Sales and marketing scenarios

The first sales agent works well with its CRM connections. Three more business units want the same thing.

1. Design and test one reference sales agent.
2. Copy its configuration for each business unit.
3. Tailor each copy's instructions, knowledge and rules.
4. Reconnect and re-scope each copy's skills and tool access.
5. Deploy each to its own channels.

Four units with different products, owners and rules don't belong in one agent. Copying a proven setup saves time while keeping them separate.

• Do the units share a CRM account?
• Are contacts and opportunities separated by unit?
• Can connections be reused without widening access?
• Does each unit need its own scoring or routing rules?

• Assuming a copy is complete. For security, copying an agent doesn't carry over every skill, key or connection. Check each channel, key, knowledge store, permission and bridge after copying.

After client meetings, reps want the meeting logged in the CRM and follow-up tasks created, without the admin work.

1. A transcript arrives, uploaded by the rep in Copilot or delivered by a meeting-recording tool where that's available.
2. A meeting analyst agent with live CRM access finds the client contact from attendee details.
3. It identifies the rep who owns the record.
4. It logs the meeting as an activity.
5. It proposes three follow-up tasks.
6. The rep approves all, some or none, or edits them.
7. Only approved tasks are created, linked to the right contact and, where relevant, opportunity.

The agent does the writing; the rep keeps control of what goes into the CRM under their name.

• Creating tasks before approval. Never.
• Assigning to the wrong owner.
• Guessing between opportunities. Ask when more than one matches.
• Losing the evidence. Keep references to the transcript.

The business wants each lead scored from 0 to 100 for intent and fit.

1. Agree the scoring rubric: factors, weights and exceptions.
2. Collect labelled examples of real leads, ideally 10 to 50.
3. Turn them into reviewed instructions or examples for the agent.
4. The agent scores each lead against the rubric and returns the score, the factors behind it, the evidence and any exclusions.
5. Optionally, the score is saved to the CRM.

A score is only as good as the rules behind it. Examples teach the agent where the boundaries are.

• Is 0 the least qualified?
• How are intent, company size, sector and timing weighted?
• When should a lead be "unscored" rather than 0?
• What does a high score trigger?

• Inventing a rubric and treating it as company policy.
• Confusing lead scores with conversation scores. They measure different things.

When a new lead enters the CRM, the business wants it enriched, the record updated, and a tailored first email prepared.

1. A new lead in the CRM fires a webhook that starts a Loop.
2. An agent with access to an enrichment provider, such as Apollo, looks up the person and company.
3. It updates the CRM record through a scoped tool.
4. It drafts a tailored email.
5. A person approves, edits or sends it back.
6. The email is sent and the outcome recorded.

One lead, one event, one message: this is a Loop, not a campaign. Review the first batch by hand, then automate more once quality is proven.

• Promising tools you haven't confirmed. Check exactly which actions your enrichment provider's connection supports.
• Consent, duplicates, enrichment credit costs, invalid addresses, CRM ownership and sending limits. Settle them before go-live.
• Choosing the wrong tool. If the real request is to message a whole list, use raia Control (S24).

The business wants a regular search for prospects matching a profile, enriched, ranked and prepared for outreach.

1. A schedule, or a person, starts a Loop.
2. A search-enabled agent finds companies and contacts that match the profile.
3. It enriches them within an agreed credit budget.
4. It scores them against the business's rubric.
5. It adds them to an approved CRM list, or drafts messages.
6. A person approves.
7. Approved prospects are handed to the outbound system or campaign.

This is a preparation pipeline. It may feed a raia Control campaign, but each stage doesn't need to live in Control. Decide where the hand-off happens.

• What's the target segment, and how many prospects?
• What's the scoring rubric?
• What are the duplicate rules and enrichment limits?
• Who reviews?
• Are messages drafts only, or sent automatically?

• Broad, unsupervised prospect collection, or emailing from unchecked search results.

Every morning, the marketing team wants a timely topic found, a blog drafted and the draft sent for review.

1. A daily schedule starts a Loop.
2. A research and writing agent uses approved web search or industry sources.
3. It analyzes current, cited information and writes a draft.
4. An editor approves or sends it back with feedback.
5. The draft is saved to the content system, and published only when explicitly authorized.

• Who's the audience, and what's the brand voice?
• What source quality is required, and which topics are allowed?
• Who checks facts, and who's the editor?
• What content system is used, and does an integration exist?

• Invented statistics or quotes. Every factual claim needs a source.
• Automatic publishing without the owner agreeing to that level of autonomy.

The business has a list of contacts or leads and wants a campaign with follow-ups over time.

1. Validate the list and confirm everyone on it is eligible to contact.
2. Set up a raia Control mission using the right agent and channel.
3. Track responses, and escalate follow-ups through authorized workflows.

Lists and ongoing sequences are what raia Control is built for. A single new lead is better handled by a Loop (S12).

• Where did the list come from, and what's the legal basis for contacting these people?
• What's the cadence?
• How are unsubscribes and suppression handled?
• Which channel, and who owns the campaign?
• Is this transactional or marketing?

• Treating an agent plus a Loop as a campaign. Control is a separate product with its own setup. See raia Control.