What a B2B AI BDR needs to know about your offer
Estimated reading time: 7 minutes
Before a B2B AI BDR researches its first account, give it an approved explanation of what you sell, who can buy it, and what it must not claim.
A website can supply the first draft. Someone who knows the business still needs to check the delivery limits, customer proof, and reasons to exclude an account. A confident description of the wrong offer gives the campaign the wrong starting point.
We would treat automated onboarding as preparation for that review. Finishing the setup screen does not approve the targeting or the messages. The useful output is a brief that lets a reviewer explain why an account fits and trace each sales claim to an approved source.
Separate your offer from the account you are researching
Keep two sets of information distinct:
| Seller knowledge | Account research |
|---|---|
| What your business supplies | What the target company does |
| The conditions under which you can supply it | Evidence that those conditions might apply |
| Claims and customer proof you have approved | Sources supporting observations about that account |
| What your team must confirm in a conversation | Questions that public research has not answered |
Suppose you sell permanent recruitment services. An employer's vacancy can help identify a relevant account. It does not expand your service into temporary staffing, prove the employer uses agencies, or establish a budget.
The seller brief defines the offer. Account research tests whether a particular business might fit it. Keep an unanswered question visible instead of filling it with a plausible answer.
Current product documentation illustrates the distinction. HubSpot's prospecting-agent setup separates audience selection from selling context, including what the company sells and which problems it addresses. Microsoft's Sales Qualification Agent setup separately configures company and product information, selection criteria, email instructions, and handoff criteria. These are examples of configurable inputs, not evidence that a completed setup produces qualified opportunities.
Build an AI SDR onboarding brief your team can approve
Start with one offer and one target segment. If different services need different qualification rules, give them separate sections or briefs. A broad statement such as “we help companies grow” gives the reviewer little to check.
Use the following fields. Keep the main brief short enough for someone to review; attach the source material behind it.
| Field | What to record | What the reviewer should challenge |
|---|---|---|
| Offer | What you deliver and what the customer receives | Does the description add a service you do not provide? |
| Business task | The specific work or decision the offer helps with | Is the problem relevant to this offer, or merely common in the industry? |
| Account fit | Relevant operations, locations, delivery conditions, and exclusions | Can you explain why an account qualifies beyond its industry label? |
| Buying roles | The responsibilities you want to identify | Has a job title been mistaken for purchasing authority? |
| Proof | Approved examples, supporting documents, and permission to use them | Does the proof support this exact claim for this audience? |
| Commercial facts | What you may say about pricing, scope, commitments, and the next step | Is a proposal-dependent fact being presented as a fixed promise? |
| Unknowns | Facts that need a conversation or another source | Will the agent hold the question open or invent an answer? |
| Ownership | Who approves the brief, reviews exceptions, and handles buyer replies | Is there a person responsible when the source or offer changes? |
Include negative criteria alongside positive criteria. If the service requires a particular delivery region, an account outside it should not qualify because it matches several attractive characteristics.
Distinguish a hard exclusion from missing information. An unsupported location is a reason to investigate or hold. A verified location outside your service area is a reason to exclude. Decide which action applies before you ask the system to produce a list.
Give each usable claim a source and an owner
Treat an AI sales knowledge base as source material for review. A brochure, old proposal, product note, or case study still needs a decision about what the campaign may use.
Keep a small claims ledger beside the brief:
| Record | What to store |
|---|---|
| Approved wording | The claim the message may make, with its conditions |
| Source | Document name or URL, relevant section, and version or date |
| Permission | Whether the material may be used publicly, shared with this audience, or kept internal |
| Limit | What the source does not establish, including excluded products or customer groups |
| Owner | The person who can resolve a conflict or approve a revision |
| Review trigger | A product change, expired offer, withdrawn permission, or other reason to check it again |
For example, a customer result from one service should not silently become a promise for another service. A proposal written for a particular client should not establish a public starting price. An old feature description needs checking if the current product changed.
When two sources conflict, ask the owner to resolve the conflict. Do not let the agent choose whichever sentence makes the stronger pitch. Until the conflict is resolved, hold the disputed claim out of the campaign.
An illustrative B2B offer brief
The agency, offer, and message below are fictional. They demonstrate the brief's decisions; they are not Quantonica or client results.
Assume Northline Talent recruits permanent industrial-maintenance technicians for manufacturing employers in its North and Central delivery regions. Each employer approves the role and makes the hiring decision. The agency does not provide temporary staffing or executive search.
| Brief entry | Approved instruction |
|---|---|
| What we sell | Permanent technician recruitment for manufacturing employers within the stated regions |
| Useful account evidence | A current, relevant vacancy with a work location inside the delivery scope |
| Person to research | The person responsible for hiring for that team; verify responsibility rather than relying on a title alone |
| Exclusions | Temporary staffing requests, executive roles, and locations outside the delivery scope |
| Commercial boundary | Pricing follows an agreed role brief; do not invent a fee or discount |
| Proof boundary | No customer result or placement-time guarantee is approved for this example |
| Unknowns | Whether the employer uses agencies, the budget, and willingness to discuss outside support |
| Next step | Ask whether an agency discussion is relevant; Northline’s recruiter handles the reply |
Now suppose a reviewer has checked a current vacancy and its work location. A first-message draft could use those two facts:
Your careers page lists an industrial-maintenance technician role in [approved region]. We recruit permanent technicians for manufacturing employers. Are you considering outside recruiting support for this role?
The first sentence needs the checked vacancy source. The second stays within the agency's approved offer. The final question leaves agency use unresolved.
The same brief should reject “we can fill this position within two weeks.” Nothing in the fictional offer establishes that guarantee. It should also reject outreach about a temporary role, even when the employer is otherwise a strong fit.
If the location cannot be confirmed, the next action is to check it. Writing a more polished message does not resolve the missing qualification fact.
Test the brief before approving a live campaign
Ask the system to apply the brief to difficult cases as well as obvious matches. Use cases drawn from your own offer, and record the expected decision before reviewing the output.
- Give it an attractive account that fails one hard exclusion. Check that it excludes the account.
- Ask for a price that the approved material does not contain. Check that it leaves the price unknown.
- Ask it to turn a limited customer example into a general guarantee. Check that it keeps the original scope.
- Supply conflicting product descriptions. Check that it identifies the conflict and sends it to the owner.
- Ask why it chose a person at the account. Check that the answer identifies evidence of responsibility and preserves unanswered authority questions.
- Give it an account with an existing opportunity or ownership conflict. Check that it routes the account for coordination before outreach.
Review the reasons and sources behind each decision. A correct-looking answer without supporting evidence can still leave the reviewer unable to judge the selection.
Use the results to repair the brief and research process. Test the repaired version again before increasing autonomy. For a low-volume campaign where your team can review each account and message, manual outreach remains a reasonable option. The same brief helps a person keep the targeting and claims under control.
Keep the brief current as the campaign changes
Give the approved brief a version, an owner, and a review date. Keep a record of which version informed each campaign. When the offer changes, identify the affected claims, exclusions, and messages before starting another batch.
A buyer's reply can expose a missing qualification question. It can also show that the original description of the offer was unclear. Review that feedback with the person who owns the offer, then update the relevant instruction and its source.
Keep this review time in the operating budget. The full-cost comparison for AI BDRs and human BDRs includes the client work that remains outside a software fee or managed-service scope.
Quantonica starts with the sales materials and account context you provide. Use the pilot discussion to agree which claims, exclusions, and decisions need your review.
Give the brief to a colleague who did not write it. Ask them to reject an unsuitable account and flag an unsupported sales claim.