesson
- Build a "sales outbound specialist" Bot: connect your CRM, intent sources, and outreach mailbox, and research the accounts in a designated CRM view one by one
- Have the Bot score and rank accounts against your ideal customer profile and recent intent, picking up to three relevant contacts per account
- Draft email and LinkedIn outreach based on the writing-style samples you provide, automatically skipping people already in active sequences
- Schedule a nightly research routine: a reviewable list waiting for you every morning, with sending and sequence enrollment always released by you
Why use a Grok Bot instead of web-based AI?
| Capability | Web-based AI | Grok Bot |
|---|---|---|
| Can read your CRM | ❌ Export reports, copy and paste into the chat box | ✅ Connect directly where a connector exists; for CRMs without one, log in via the cloud computer browser |
| Remembers your profile and writing style | ❌ Re-paste the profile and sample emails in every conversation | ✅ Hand the profile and writing-style samples to the Bot, with context accumulating across conversations |
| Delivers the list on time every day | ❌ The web page won't come find you on a schedule | ✅ The nightly research routine runs in the cloud, even with your laptop closed |
| Scores with explainable rationale | ❌ Says "great fit" based on the model's gut feeling | ✅ The official workflow scores against profile and intent, with output stopping at a reviewable list |
| Sends cold emails for you | ❌ Can only offer text suggestions | ✅ It can send, but drafts come first; first contact is one email per person with one-click unsubscribe |
Your current struggles
- Hundreds of accounts on your list, and who to contact first is pure gut feeling; researching one account means half an hour of digging through websites, news, and LinkedIn
- Cold emails written until your hands ache, with reply rates so low you can't tell whether the list or the copy is the problem
- The most embarrassing part is duplicate outreach: people already in active sequences get "prospected" again as new customers
When to use this technique
When you need to:
- Research a batch of accounts every day, ranked by "how much they resemble your customers and whether they've shown recent activity"
- Compress "browsing a dozen websites to do homework" into "reading one ranked list with rationale"
- Email and LinkedIn outreach drafts waiting in the chat every morning, sent only as you approve them one by one
And you don't want to:
- Manually browse websites, news, and LinkedIn to do homework on every account
- Mass-send templated cold emails, or prospect people who are already in a sequence all over again
🎒 Preparation Before You Start
- Read 3.2 Skills and Teaching by Demonstration — step 4 of this lesson saves the research workflow as a skill, and this is also the common prerequisite for all Route A scenarios
- Have a Bot you can assign work to: if you don't have one yet, first go to 1.5 Create Your First Bot
- Think through three data sources: CRM (connector, or web address + login method), intent source (your existing intent data, such as a product usage sheet), and outreach mailbox (Gmail or Outlook)
- Prepare two materials: a description of your ideal customer profile (ICP) + 2–3 of your best-written outreach emails as writing-style samples
- About 12 minutes, plus one round of human feedback after the first list comes out
Follow Along With Me
Step 1: Write the sales outbound specialist's responsibilities into the Bot's description
Why: The first step of the official six-step method is to write the responsibilities, data sources, output format, and standing boundaries into the Bot's description — it relies on this text every day to understand its job.
How: Create or open a Bot (sidebar + → Create a Bot), and write the description like the following, adapted to your actual situation:
You are my sales outbound specialist, responsible for account research, contact ranking, and reviewable outreach drafts.
Data sources: my CRM, intent data, and outreach mailbox, plus the ideal customer profile and writing-style samples I provide.
Output format: a ranked account list (each entry with scoring rationale) + up to three relevant contacts per account + email and LinkedIn outreach drafts written in the style of the samples.
Standing boundaries: research and draft only; do not send emails or add anyone to a sequence; skip anyone already in an active sequence; state clearly when information cannot be found, and never fabricate it.What you should see: After saving the description, this text appears on the Bot's detail page. From then on, no matter which conversation you assign work in, it operates according to this positioning.
Step 2: Connect CRM, intent sources, and email
Why: The facts for research can only come from the data sources you connect—channels Grok Bot can't read never make it into the ranked list.
How: Handle data sources one by one; for details of the 5-step connection method, see 2.5 Connect apps:
- Outbound email: search Gmail (or Outlook) in Connect apps and add it; when the Bot asks for Authorize / Authenticate, complete the login in the browser
- CRM: if it shows up in Connect apps, connect it directly; if there's no connector, have the Bot log into the web version using the browser on its cloud computer—the login credentials go into Vault, and the Bot logs in on your behalf but can't see the values; if your CRM offers an MCP server, you can also have the Bot add it by URL
- Intent sources: the official guide doesn't name specific tools—just connect your existing intent data (such as product trials or usage forms); public mentions on X can also serve as intent signals, and the Bot can search public content without any connection (read-only, not logged in as you)
- If authorization is interrupted, click Reconnect (on the app) or Reopen (on the connection card in the chat) to get back to the login page
You should see: the email shows Installed; the CRM opens in the Bot's browser and stays logged in. Note that Added ≠ logged in—when it shows Needs auth or Disconnected, you must click Authenticate (Authorize on mobile) to re-authorize for it to count.
