← Blog

Kommo Salesbot in 2026: how to automate WhatsApp without annoying the lead (and without killing your conversation)

If you want to use the Kommo Salesbot to automate customer service on WhatsApp without annoying the lead and without killing the conversation, the rule is simple: automate low-friction screening, routing, and follow-up — and pass to a human at the right moment (price, real objection, negotiation, and closing). I have been implementing this in Brazil from the start: method, SLA, and data. No improvisation.

What I’m going to give you here is an angle that almost no one covers: the Salesbot is not “a robot that sells”. It is a Process engine. When you design it as a process (response time, queues, tags, funnel stages, handoff criteria), you gain ROI. When you design it as a “cute chatbot,” you lose money and damage your brand.

The biggest conversion killer: using a bot to “chat” instead of using a bot to “operate”

When I take on a stuck operation, 80% of the time the Salesbot is making one of these two mistakes:

  • Too many questions (full registration, a thousand fields, “tell me more”) before delivering value.
  • Robotic and long text that doesn’t respect WhatsApp’s context (where people want speed).

WhatsApp is an intent channel. The lead entered there because they want to solve something quickly. So the Salesbot has to be a bridge to decision, not a barrier.

What to automate with Salesbot (what gives ROI) vs. what NOT to automate

Let’s go straight to what works in practice.

Automate (high ROI, low risk)

  • Immediate confirmation: “Received your message + next step” within 10–30 seconds.
  • Objective screening: 1 to 3 questions maximum (e.g., city, service type, urgency).
  • Routing: send to the right seller’s queue/channel/field with a clear rule.
  • CRM hygiene: apply tag, fill field, advance funnel stage automatically.
  • Automatic follow-up for “unanswered” and “cold conversations,” with a short cadence.
  • Reactivation of old leads by segment (no spam, with context).

Do not automate (high risk of annoying and losing sales)

  • Negotiation (discount, deadline, conditions): bot becomes a “wall” and kills trust.
  • Complex objections: “it’s expensive,” “I want to think,” “I’ll compare” — that’s human.
  • Consultative selling (B2B, industry, projects): bot can support, not lead.
  • Sensitive discussions (health, legal): here the reputational risk is high.

If you want to dive deeper into what true sales automation is (what can and cannot be automated), I explain in detail in this post: Sales automation: what it really is, what can be automated (and what can’t).

The playbook I use: 5 automation blocks that don’t ruin your operation

I design Salesbot in 5 blocks. This becomes the implementation standard and avoids “Frankenstein.”

1) Welcome block (response in seconds)

Objective: reduce first response time and keep the lead engaged in the conversation.

  • Short message
  • 1 objective question or 3 choice buttons
  • Make it clear there’s a human behind

Practical example (good): “Perfect. To direct you quickly: do you want (1) a quote (2) schedule (3) ask a question?”

Example (bad): “Hello, I am the virtual assistant. Please provide full name, CPF, email, neighborhood, how you found us, and describe your problem in detail.”

2) Minimal screening block (1–3 questions)

Objective: classify and routing without creating friction. Here I think about “what is the minimum for the next action?”

  • 1 context question (e.g., city)
  • 1 need question (e.g., service A/B/C)
  • 1 timing question (e.g., urgent/this week/no rush)

You collect the rest with a human or forms when it makes sense. If you overdo screening, you create an invisible queue: the lead gives up before reaching the team.

3) Routing block with SLA (the part almost no one implements)

Goal: it’s not “send to a seller.” It is deliver the right lead to the right person with a traceable rule.

What I configure:

  • Distribution rule (round-robin, by region, by product, by portfolio)
  • SLA (e.g., if no response in X minutes, reassign)
  • Audit (responsible field, timestamps, tags)

This topic relates directly to CRM governance. If your operation becomes a mess, you can’t scale. Read later: Kommo CRM in 2026: how to design governance (SLA, fields, and audit) so the CRM doesn’t become a mess.

4) Automatic follow-up block (no spam, with cadence)

Objective: recover opportunity that would die due to lack of team routine.

I work with follow-up in two layers:

  • Operational follow-up: “I saw your message, just confirming if you want to proceed with X.”
  • Value follow-up: “To help you: here’s a summary of how it works / what I need to give you a quote.”

Typical cadence (simple example):

  • +2h: quick touch
  • +24h: touch with context
  • +72h: last attempt + graceful exit (“can I close here?”)

If you want a practical guide just on this, it’s here: Automatic sales follow-up: what it is, why almost everyone does it wrong, and how to set it up.

5) Handoff to human block (the critical point of “not annoying”)

This is the turning point. The question that drives the result is:

“What event triggers a human now?”

