We turned Kommo CRM lead routing into a shift system the sales manager runs by writing to an AI agent. Our client is a health tourism clinic in Türkiye. Leave, a sick day or someone leaving the team now takes one message: the agent changes the rule, deploys it, checks it live and reports back. All 13 routing changes in about three months were made this way.
Sources: shift rules and change records, Aug – Nov 2026. Response time measured live. Lead volumes are not shared.
A health tourism clinic in Türkiye answers patients who come from abroad. The team uses Kommo as its healthcare CRM; leads arrive from WhatsApp, ad forms and the website. The shift plan lived in the sales manager's spreadsheet, but real life didn't fit it: someone goes on leave, someone falls ill, someone leaves. Every change had to reach routing the same day.
When a rep went on leave or fell ill, their leads had to go to someone else that day. Someone had to remember it and enter it into the system.
When someone left the team, the plan had to be rebuilt. If a gap slipped through, a new patient would land with someone who was no longer there.
Of the two reps sharing the day shift, one always got the busier hours. A gap that looks small in a single week grew over the months.
Patients often fill in a form and message on WhatsApp minutes later. With leads handed out in turn, the same patient could land with two different reps.
Kommo resends a notification when it gets no response within 2 seconds. Late retries pulled a lead already handled back to the first stage, sometimes to another rep.
The interface is a chat: no code, no dashboard. Whoever runs the team writes the change in plain words; the AI agent applies it, checks it and reports back.
“Send every lead that comes in between 1:35 pm and 6:00 pm today to one rep.”
How we workSEMROI built the agent. We build agentic processes for you too: we can run them for you, or hand them over to your team.AI agent development
message(off: A, to: B, when: tomorrow)exception = handover(A → B, tomorrow)rota.add_exception(handover, A → B)deploy → liverouter.health_check → oknotify(manager, A → B, tomorrow)Every change is stored as a dated record with its reason.
When a new lead opens in Kommo, the system answers Kommo first, then works out who is on shift that minute. Exceptions take precedence over the plan; the lead lands in that person's pipeline at the first-contact stage.
The shift rule is generated from the sales manager's 4-week rota; it matched the 224-day plan exactly in 1,124 checks.
First we made the routing itself solid, then we turned every real-life situation into an exception type. Today, when the sales manager asks for a change, the agent knows which type it needs; nobody touches the code.
Built on the clinic's existing Kommo account, and running there.
OutcomeThe rota and the routing in Kommo now say the same thing.
OutcomeThe split is even; no patient is divided between two reps.
Outcome13 routing changes in about three months, all made by writing to the agent.
OutcomeRetries no longer pull leads back or hand them to another rep.
OutcomeThe team got smaller; not one lead was routed to the rep who left.
OutcomeNobody copies form leads by hand or misses the moment of sale.
OutcomeThe rota and the routing in Kommo now say the same thing.
OutcomeThe split is even; no patient is divided between two reps.
Outcome13 routing changes in about three months, all made by writing to the agent.
OutcomeRetries no longer pull leads back or hand them to another rep.
OutcomeThe team got smaller; not one lead was routed to the rep who left.
OutcomeNobody copies form leads by hand or misses the moment of sale.
The person running the shifts does not need to learn a dashboard. Because the rules live as data, "A is off tomorrow" becomes a safe, one-line change. The agent applies it, checks it and reports back.
If a team trusts a spreadsheet, the new system must first give the same answer. Here we tested the rule against the rota with 1,124 checks over 224 days; the trust came from that match.
If the same person writes on two channels minutes apart, routing in turn splits one conversation in two. Hour-by-hour routing keeps the patient with one rep; rotating hours keep the split fair.
Systems like Kommo resend a notification when they don't get a quick response. Answering first, doing the work in the background and checking a lead's state before touching it makes late retries harmless.
We gave the interim rota no end date; it runs until the new plan arrives. That way a forgotten date is never enough to send leads to someone who has left.
For definitions, see what agentic AI is; for applications, AI agent examples.
When a new lead opens in Kommo, a webhook tells the routing system. The system works out who is on shift that minute and assigns the lead to that rep's pipeline. The rules come from the shift plan; exceptions such as leave or sick days take precedence. So no lead waits for someone who is off.
The sales manager writes the request in plain words: "A is off tomorrow, send their leads to B." The AI agent picks the right exception type, adds the rule, deploys it, checks it live and reports back. No code or dashboard needed. In this project, all 13 changes in about three months were made this way.
They can if leads are handed out in turn. Here, the two reps sharing the day shift take leads hour by hour; a patient who fills in a form and messages on WhatsApp minutes later lands in the same hour, so with the same person. Kommo's duplicate check by phone number is on too.
Kommo resends a notification 5 and 15 minutes later when it gets no response within 2 seconds. If the receiver responds only after finishing its work, a late retry can pull the lead back to the first stage. The fix: respond first, work in the background and check whether the lead has moved on before touching it.
Yes. In a team that uses Kommo as its WhatsApp CRM, a new WhatsApp lead can be assigned to the rep on shift the moment it opens. Here the rule is generated from a four-week rota, with public holidays defined in advance. The rep finds the lead in their own pipeline at first contact; nobody sorts leads by hand.
In this project, no leads went to the departing rep in their last two weeks; the rest of the team covered their day hours. Then a three-person interim rota was set up. It has no end date on purpose and rotates until the new plan arrives. So a forgotten date never sends a lead to someone no longer on the team.
In this team, people decide and the agent acts. The sales manager says who gets leads and when; the agent writes the rule, deploys it, verifies it live and keeps every change as a dated record with its reason. The team learned no new dashboard; the only new habit is writing a message to the agent.
We build agentic processes and AI agents for business. Tell us about your shifts and lead sources; we set it up, and your team runs it by chat.
