By Luiz Otávio Gonçalves Published 2026-09-11

Kommo Salesbot: what it is, what you can automate, and how to configure it without annoying the lead

Kommo Salesbot: what it is, what you can automate, and how to configure it without annoying the lead? Straight to the point: the Salesbot is the heart of Kommo’s native automation, but it works as a state machine (conditions → actions → transitions), not like an “AI chatbot” that talks like a human. When you use it right, it qualifies, distributes leads, triggers messages at the right time, and moves funnel stages. When you force “infinite humanized conversation,” it breaks, creates noise, drops conversion, and can even burn your WhatsApp number with too many sends, too fast, and outside hours.

I’m writing this because I see the term “salesbot kommo” already nearing the top 10 and yet most content treats Salesbot as just a welcome message “hi, how are you?”. In production, in real operation, a well-made Salesbot is invisible. Poorly done is blocking.

My thesis (and what changes in your execution)

If you understand one thing, the reading is worth it: Salesbot in Kommo is not AI. It’s logic. It’s a state machine that reacts to events and conditions and pushes the lead through the process. This is great for Process (qualification, routing, follow-up, SLA, stage change). And it’s terrible for “improvisation” (trying to imitate a human for 30 messages).

Those who try to make the Salesbot “converse” usually make three mistakes:

• Endless flow: the lead responds with anything and you have no clear exit (fallback), it becomes a loop.
• Too many messages: long sequence, several “touches” in minutes, looks like spam.
• No time lock: sends messages at dawn, on holidays, or outside your real SLA. Result: irritation + complaint + risk of blocking.

What Kommo Salesbot is (without romanticizing)

The Salesbot is Kommo’s conversational automation builder for channels like WhatsApp (depending on integration), Instagram/Meta (when applicable), website chat, and other connectors. In practice, I treat it as an engine for:

• Input: event that triggers the flow (message received, lead created, stage change, tag applied, etc.).
• State: where the lead is in the flow (e.g., “Waiting for choice 1/2/3”).
• Transition: rule that moves from one state to another (e.g., “if answered 2 → goes to ‘Scheduling’”).
• Action: send message, create task, change responsible, apply tag, move stage, call webhook.

This is process engineering, not “chat.” And when you start seeing it this way, you stop fighting the tool and start to build automation with ROI.

What can truly be automated (and what I implement the most)

This is where Salesbot shines. These automations seem simple but unlock scale because they take the team out of “copy and paste” and reduce response time.

1) Conditional greeting (without sounding like an annoying robot)

A “generic” greeting is the quickest way to sound like a bot and lose the lead. I prefer conditional greeting based on Source, time, Stage and responsible person.

• By time: if it arrives outside business hours, I inform the real SLA and collect one essential piece of data.
• By source: a “price” ad requires a different question than a “scheduling” ad.
• By history: if the lead/contact already exists, I avoid repeating questions and go straight to the point.

Golden rule: the bot’s first message is not to “make friends.” It is to reduce friction and direct the next step.

2) Objective qualification (2 to 4 questions, maximum)

Qualification works when you ask the minimum to decide the next routing. In production, I recommend:

• 2 to 4 questions (above that, response drop-off is normal).
• Multiple-choice questions multiple choice whenever possible (reduces ambiguity and interpretation errors).
• An exit to “talk to a human” at any stage (without humiliating the lead).

Examples that deliver ROI:

• “Do you want to Budget or schedule?”
• “Which unit/city?”
• “When do you want to resolve this?” (urgency)

3) Lead distribution (routing) with clear rules

If you have more than one salesperson/attendant, manual routing is a bottleneck. Salesbot handles the basics very well:

• round-robin (rotation) by queue
• By region (city/state/neighborhood)
• By product/service
• By channel (WhatsApp vs Instagram vs website)

What I insist on implementing together: task + SLA. Routed? It already creates a service task with a deadline. Without this, distribution becomes “I passed the lead and now it’s someone else’s problem.”

4) Automatic follow-up (short, spaced, and locked)

The classic mistake is turning follow-up into spam. I detailed this problem (and the right way) here: Automatic sales follow-up: what it is, why almost everyone does it wrong, and how to set it up.

In Salesbot, what works:

• Follow-up of 1 to 3 touches, spaced (hours/days, not minutes).
• Message with an action (reply 1/2, click, send info).
• Stop automatically when the lead responds or when a human takes over.

5) Move stage in the funnel (native automation with governance)

Automatic stage movement is powerful because it creates Process. But it needs explicit rules, otherwise it becomes a mess.

• If answered “I want to schedule” → stage “Scheduling.”
• If city X was informed → stage “Regional Screening.”
• If no response in 24h → stage “Follow-up 1.”

Want a broader view of commercial automation (what works and what doesn’t)? I wrote about it without romanticizing here: Sales automation: what it really is, what can be automated (and what can’t).