💡 This is the unique value of the Grok Bot
Web-based AI can only look at the exported reports you paste to it; the Grok Bot has its own cloud computer and browser, so it can log into the web version of a CRM that has no connector and look things up itself. There's also a free route for intent signals: the Bot can search and read public content on X without connecting an X account—searching posts, reading a given account's posts and mentions, seeing who quoted or reposted, counting recent post volume on a topic, checking trends and news, all read-only.
Step 3: Run your first AI sales outbound research with the official starter prompt
Why: The official starter prompt deliberately stops at a read-only, no-modification review point—validate the list quality first, then talk about automation.
How: Send your ideal customer profile and writing-style samples to the Bot along with your message (drag them into the input box as attachments), then send this Chinese-localized version of the official Sales outbound starter prompt:
Research the 25 accounts in this CRM view. Score them against our ideal customer profile and recent intent,
find up to three relevant contacts for each account, and draft emails and LinkedIn outreach following the attached writing-style samples.
Skip anyone already in an active sequence. Return a reviewable list; do not send, and do not enroll anyone into a sequence.If you only want it to look at a specific app, type @ in the chat to attach the app to this task.
You should see: the Bot reads the data sources one by one and returns three things—a ranked account list with scoring rationale, up to three relevant contacts per account, and outreach drafts written in your style; anyone already in an active sequence is skipped, and no email is sent and no sequence is touched throughout.
Step 4: Refine the output until it's reviewable, then save it as a skill
Why: The official six-step method requires refining the results until they're "reviewable without you watching," then saving the workflow as a skill and testing it with a second input—for outreach, the reviewable standard is: scoring has rationale, contacts are well chosen, and the writing sounds like you.
How: Check the Bot's output item by item and send it back for rework if it falls short (the "scoring criteria feedback" magic prompt below makes this easier); once it's reviewable, lock in the workflow:
Save the complete account-research workflow you just did as a skill named "Nightly Account Research."
Requirements: research the accounts in the specified CRM view, and score and rank them against the ideal customer profile and recent intent;
up to three relevant contacts per account; draft emails and LinkedIn outreach following the writing-style samples;
skip anyone already in an active sequence; return a reviewable list, send nothing, and enroll no one into a sequence.You should see: the Bot confirms the skill is saved. Afterward, ask for it by name in any conversation—just say "run the Nightly Account Research skill" to invoke it (@ is for Bots, groups, routines, and apps; it doesn't summon skills). The official guide requires testing with a second input—run it again with a different CRM view and check whether the scoring and writing style remain stable.
The private skill library is shared across all Bots, so your other Bots can use it too. For saving details, see 3.2 Skills and demonstration.
Step 5: Schedule a nightly research routine—stop at the review list
Why: The official guide says explicitly after the starter prompt: once the output is reliable, schedule a nightly research routine that likewise stops at a reviewable list—all that's left for your morning is reading the list and approving the drafts.
How: Have the Bot create the routine with one plain-language sentence (this is exactly the phrasing of the official example):
Every night at 11:00, run the "Nightly Account Research" skill: research the newly added and updated accounts in this CRM view,
post the ranked list with scoring rationale and the outreach drafts to this chat, and stop at a reviewable list.
If the CRM or an intent source can't be connected, report the failure directly—don't pad the results with stale data.The Bot will first confirm 6 items with you: which Bot is responsible, the schedule and time zone, input sources, expected results, approval boundaries, and what to do when a data source is missing. Once it's created, open the routine and check When to run—it shows the saved schedule in plain language; if it doesn't match what you intended, click Edit to have the Bot change it.
You should see: the routine appears in this Bot's routines list. Don't miss these three official details:
- A new routine waits for the next scheduled time to run: a "daily at 23:00" routine created at 9 a.m. first runs that night at 23:00—to verify immediately, click Test
- The time zone follows Settings → Bot → Timezone (Auto-detect follows the local machine); if the 23 o'clock looks wrong, check here first
- Routines run in the cloud, so closing your laptop doesn't affect them
For more scheduling details, see 3.3 Routines.
Step 6: Hold the approval line—drafts are drafted automatically, sending never is
Why: The final step of the official six-step method: consequential external actions always stay behind approval. The official starter prompt's exact words are do not send or enroll anyone—neither sending nor sequence enrollment may happen automatically.
