Sales & Marketing
AI Sales Desk
Outbound run by AI, approved by your reps, sent from their own mailboxes.
Available as a pilot · Human approval · Complete audit trail
The operational problem
The hard part of outbound is control, not writing
Generating an email is easy. Running outbound without contacting the wrong account, repeating old outreach, inventing a trigger or losing a reply is the operational problem.
AI Sales Desk coordinates the full path from a target record to an approved draft. Research stays cited, existing relationships are checked, the rep approves the exact message and outcomes return to the CRM. The first pilot sends nothing.
Workflow
What the Sales Desk coordinates
Target records
Accounts and contacts start from your CRM or an agreed list, with the fields needed for matching and ownership.
Deduplication and suppression
Existing accounts, prior outreach and exclusion rules are checked before research or drafting continues.
Cited research
The agent uses allowed sources, keeps a source next to each material claim and flags uncertainty.
Rep approval
The sender reviews the complete message as it would leave—not an abstract summary of it.
Mailbox and CRM
Production sending uses the rep’s mailbox under an agreed cap; replies and outcomes return to the account history.
Method
How one prospect reaches an approved draft
- Resolve the account firstThe target is matched against the CRM, ownership, previous contact and suppression rules.
- Research from allowed sourcesThe agent collects only the facts needed for relevance and keeps citations beside them.
- Separate fact from inferenceUnsupported or ambiguous signals are flagged; they are not rewritten as confident claims.
- Draft for one named senderThe message is composed in the rep’s voice and connected to the target record and evidence.
- Approve the exact wordingThe rep approves, edits or rejects the complete message. Approval is recorded per message.
- Send within explicit limitsAfter a successful dry run, approved messages can leave from the rep’s mailbox under daily caps and logging.
Example output
The rep sees the evidence before the message
Illustrative dry-run approval record — no real prospect data
- Company
- Mid-size packaging manufacturer, DE
- Trigger
- New plant announced, 3 sources
- Sender
- Your rep, own mailbox
- Daily cap
- 10 messages
- Lawful basis
- Confirmed
Commercial control
The rules behind responsible outbound
No contact without identity resolution
A likely duplicate or existing relationship is a stop condition, not a reason to create another lead.
Claims keep their sources
A trigger used in the message remains traceable to the page or record that supports it.
The sender approves the message itself
A person sees the final recipient, subject and body before anything can be sent.
Volume is a controlled parameter
Mailbox identity, daily caps, suppression rules and reply handling are part of the system—not afterthoughts.
Fixed start
What the 25-contact dry run proves
- Twenty-five resolved targetsEach target is checked against the available CRM and outreach history before it enters the workflow.
- Research with citationsThe facts used for qualification and personalisation remain inspectable beside each profile.
- Approval-ready draftsReps can review the proposed recipient, evidence and exact wording, then approve, edit or reject.
- CRM and mailbox mapThe pilot documents the fields, permissions, write-backs, sender controls and gaps required for production.
- Zero messages sentThe dry run tests the whole decision flow without contacting a prospect. Production sending is a separate approval.
Fit
When an AI Sales Desk is—and is not—the right move
A good fit
- Your team has a defined target profile and wants deeper research without adding manual administration.
- Account ownership, previous communication or CRM deduplication matters before a first message.
- Reps should keep judgement and sender identity while AI prepares repeatable work around them.
Not the right fit
- The goal is anonymous bulk email with no per-message review or evidence.
- There is no reliable account list, CRM history or owner who can approve messages.
- You want the system to determine legal basis without your policy, counsel or accountable owner.
Where the human approves
Your rep approves the exact wording. The AI works through a fixed list of allowed operations; every approval and result is logged.
Sales and compliance questions
Questions a sales leader should ask
Does the first pilot send emails?
No. The dry run takes twenty-five contacts through targeting, deduplication, research, drafting and approval with zero external messages sent.
Can the AI contact anyone it finds?
No. Targets come from an agreed list or profile, allowed sources and suppression rules. Production sending also requires the exact draft to be approved.
Who decides whether outreach has a lawful basis?
Your organisation remains accountable for its outreach policy and legal basis. The workflow can record and enforce the rule you approve; it does not replace legal judgement.
How do you prevent duplicate or conflicting outreach?
The workflow checks the available CRM, ownership and prior outreach before drafting. Ambiguous matches are held for a person rather than treated as new prospects.
Does every message need approval?
The default design requires the rep to approve the exact wording for every first message. Any later change to that control would need a separate, explicit policy decision.
What happens to replies?
Replies stay in the rep’s mailbox, are classified for follow-up and can update agreed CRM fields. The production scope defines which write-backs are automatic and which need approval.
Experience behind the service
Built from an approval-led outbound workflow
The service is based on an approval-led outbound workflow that connects CRM records, Microsoft 365 mailboxes, allowed research sources, identity controls and operating logs. We show the workflow, safeguards and acceptance scenarios without exposing a customer, target market or conversion data.
Technical stewardship: David Máj, Founder & Technology Consultant. Last reviewed 24 September 2026.