AI Systems · Sales operations
Where AI belongs in your inquiry pipeline — and where it doesn't
Most lost inquiries are lost to delay and inconsistency, not to bad salesmanship. A practical map of which steps to automate, which to assist, and which to leave with people.
This is a preview edition under review. It may change before publication.
When a business says it wants “an AI chatbot for leads”, the real problem is usually narrower and more expensive: inquiries that wait too long for a reply, qualification that changes depending on who answers, and follow-up that relies on someone remembering. A chatbot can help with one of those. A system that handles the whole pipeline can help with all three.
The useful question is not whether to use AI, but at which step — and with what authority. We break an inquiry pipeline into six steps and treat each differently.
1. Capture: no AI needed, just plumbing
Inquiries arrive through web forms, email, WhatsApp, marketplace platforms and phone calls. Before anything intelligent happens, they need to land in one place with a consistent structure. This is integration work. It is unglamorous, and it is where most of the value starts: you cannot respond consistently to inquiries you cannot see in one list.
2. First response: automate, within strict limits
An immediate, relevant first reply is one of the most valuable things AI can do in a sales process. It acknowledges the request, answers simple factual questions from approved content, and asks the next question that matters. The limits matter as much as the capability:
- Answers come only from content you have approved — services, availability, policies — never from the model's general knowledge.
- No commitments: prices, discounts, deadlines and contract terms are always confirmed by a person.
- The assistant identifies itself as automated and offers a human at any point.
3. Qualification: assist, and show the reasoning
Qualification criteria usually live in someone's head. Writing them down — budget range, location, timing, fit with your services — is the first deliverable, and it improves the process even before automation. AI can then ask the qualifying questions in natural language and propose a score. The score should come with its reasons, so a salesperson can disagree with it in seconds.
A useful test
If your team cannot agree on how to qualify ten past inquiries by hand, an AI system will not agree with them either. Align the criteria first.
4. Routing: rules first, AI for ambiguity
Most routing is deterministic: this service goes to this team, this region to that office. Rules handle it reliably and cheaply. AI earns its place when the request is ambiguous — a long email that mixes a complaint with a new order, for example — and needs interpreting before it can be routed.
5. Follow-up: automate the schedule, not the relationship
Scheduled reminders, booking links and status updates are ideal for automation. Persuasion is not. A good system makes sure no qualified inquiry is left without a next step, then hands the conversation to a person with a summary of everything that has happened so far.
6. CRM updates: automate completely, with an audit trail
Updating the CRM is the step people skip when they are busy. It is also the step that makes reporting possible. Every message, score and status change should be written automatically, with a record of whether a person or the system made it.
What to measure
- Time to first response, by channel and hour of day.
- Share of inquiries that receive a qualification decision.
- Agreement between the system's score and your team's judgement.
- Appointments booked and follow-ups completed.
Measure the current process before changing anything. Without a baseline, any improvement is an opinion.