CRM with qualification
Every incoming request is analysed, scored and ready to handle.
Every incoming request is read, scored and routed before a salesperson even opens their inbox.
Reference system: a need I come across often, documented with the exact architecture I would put in place. Other client work is presented in meetings.
Problem
Requests come in through the website form, email and LinkedIn. Nobody knows which ones to handle first.
Good opportunities wait for days, weaker ones take up as much time as the rest, and follow-up lives in a spreadsheet that everyone fills in their own way.
Context
- Users
- A sales team of 3 to 10 people and a management team that wants to follow the pipeline.
- Current process
- Every request read by hand, copied and pasted into a spreadsheet, follow-ups done from memory.
- Constraints
- Keep the existing email and CRM, no re-keying for the sales team, a fast reply on priority requests.
- Tools in place
- Website form
- Spreadsheet
- Existing CRM
Solution
A qualification layer placed between the incoming channels and the CRM. Each request is normalised, analysed by a language model against criteria defined with the team (budget, timing, fit), then routed with a draft reply.
The salesperson approves. They no longer sort.
Why this approach. Replacing the CRM wouldn't have solved anything: the problem was sorting, not the tracking tool.
Architecture
Input
- Form The website form sends each request by webhook, the moment it arrives.
- Incoming email Requests received by email are captured and go through the same pipeline.
Automation
- n8n The workflow normalises the request, calls the AI, writes to the database and distributes the results.
Data
- Supabase Single source of truth: requests, scores, reasons and history, available to the whole team.
AI
- Claude Analyses the request against the team's criteria, gives a score with its reasons and drafts a reply.
Output
- CRM The record is created with the score and the reason: no re-keying.
- Slack alert High-priority requests trigger an immediate alert to the right salesperson.
Interface
- Pipeline view A simple view of the pipeline by priority, for the team and management.
Features
- Qualification on explicit criteriaBudget, timing and fit are defined with the team, written down in black and white, and adjustable.
- Visible score and reasonEvery score comes with its justification: you can see why a request moves up the queue.
- Draft replyThe AI drafts, the salesperson reviews and sends. People stay in control.
- CRM record without re-keyingThe record is created and filled in automatically from the original request.
- Alerts on hot requestsPriority opportunities surface immediately, to the right person.
- Searchable historyEvery sorting decision is logged, so the criteria can be adjusted over time.
Target result
- No priority request left without a reply the same day.
- Zero re-keying between the channels and the CRM.
- Qualification criteria shared by the whole team.
How it is measured
- Average time to first reply
- Share of requests qualified without intervention
- Conversion rate by priority level
What I do
- Workshop on qualification criteria
- Flow design
- n8n workflow
- AI prompts and guardrails
- Supabase data model
- CRM and email integrations
- Production launch and follow-up
Try it.
A working version of the system, with sample data. Nothing is pre-recorded: it all runs in your browser.
Incoming requests
- Received by webhook
- Analysed by Claude
- Priority score
- Reply drafted
- CRM record created
Pick a request, then run the agent.