How: You don't need to do anything—just keep the defaults: Auto Review is on by default, outbound emails come out as a draft card first by default (Needs approval), emails sent by routines are always saved as drafts awaiting your click of Send, and the official guide states plainly that there is no unattended-sending option. First-contact emails also have one hard rule: one email per person, with one-click unsubscribe; if someone unsubscribes or marks an email as spam, the Bot stops sending to that address and tells you who was skipped.
What you should see: every morning you receive a ranked list and a set of outreach drafts, which only go out after you approve them one by one.
Two easy-to-miss details: the nighttime routine stops at the list and drafts, so it doesn't actually need your sign-off in the middle of the night—and don't add actions that require approval to it later, since approval cards triggered by a routine expire after 30 minutes if unhandled (ones initiated in chat wait indefinitely); as for how many new addresses you can email, there is an official protective limit (to prevent addresses from being marked as spam), and the exact number is not published—just treat the list as a batch plan. For approval rule details, see 3.1 Approvals and Auto Review.
📋 Magic Prompt
📝 Account Research Checklist (official starter prompt structure version)
Expected result: the Bot returns a reviewable checklist—accounts scored and ranked against the profile and intent, each entry with supporting evidence, at most three relevant contacts per account, plus outreach drafts written in your style.
## Role
You are my sales outreach specialist, responsible for account research and reviewable outreach drafts.
## Task
Research the accounts in the specified CRM view, score and rank them against the ideal customer profile and recent intent, find at most three relevant contacts for each account, and draft email and LinkedIn outreach following the style samples.
## Input Information
### Required Fields
- CRM view: [the account list or view name to research]
- Ideal customer profile: [hard criteria such as industry, size, roles, pain points]
- Style samples: [2–3 of your best outreach emails]
### Optional Fields
- Intent data: [recent signals such as trials, usage, website activity?]
- Contact rules: [how many per account at most, which seniority levels?]
## Output Specification
A reviewable checklist: accounts ranked by score, each entry with scoring rationale and relevant contacts; outreach drafts written following the style samples, one draft per person.
## Constraints
- ✅ Skip people already in active sequences; scoring rationale must be traceable back to the data source
- ✅ Say explicitly when information can't be found; don't fabricate
- ❌ Do not send any emails or messages
- ❌ Do not add anyone to a sequence📝 Outreach Drafts (email + LinkedIn version)
Expected result: for the contacts you select from the checklist, one draft set per person: an email version + a LinkedIn version, both sounding like you wrote them yourself.
## Role
You are my sales outreach specialist; you just finished the account research and are now drafting outreach for me.
## Task
Draft one email and one LinkedIn outreach for each contact I've selected, personalized to this person and this company.
## Input Information
### Required Fields
- Target contacts: [contacts selected from the reviewable checklist]
- Style samples: [your sample texts; tone and length follow them]
- Angle: [why this account is worth contacting right now]
### Optional Fields
- Email subject lines: [provide a few options?]
- LinkedIn word limit: [e.g., within 300 words?]
## Output Specification
One set per contact: an email draft (subject + body) + a LinkedIn draft; open with the angle, no canned filler.
## Constraints
- ✅ One draft per person, personalized to the account and role
- ✅ Tone close to the style samples
- ❌ Don't send anything; save all as drafts and wait for my review
- ❌ Don't promise discounts or demos I haven't authorized🔧 Scoring Criteria Feedback (tuning profile judgment)
Expected result: the Bot updates the scoring criteria based on your feedback and writes them into the skill; the next day's list is noticeably more accurate—the ones that should rank high do, and the ones not worth contacting stop appearing.
## Role
You are my sales outreach specialist; you just sent me today's ranked account list.
## Task
Based on my feedback, update the scoring criteria for "what counts as high intent and what counts as a poor fit," and write them into your nighttime account research skill.
## Input Information
### Required Fields
- Misjudged accounts: [list the ones ranked too high and the ones missed, with the correct judgment]
- Judgment basis: [the corresponding profile criteria or intent signals]
### Optional Fields
- New signals: [intent sources not counted this week but to count going forward?]
## Output Specification
First restate the new criteria you've summarized; update the skill only after I confirm.
