AI agent team guide
Build a Customer-Proof Campaign With an AI Team
A source-permission register and reusable customer-proof campaign

A customer-proof campaign starts with evidence you are allowed to use. Identify the customer’s actual problem, what happened, which statements are supported and what permission applies. Then ask an AI team to build a small draft package that preserves those limits across an article, a short post and a sales excerpt.
Content Manager leads the asset, Customer Voice Researcher prepares the traceable evidence, Brand Voice Editor improves the story and Content Repurposing Producer adapts the approved version. The output can be useful even when the evidence supports a process story rather than a performance claim.
Build your four-agent team
| Agent template | Its job on the team | Bring to its setup | Reviewable return | Team |
|---|---|---|---|---|
Content ManagerLead agent | Own the story brief and approval-ready asset. | Audience, intended claim, evidence pack, approval owner and distribution constraints. | A draft story, open factual questions and a clear review handoff. | Add to team |
Customer Voice ResearcherSupporting agent | Build the source and permission ledger. | Authorized interviews, customer statements and outcome records. | A traceable problem-evidence pack separating observation, interpretation and missing proof. | Add to team |
Brand Voice EditorSupporting agent | Make the story clear without overstating it. | The source-linked draft, approved voice samples and exact claims. | A voice-preserving edit with meaningful changes and unsupported language flagged. | Add to team |
Content Repurposing ProducerSupporting agent | Adapt only the accepted source version. | Approved story, allowed quotations, channel constraints and attribution limits. | Faithful derivative drafts that link back to the full context. | Add to team |
Set up your team in Finta
- Add each template individually. Use the Add to team action for each named coworker, review the import setup and complete its Getting Started skill.
- Give each role the right context. State the shared objective, then supply the authorized documents, notes or exports that role needs. Do not assume an unconnected system is available.
- Review the handoff. Accept or correct one return before giving its evidence and open questions to the next coworker. Importing four templates does not itself run a bulk team.
- Make recurrence deliberate. Complete a useful manual pass first. Review any supported routine and enable it only when the scope, inputs and permissions are right.
Prepare the evidence
- A defined audience and the practical question the story should answer.
- Permissioned source notes with exact quotes, attribution rules and review contacts.
- The actual outcome definition, period and supporting records when a result is claimed.
- The current approved product facts and the scope of any customer publication permission.
Run the workflow
1. Separate the story you want from the evidence you have
Have Customer Voice Researcher map every proposed claim to a source. A customer saying a workflow felt clearer is not proof of a measured time saving. Keep subjective statements attributed and quantitative claims tied to their real definitions.
2. Choose the strongest honest asset
Content Manager decides whether the sources support a customer story, a process example or a general educational piece. Do not create a named case study when customer evidence or permission is absent. Preserve unanswered factual questions for the reviewer.
3. Review facts before polishing the headline
Brand Voice Editor can make a precise story compelling without adding a percentage or implying a guaranteed result. The responsible reviewer checks the final version, exact quotes, product statements and permission scope. Draft approval and customer approval are separate where both are required.
4. Repurpose the approved version, not the early draft
Give the producer the accepted source and allowed uses. A short sales excerpt should retain the conditions that make a result meaningful. Record which version each derivative used and verify actual destinations when authorized publication occurs.
A fictional proof ledger that changes the headline
The customer, statements and records below are invented for teaching. They must not be used as a real testimonial, case study or Finta result.
| Proposed claim | Available fictional evidence | Editorial decision |
|---|---|---|
| “Reporting took half the time.” | No comparable time measurements are supplied. | Omit the performance claim. |
| “The team clarified who owned the final review.” | A permitted interview account and the accepted owner register support the process description. | Describe the process with the agreed attribution. |
| “Every customer can get the same result.” | Only one contextual account exists. | Do not generalize the result. |
| A short social excerpt | Permission covers a reviewed excerpt but not raw interview notes. | Adapt the approved wording and keep the source notes private. |
The stronger asset explains a believable change in work. It does not need a fabricated metric to be useful to the next reader.
Copy the team brief
Prepare a customer-proof draft package for [reader and useful decision].
Use [permissioned evidence], [exact quotes and attribution limits], [verified outcomes if any] and [approved product facts].
Customer Voice Researcher: build claim, source, definition, date, permission and uncertainty fields.
Content Manager: choose the asset the evidence can support and draft it.
Brand Voice Editor: improve clarity while preserving exact facts and limits.
Repurposing Producer: adapt the approved version for [channels] without dropping material context.
Return factual questions, review owners and version-linked derivatives.
Do not invent a customer, quote, result, endorsement or permission. Do not contact customers or publish.Review the return
- Each material claim has a suitable source and an applicable permission.
- Numbers preserve their period, denominator and attribution limits.
- Every derivative maps to the approved version rather than an unreviewed draft.
Questions before you start
Can AI write a case study if we only have a few notes?
It can help prepare a draft from authorized notes, but the notes must actually support the story and its claims. If customer proof or permission is missing, choose an educational example and label any fictional material clearly.
Can a quote be shortened?
Only when the edit preserves its meaning and fits the permission and review requirements. Keep the source wording for comparison and have the appropriate owner approve material changes.
Sources and next steps
This playbook uses the current published Finta marketplace packages linked in the team table. Open each profile to review its instructions, context, Skills and any routines before importing.