Ep 02
Follow-up and intake
A missed enquiry produces no signal at all — which is exactly what makes it the most expensive leak to leave unfixed.
The voices in this episode are AI-generated. The research, writing and opinions are Ramanjit Singh's.
Two hosts walk through why "did we follow up with that lead" is often genuinely unanswerable in a small business, why putting a chatbot on the busiest channel makes the problem worse, not better, and the four-step order — converge, assign, age, then automate — that actually fixes it.
Chapters
- 00:00 — The problem: an unanswerable question
- 02:10 — Why it happens: five channels, five owners, no owner for the set
- 05:00 — What most people try first, and why it backfires
- 08:15 — What actually works, in order
- 11:00 — Where a human stays in the loop
- 12:40 — One thing to check this week
What this episode claims
- A missed enquiry produces no signal — no complaint, no angry email — which is what makes it more expensive than a slow reply.
- Automating replies on the busiest channel before building a queue removes the only feedback loop you had.
- AI belongs at step four of intake — converge, assign, age, then classify and draft — never earlier.
Read the full written version: /insights/follow-up-falls-through-the-cracks
Full transcript
A: Ask an owner how many open enquiries the business has right now. Not roughly — the actual number.
B: And if it takes more than about ten seconds to answer —
A: — the problem isn't reply speed. It's that no single place knows what exists in the first place.
B: Enquiries come in on WhatsApp, email, a web form, the phone, sometimes Instagram DMs. Each of those channels has an owner. Somebody's watching the inbox, somebody's got the phone.
A: But the set of channels — all of them together — has nobody.
B: Which is why "did we get back to that lead from Tuesday" turns into "probably." That's the honest answer, and it's genuinely unanswerable most of the time.
A: And here's the part that makes it expensive rather than just annoying. A slow reply gets you a complaint. You hear about it, you can fix it.
B: A missed enquiry produces nothing. No signal at all. The person just goes somewhere else, and the business never learns it happened.
A: So what's the instinct, when an owner notices this?
B: Put an AI on the busiest channel. WhatsApp usually. So nobody waits.
A: Which sounds reasonable on paper.
B: It's actually where this goes wrong first, and for a reason that has nothing to do with whether the AI is good. If you can't currently see the queue, you also can't see what the bot got wrong.
A: You've added speed and removed the only feedback loop you had.
B: There's a second problem too, a subtler one. Automating replies on the busiest channel makes that channel more attractive to customers. So volume shifts even further toward the one place a human is no longer reading.
A: A chatbot doesn't fix scattered follow-up. It adds a fifth channel.
B: So what's the actual order? Because this is one of those cases where the fix is almost insultingly boring.
A: Converge first. Every enquiry, from every channel, lands as a row in one place — tagged with its source. WhatsApp, form, email, call log. No channel keeps its own private memory anymore.
B: Then force an owner. Every item gets a named person, assigned automatically on arrival — round robin, by category, whoever's on shift. "Unassigned" isn't a valid state the item can sit in.
A: Then make ageing visible. An item untouched for some number of hours changes colour. This one feature does more work than any AI in the system, because "nothing is overdue" becomes something you can check at a glance instead of something you hope is true.
B: And only then — step four — does AI come in. Classifying urgency and category on arrival. Drafting a first reply for the owner to approve. Summarising long threads so whoever picks it up doesn't read forty messages first.
A: Never the send, though.
B: Right — that's the loop-closing part. A draft prepared automatically, put in front of the person who owns the item, with the original request sitting right beside it. They edit, they send. The judgment stays exactly where it was.
A: Which is the human-in-the-loop piece — a person is still the one who decides whether a reply actually goes out, every time, on anything customer-facing.
B: One honest limit worth naming here — a queue doesn't create capacity. If four people genuinely can't handle the volume coming in, seeing that clearly is useful, but it's not itself a fix.
A: That's a staffing conversation the dashboard makes visible. It doesn't have it for you.
B: So — one thing to actually check this week.
A: Time three questions, don't estimate them. How many open enquiries exist right now. Which is the oldest one nobody's replied to. And who owns it.
B: If you can't answer all three inside a minute, that's the workflow to fix — before any AI touches it.
A: Written version's on the site, atinnovators.in, under insights. That's this one.