## Constraints
- ✅ Write the criteria into the skill, taking effect from tomorrow night
- ✅ Only adjust the scoring criteria; don't change the output format or approval boundaries
- ❌ Don't rank all accounts high just to please me
- ❌ Do not send any messagesCheckpoints ✅
- Email shows Connect apps as Installed (not Needs auth / Disconnected), and the CRM can be connected (via connector or browser login)
- Every entry in the first list has scoring evidence, and people already in active sequences are skipped
- No emails were sent and no one was added to a sequence throughout
- The research workflow is saved as a skill and can be rerun by name with a different CRM view
- The nighttime routine's When to run shows the schedule you want, and outreach drafts only send when you click Send
Pitfall Reminders
❌ Scores have no evidence, just "this company is a great fit"
This means the profile wasn't written concretely. Scoring rationale must be traceable back to the data source—which profile criterion was hit, which intent signal appeared. Use the "Scoring Criteria Feedback" magic prompt to bounce it back a few rounds, and confirm that "scoring rationale must be traceable back to the data source" has been written into the skill.
❌ People already in sequences get prospected again
Two possibilities: ① the Bot can't read the sequence status in the CRM—give it access to the view containing the sequence field, or state in the instructions "check sequence status first, then research"; ② you later manually added people to a sequence—have the Bot re-pull the account list before each research run instead of reusing the old list from last time.
❌ Drafts reek of AI and don't sound like you
Your style samples are too few or too generic. Provide 2–3 emails you actually sent and got replies to as samples, and hard-code into the instructions "length, tone, and opening style all follow the samples." If it still doesn't sound like you, just send it back: "This reads like a template; rewrite it following the samples."
❌ Wanting the routine to send the emails directly—not possible
The official word: emails sent by a routine are always saved as drafts waiting for your Send; there is no unattended sending option. This isn't a flaw—it's the official design for outreach scenarios: the starter prompt stops at the review point, and sending always stays behind approval. To save effort, make "approving drafts" a fixed part of your morning routine.
❌ Sent too many cold emails, and new addresses suddenly can't receive
There's an official protective limit on "how many new addresses you can email," to prevent addresses from being marked as spam; the exact number is not published. Don't aim for a mass blast overnight—the list is a batch plan by design; for first contact, one email per person with one-click unsubscribe, and if someone unsubscribes or marks it as spam, the Bot automatically stops emailing that address and tells you who was skipped.
Advanced Tips (optional)
- Public X mentions as intent signals: the Bot can search and read public content on X without any connection (read-only, not logged in as you)—having it count recent post volume on a topic or read mentions of a certain account are ready-made intent signals; private content like DMs, bookmarks, and the timeline requires the X plugin. See 2.5 Connect apps.
- Archive ranked lists as Pages: have the Bot also organize each day's ranked list into a Pages page (officially positioned for readable pages like reports and dashboards), stored in the Bot's Library—a month later, looking back at which accounts keep showing up is faster than digging through chat history. See 2.2 Files and reviewable results.
- Swap the data sources and it's a new scenario: with the same six steps, swap CRM, intent sources, and email for CRM, product usage, support tickets, and billing, and you get a ranked list of risk and expansion signals for A5 Account Health; to send content digests from the Bot's own mailbox, see C3 Email Digest.
Lesson Summary
- ✅ One "sales outreach specialist" Bot + three types of data sources (CRM, intent sources, outreach mailbox, plus profile and style samples) produces a reviewable AI sales outreach list every day
- ✅ Start with the official starter prompt: research accounts, score and rank, pick contacts, draft outreach, and stop at the review point—no sending, no adding to sequences
- ✅ Trace scoring rationale back to the data source, iterate until it's reviewable, then save as a skill, and test with a different CRM view
- ✅ Schedule it as a nightly research routine: cloud execution shows up on time; sending and sequencing always stay behind drafts and approvals
| Key item | Value for this lesson |
|---|---|
| Bot role | Sales outbound specialist — account research, contact ranking, reviewable outreach |
| Connected apps | CRM (via connector or browser login), intent sources (e.g., a product usage sheet), Gmail/Outlook |
| Ranking criteria | Your ideal customer profile (ICP) + recent intent signals |
| skill | "Nightly Account Research" (name it yourself, invoke it by name, shared across all Bots) |
| routine | Every night at 11:00 (example), stopping at a reviewable list; the timezone is under Settings → Bot → Timezone |
| Approval line | Outreach goes out as a draft card first; routine emails are always saved as drafts waiting for you to Send; first contact is one email per person with one-click unsubscribe |
Next lesson preview
Next lesson A4 Talent scouting: still the same six steps, swapping the information sources to your ATS, sourcing tools, email, and calendar, letting the Bot source candidates, research backgrounds, and draft outreach for you. You will learn:
- Have the Bot find candidates who meet the hard requirements in the job description, excluding people already in the ATS
- Every match comes with evidence-based explanations, and personalized outreach drafted in your tone
- Official special requirement: outreach stays behind approvals, and the Bot respects candidate privacy and each source's terms