Use case
Sales research and follow-up
An agent prepares account briefs with a source for every claim, drafts follow-ups after meetings and proposes CRM updates. Reps approve what gets sent and what gets written.
The problem
How the work goes today
Before a call, a rep opens the CRM, the company website, LinkedIn, recent news and old email threads, and pieces together a brief in their head. For a busy rep that happens in the five minutes before the meeting, or not at all.
After the call, the follow-up email waits until the end of the day. Next steps live in a notebook. The CRM gets updated on Friday afternoon from memory, so stage, close date and contacts are often out of date when the forecast is reviewed.
Managers then chase reps for updates, and the data used to plan the quarter is the least reliable part of the process.
70%
Share of their time sales reps said they spend on tasks other than selling, such as admin and meeting prep, in a survey of 5,500 sales professionals across 27 countries. This is a vendor survey, run by a CRM company.[1]
25.1% faster
How much faster 758 BCG consultants finished tasks inside GPT-4's capabilities, with about 40% higher quality. On a task outside those capabilities, consultants using AI did worse, which is why research claims need a person to check them.[2]
How it works
Step by step
Highlighted steps are where a person decides. Everything else runs on its own and is logged.
Systems involved
- CRM (Salesforce, HubSpot, Microsoft Dynamics)
- Email and calendar (Microsoft 365, Google Workspace)
- Call recording and transcripts (Gong, Chorus)
- Enrichment (ZoomInfo, Apollo, Clay)
- Sales engagement (Outreach, Salesloft)
- Product analytics for usage signals (Amplitude, Mixpanel)
- 01System
Trigger
A new inbound lead, a booked meeting or a deal stage change in Salesforce or HubSpot starts the workflow.
- 02Agent
Research the account
The agent gathers CRM history, enrichment data from tools like ZoomInfo or Clay, public filings and news, and links every claim to its source.
- 03Agent
Score and route
It scores fit against your criteria and assigns the lead to the right owner in the CRM.
- 04Agent
Prepare the meeting brief
A one-page brief with open opportunities, past conversations and recent account news is sent to the rep's calendar invite or inbox.
- 05System
Capture the call
The transcript arrives from the call recorder (Gong, Chorus or Teams) once the meeting ends.
- 06Agent
Summarise and extract
The agent pulls out next steps, qualification fields and risks from the transcript.
- 07Agent
Draft follow-up and updates
It drafts the follow-up email in Outlook or Gmail and proposes CRM field changes such as stage, close date and contacts.
- 08Person
Rep approves
The rep edits and sends the email, and confirms or rejects each proposed field change before anything is written to the CRM.
- 09System
Write with audit trail
Confirmed changes are written to Salesforce or HubSpot with a field-level record of who approved them.
Measurement
What we measure
Agreed before anything is built, with a baseline taken from how the work runs today.
- Research time per account
- Minutes of rep time to produce a usable brief, measured on a sample of accounts before launch and again after.
- Follow-up within SLA
- Share of external meetings with a follow-up email sent within 24 hours, taken from calendar and email logs.
- Brief accuracy
- Share of facts in a weekly sample of briefs that a reviewer confirms against the linked source.
- Draft acceptance rate
- Share of drafted emails sent with no or minor edits, used as a rough quality signal per template and segment.
- CRM completeness
- Share of open opportunities with stage, close date, amount and next step filled in and updated within the last 14 days.
Pitfalls
What usually goes wrong
- Personalisation built on wrong facts
- Link every claim to a source, ignore sources older than a set age, and have the rep check the brief before anything is sent.
- Volume damages the sending domain
- Throttle sends, approve templates per segment, and follow Gmail and Yahoo bulk-sender rules and anti-spam law such as CAN-SPAM and PECR.
- The CRM fills with machine-written guesses
- Propose changes and write them only after the rep confirms, with a field-level audit record of who approved what.
- Enrichment data with no lawful basis
- Check the legal basis for each data source, especially for EU contacts, and keep new contacts at an account behind rep approval.
Build or buy
When a ready-made tool is enough
Custom work isn't always the answer. Here's how we'd decide.
- Salesforce Agentforce, HubSpot Breeze, Gong, Clay, Apollo, Outreach and Salesloft already cover research, call summaries and email drafting for a standard CRM-centred sales process.
- We found no independent, controlled study of the revenue effect of AI sales agents or automated outbound. Vendor claims exist but can't be checked, so we measure time saved and data quality rather than promise pipeline.
- Custom work pays off when research depends on your own data, such as product telemetry, warehouse signals or industry registries, or when the deal involves many stakeholders in a regulated sector.
- It also makes sense when the workflow crosses several systems with approval steps, or when customer data has to stay in your environment.
Industries
Where this applies
Sources
- 1.Salesforce, State of Sales, 6th edition (2024)
- 2.BCG, There's a time and place for generative AI (Oct 2023)
Figures are as reported by each source on its publication date. Examples are public industry cases, not AI Sewa clients.
Next step