An AI sales assistant agency playbook is a repeatable way to turn each client brief into researched accounts, reviewed outreach, handled replies, and a CRM handoff. The agency keeps client instructions, source information, approval decisions, and suppression rules visible at every step.
What should an agency set before research begins?
Create a client brief with the target market, offer, approved claims, geography, owner, data sources, and exclusions. Record who can approve outreach and who handles legal or sensitive replies. Keep each client’s records and instructions separate; check the actual access controls of any tool before adding client data.
For international campaigns, record the relevant market and language, any reseller or partner attribution, and the permitted use of event or company lists. Agree on event routing, follow-up states, and separate acquisition, sales, and partner reporting before work starts.
How does the work move from signals-to-pipeline?
Use the same review path for each client:
Signals -> Leads -> Enrichment -> Decision -> Outreach -> Pipeline| Stage | Agency action | Output for review |
|---|---|---|
| Signals | Collect client-approved CRM notes, public research, and any licensed shipment or customs records. | Source list with permitted use and gaps. |
| Leads | Match company name, domain, location, and existing CRM record; flag possible duplicates. | Account candidate with conflicts marked. |
| Enrichment | Add relevant roles and contact candidates; distinguish sourced facts from AI inferences. | Field notes and missing-data list. |
| Decision | Decide whether to research, prioritize, hold, or exclude the account. | Reason and named approver. |
| Outreach | Prepare message drafts using approved facts and check suppression and consent metadata. | Copy ready for human approval. |
| Pipeline | Classify replies, assign an owner, and prepare a CRM update. | Handoff with information, reply context, and next action. |
Shipment activity may show that a company traded a product or used a route. It does not, by itself, prove current buying intent or identify the right contact. Check identity and permitted data use before acting on it. The Trade Data page explains this research layer.
Where should people approve the work?
Use approval gates at the client brief, target account, first message, sensitive reply, and CRM handoff. A reviewer should see the source, the proposed action, and what remains uncertain. AI output is candidate material; the client’s authoritative CRM records remain the reference for ownership and history.
| Review point | Question to answer | Hold the work when |
|---|---|---|
| Account selection | Is this the right company and client? | Leads, ownership, or source rights are unclear. |
| Outreach | Is each claim supported and is contact permitted? | Consent, opt-out, suppression, or restricted-use status is unresolved. |
| Reply and handoff | Who responds and what enters CRM? | A reply asks for legal, privacy, pricing, or other sensitive guidance. |
How can an agency run a first client pilot?
- Choose one client and one narrow segment. Name the delivery owner and approval owner.
- Review a small sample of account information and resolve duplicate or uncertain identities.
- Add only contact data the client may use; record source and suppression status.
- Prepare a prioritized queue and message drafts. Have a human approve the first touch.
- Route replies to an owner. Process opt-outs and exclusions before any follow-up.
- Check that CRM notes preserve the source, message, reply, owner, and next action.
- Review research quality, operator effort, reply handling, and client reporting before expanding.
If a source is missing or a contact cannot be verified, hold the record for research. If a client cannot approve a claim, remove it from the draft. Do not treat missing information as permission to guess.
Where does Dureach fit?
Dureach positions its work as an operator-reviewed AI workforce: a business brief is assigned to named specialists with defined deliverables and approval checkpoints. An agency can use that model to frame research, outreach preparation, and CRM handoff jobs. Confirm the controls and integrations needed for each client’s delivery plan during setup. See the Marketing Agency solution and AI sales agent guide for related context.
FAQ
What is an AI sales assistant agency playbook?
It is a repeatable set of inputs, review gates, outputs, and owners for using AI in agency sales delivery across clients.
Can AI approve outreach for a client?
The client and agency should assign a human approver. AI can prepare drafts and research summaries for that decision.
What should happen when two records may represent the same company?
Hold outreach, compare company identifiers and CRM history, and resolve the match before enrichment or handoff.
Can shipment data prove a company is ready to buy?
No. It can support an account hypothesis, but the agency still needs to check identity, timing, relevance, and permitted use.
How should an agency handle an opt-out?
Record and apply the suppression before further outreach, then check any related sequences or client lists that could reuse the contact.
What belongs in a CRM handoff?
Include the account rationale and sources, approved message, reply context, owner, open risks, and next action.