I use objective triggers, for example:

  • Lead chose “quote” and answered minimal screening
  • Lead asked price/condition (“value”, “cost”, “installments”)
  • Lead showed urgency
  • Lead is premium category (e.g., high ticket / B2B)

At this moment, Salesbot steps aside and a message appears like: “I will connect you with a specialist now.” This reduces friction and increases the perception of customer service.

How to measure if your Salesbot is helping or sabotaging (adult metric)

I don't evaluate Salesbot by “how many messages it sends.” That's vanity. I evaluate by funnel impact.

Metric What it indicates How the bot influences
1st response time Contact speed Immediate welcome + routing
% of leads who answer the 1st question Friction in screening If it drops, screening is long/boring
Scheduling rate / next step Real funnel progress Bot must push for action, not for chat
No-show / dropout rate Pre-qualification quality Bot can confirm and reduce absences
Conversion by source Efficiency by channel Bot can standardize service by channel

If you want to measure automation ROI with a method (not just “I think”), read: How to measure the ROI of business automations.

Real costs you need to account for (no self-deception)

I'll be direct: the cost of Salesbot is not just “activating the bot.” You pay on three fronts:

  • Kommo license (per user): the CRM is the foundation of the operation.
  • WhatsApp Channel: if it is Official WhatsApp API, there is a cost per conversation/category and charges from the provider + Meta (this varies by country and category and changes over time).
  • Implementation: funnel design, SLA, fields, messages, tests, exceptions, and team training.

About Kommo's price and what people forget to add, I detail here: How much does Kommo really cost per month? Value per user + what no one adds to the bill.

Real (and important) data: Kommo offers 14-day trial (free trial) — this is useful to validate flow, but does not validate scale (real SLA, peaks, team quality). If you want to understand this trial and what can be properly tested, see: Does Kommo have a free trial? How the 14-day trial works and what you can validate before paying.

Who Kommo's Salesbot serves (and who it DOES NOT serve)

Let's cut the illusion.

It's for:

  • Operation with lead volume and response time becoming a bottleneck
  • Team that needs Standardization (SLA, routing, tags, stages)
  • Businesses that sell via WhatsApp and want Predictability in the funnel

Does NOT work for:

  • Those who want “a robot that closes deals alone” (that's fantasy)
  • Operation without process (no stages, no owner per stage, no passing rules)
  • Company that won't measure anything (if you don't measure, you just automate chaos)

Implementation in 7 days (lean) vs. 30 days (to scale without suffering)

I usually work with two mental models:

  • 7 days (MVP): welcome + minimal screening + routing + 1 follow-up.
  • 30 days (scale): full SLA + audit + cadences per stage + reactivation + reports and fine tuning.

The secret is: don't try to automate 100% on day 1. Build the basics, measure, optimize, scale. Method > improvisation.

Quick checklist: if you do this, you will stop “irritating leads” tomorrow

  • Reduce screening to 1–3 questions
  • Write short (WhatsApp is not email)
  • Leave an exit to a human at any time (“talk to an agent”)
  • Create response SLA and redistribution rule
  • Do follow-up with context, not with collection

If you want me to design and implement this in your operation (funnel, SLA, bot, routing, reports, and training) focused on results and ROI, the next step is simple: request a project.

FAQ — real questions about Kommo's Salesbot

Questions I answer most when the topic is Salesbot + WhatsApp + Kommo.

Frequently Asked Questions

Does Kommo's Salesbot replace a salesperson on WhatsApp?

No. It replaces repetitive tasks (screening, routing, simple follow-up). Negotiation, objection, and closing remain human if you want to maintain conversion and avoid friction.

How many questions should the Salesbot ask before passing to a human?

In practice, 1 to 3 questions. More than that increases friction and lowers response rate. The goal is to classify the lead and unlock the next step, not to “register.”

How to prevent the bot from irritating the lead on WhatsApp?

Short message, clear options (buttons), natural language, easy exit to human and handoff at the right moment (when price, urgency, or objection appears). And follow-up with context, not with collection.

Is it possible to measure Salesbot ROI in Kommo?

Yes. I measure by funnel impact: time to first response, % of leads who answer the first question, stage advancement (scheduling/quote), dropout rate, and conversion by source. If these metrics don't improve, the bot is just generating volume.

Is Kommo's 14-day trial enough to validate Salesbot and WhatsApp?

It is enough to validate basic flow (welcome, screening, routing, and follow-up). It is not enough to validate scale (peaks, real SLA, team maturity, and exceptions). For that, you need to run with process, metrics, and adjustments.

I want to implement this in my company → More articles →