What NOT to expect from Salesbot (so you don’t stall your project)

It serves for process. It doesn’t serve to “be human.” Points where I see people failing:

• Real AI: Salesbot is not an LLM. It doesn’t “understand context” like ChatGPT. It executes rules.
• Long and open dialogue: if you want open questions and unambiguous interpretation, you need another architecture (e.g., agent/AI + security layer).
• Complex negotiation: discount, objections, comparison, proposal… that is human (or well-integrated, well-tested AI with limits).

Want to combine automation and AI the right way (without hype)? I explain how they complement each other here: Automation and artificial intelligence: how they complement each other.

How to configure Salesbot without annoying the lead (production checklist)

Now let’s move to execution. This is what I require from my team and projects: configuration with governance.

1) Define the flow’s objective (1 objective per bot)

A Salesbot that tries to do everything becomes a monster. I define 1 objective per flow:

• Qualify and route
• Schedule
• Confirm attendance
• Recover cold lead

2) Design the states before Kommo (paper, Miro, or text)

I literally write like this:

• State A: Greeting
• State B: Choice (1/2/3)
• State C1: Routing team X
• State C2: Routing team Y
• State D: Close + task

This avoids “clicking and seeing.” Method > improvisation.

3) Use multiple choice and normalize answers

If you leave an open question (“which service do you want?”) you get: “hi,” “good morning,” “yes I want,” audio, sticker. The automation dies.

Prefer:

• “Type 1 for Quote, 2 to Schedule, 3 to Talk to a human.”

4) Always provide an escape path to a human

I consider it mandatory. And it’s not just “talk to an attendant.” It is:

• Change responsible person
• Create task
• Send a short final message (“Perfect, I've connected you with someone from the team.”)

5) Anti-spam locks (this prevents blocking)

This is the point that causes the most damage.

• Time lock: if outside business hours, do not start a long sequence. Collect 1 piece of data and stop.
• Realistic delay: avoid message bursts. Space them by seconds and, in follow-up, by hours/days.
• Attempt limit: 1st attempt + 2nd attempt and then stop. No chasing.
• Auto-stop: if the lead responded, stop automation and pass to a human (or change status).

6) Logs and metrics: without data, you only have opinions

I track at least:

• % that answers the first question
• % that falls into fallback (“I didn’t understand”)
• Time until first human response (when necessary)
• Conversion by source (ad/channel)

If you want to measure automation ROI seriously (and not just by “feeling”), I left the method here: how to measure the ROI of business automations.

What can be integrated via webhook (Kommo → n8n) and when it’s worth it

When native automation reaches its limit, I integrate with n8n to scale logic and integrations: enrich lead, check postal code, validate CNPJ, fetch schedule, trigger events, record in spreadsheets/data warehouse, activate AI with guardrails, etc.

Common webhook cases Salesbot → n8n:

• Enrichment: takes email/phone and cross-checks in another database.
• Schedule: checks availability and returns options.
• Intelligent routing: rule by salesperson load, SLA, score.
• Record: saves in database, triggers event in BI.

My practical rule: if you can do it natively without hacks, do it natively. If you need 3 chained “if/then” and it starts to get fragile, move it to n8n and centralize the logic.

Practical flow examples (for you to copy and adapt)

Price (real data) and what I will NOT make up here

I won’t guess license cost, number cost, WhatsApp BSP cost, nor say “it costs X per month” without reviewing your account and scenario (channels, users, integrations). Price varies by plan, exchange rate, taxes, channel, and what you connect.

The real data I can safely point you to is where you consult the updated snapshot, with numbers and 2026 scenario: How much does Kommo CRM cost? Prices and plans in 2026 and also Kommo plans and prices in Brazil in 2026: which plan is worth it. (Yes: I prefer to send you to a page I update rather than “lock in” a number here and misinform you.)

Who the Salesbot is for (and who it’s NOT for)

Serves if you want:

• Respond quickly without relying 100% on humans
• Standardize screening
• Distribute leads with SLA
• Execute follow-up with governance
• Connect Kommo to integrations via webhook

DOES NOT serve if you want:

• An “AI attendant” that negotiates and closes alone, without rules or supervision
• Long and open conversations (no multiple choice, no structure)
• Sending mass messages and calling it automation (this is asking to get blocked)

The classic mistake that burns your WhatsApp number (and how to avoid it)

I’ll repeat because it causes real damage: bot that sends too many messages, too fast, without time lock. You annoy leads, increase reports, reduce responses, and put your channel at risk.

How I avoid it:

• Short cadence and meaningful (not “just another message”).
• Time lock with an expectation message (SLA).
• Automatic stop when the lead responds.
• fallback: if it didn’t understand, asks to type 1/2/3 or calls a human.

Executive summary

• Kommo Salesbot is state machine, not AI.
• Excels at: qualifying, routing, creating tasks/SLA, follow-up with lock, moving stage.
• Fails when you force “infinite human conversation.”
• Risk number 1: spam + no time lock → annoys leads and can burn your WhatsApp number.
• When you exceed the native limit, use webhook + n8n to scale with governance.

If you want me to design, implement, and optimize your Salesbot in production (with locks, metrics, and integration when needed), there’s only one way: request a project.