USE CASE · SALES & REVOPS
Your reps sell. The AI Sales Worker does the admin around it.
A production AI system that researches leads, drafts follow-ups and keeps your CRM accurate. It runs inside HubSpot, Salesforce or Pipedrive, your inbox and Slack, and a person approves every message before it goes out.
PROBLEM
Most B2B sales teams don't lose deals for lack of effort. They lose them because the effort goes to the wrong places.
- Inbound leads wait hours or days because nobody has had time to research them.
- Reps rebuild the same account research from scratch before every call.
- Follow-ups slip once a rep has more than a handful of live deals.
- CRM data is incomplete, so the forecast is a guess and handovers lose context.
Generic chat tools don't fix this. They sit outside the CRM, have no context on your pipeline, and leave someone copying and pasting.
CURRENT MANUAL WORKFLOW
- 1.A lead fills in a form or emails the sales inbox.
- 2.Someone notices it, often hours later.
- 3.A rep searches the company website, LinkedIn and Companies House by hand.
- 4.The rep checks the CRM for history, if they remember to.
- 5.The rep writes a reply from a blank page or an old template.
- 6.After the call, notes stay in a notebook or a transcript nobody reads.
- 7.The CRM gets updated on Friday, partially.
- 8.Follow-up depends on memory.
AI-OPERATED WORKFLOW
Capture
The worker picks up every new form fill and inbound email as it arrives.
Enrich
It pulls public company information and existing CRM history into a single record.
Qualify
It scores the lead against your ICP and qualification rules and writes down why.
Route
It assigns the lead to the right owner and posts a short brief in Slack or Teams.
Draft
It writes a reply in your tone from your approved messaging and places it in the rep's review queue.
Prepare
Before a meeting, it produces a one-page account brief.
Log
After the call, it turns the transcript or notes into structured CRM updates, next steps and tasks.
Nudge
It sends daily digests of deals with no next step and follow-ups that are due.
HUMAN CONTROL
You decide what the worker can do on its own and what needs a person.
Approvals
Nothing reaches a prospect without a human approving it. CRM writes can be automatic or approval-based, field by field.
Review queue
Drafts and proposed CRM changes sit in one queue where reps approve, edit or reject them. Edits feed back into the worker's examples.
Audit trail
Every action is logged with its inputs, sources, output, approver and timestamp.
Escalation
Low-confidence cases, existing-customer issues and anything outside policy go to a named person instead of being guessed.
Data boundaries
Least-privilege access to your systems, deployment in your cloud or a UK/EU-region environment, and model providers that don't train on your data.
IMPLEMENTATION APPROACH
We start with one workflow, usually inbound triage or post-call CRM updates, and get it into production before adding more.
| Phase | What happens | Timeline |
|---|---|---|
| 1. Workflow review | You show us the current process. We agree the workflow, success measures and data access. | Week 0 (one call plus a short written scope) |
| 2. Discover & design | We baseline current performance, map data flows and approval points, and build an evaluation set from real examples. | Week 1–2 |
| 3. Build & integrate | We connect to the CRM, inbox and Slack/Teams, then build the worker, review queue and logging. | Week 2–4 |
| 4. Shadow mode | The worker runs alongside the team. Outputs are reviewed and none are sent automatically. We measure against the baseline. | Week 4–5 |
| 5. Go live & operate | We switch on for real, with monitoring and a runbook, then extend to the next workflow if the numbers justify it. | Week 6 onwards |
Phases 2–5 are delivered as a Production AI Sprint. Production AI Sprint
FAQ
- Will it send emails to prospects on its own?
- No, not by default. Every outbound message goes through a human approval step. Some teams later allow specific low-risk messages, such as meeting confirmations, to send automatically. That is your decision, and it's logged.
- Which CRMs does it work with?
- HubSpot, Salesforce and Pipedrive are the usual ones. Any CRM with a usable API can be connected. We confirm access in the workflow review before any build starts.
- Is this LinkedIn automation or mass cold email?
- No. The worker supports your team's real selling: research, drafts, follow-ups and CRM accuracy. It doesn't scrape LinkedIn or send bulk sequences.
- Where does our data go?
- Your data stays in your systems and a deployment you control, in your cloud tenancy or a UK/EU-region environment. We use model APIs that don't train on your data, and we document every data flow during design.
- How do we know it's working?
- We measure the current process in week 1 and run the worker in shadow mode against that baseline before go-live. You see the comparison before you decide.
Show us how your pipeline actually runs.
Walk us through one sales workflow, warts and all. We'll tell you plainly whether an AI worker would help, and what the first production version would look like.