CRM automation is a set of rules by which the system performs repetitive actions on leads and deals on its own: it assigns an owner, creates a task, checks fields, escalates overdue items. Don’t start with flashy scenarios. Start with frequent actions where a mistake is costly and the rule fits into one sentence. Decisions that require judgment - qualification, discounts, the first reply in a complex deal - are better left to people.
In Exceltic.dev projects implementing and refining Kommo and HubSpot, we rarely walk into an empty CRM. More often it is an account where rules have piled up for years: different people set them up for different tasks, and nobody remembers why half of them exist. Later we rewrite some of these scenarios as custom code - when a rule needs to remember state, retry a failed call or keep a log of its runs.
The approach below comes from those projects. We will cover how to choose your first rules, what not to hand over to automation, where each rule should live and how to check that it works.
The problem hits hardest for a head of sales in a team of 15 or more people. Leads are handed out manually in a group chat, reps reply after several hours, and many deals have no next step. Before every team meeting the pipeline report has to be cleaned up by hand: the CRM works as a logbook, not as a system that keeps a lead from getting lost.
CRM automation: setting up rules by which the CRM, when an event or condition occurs, performs an action on a lead, deal or task by itself, without a sales rep.
Why Your CRM Is Full of Data but Manual Work Has Not Gone Down
Rolling out a CRM and automating sales inside the CRM are two different projects, and the first one does not include the second by default. A rollout usually ends with the pipeline, the fields and the database import. Reps start entering data, but everything that happens around the data stays manual: who takes the lead, when to call back, who gets a won deal.
The second scenario is worse. Automations exist, dozens of them, but reps do not trust them. Tasks are created in batches and duplicate each other, notifications arrive for any reason, and the team learns to ignore them.
Meanwhile the sales leader keeps checking everything by hand. They do not know which rule fired and which one silently broke, so they cannot rely on any of them.
Both scenarios share one cause. Rules were added as complaints came in, not by priority, and none of them has an assigned owner. Below is a way to choose what to automate first so that every rule removes a noticeable amount of manual work.
How to Choose What to Automate First
A convenient way to set priority is to score three criteria, each on a scale from 1 to 3.
- Frequency. How often the action repeats: 1 - a few times a month, 2 - every week, 3 - every day or more often.
- Cost of error. What you lose if the action is forgotten or done late: 1 - inconvenience, 2 - team time, 3 - a lead or money.
- Rule clarity. Can it be written as “if X, then Y” with no exceptions: 1 - a person decides every time, 2 - there is a rule, but with caveats, 3 - the rule is unambiguous.
The three scores are multiplied, with a maximum of 27 points. Multiplication is deliberate: a 1 on any criterion sharply lowers the total, and such a scenario does not make it into the first wave.
| Action | Frequency | Cost of error | Rule clarity | Total |
|---|---|---|---|---|
| Assign an owner to a new lead | 3 | 3 | 3 | 27 |
| Escalate an unanswered lead | 3 | 3 | 3 | 27 |
| Create a task on stage change | 3 | 2 | 3 | 18 |
| Check required fields on a stage transition | 3 | 2 | 3 | 18 |
| Hand a won deal over to onboarding | 2 | 3 | 3 | 18 |
| Merge duplicate contacts | 3 | 2 | 2 | 12 |
| Remind about a follow-up | 3 | 2 | 2 | 12 |
| Decide whether a lead is a fit | 3 | 3 | 1 | 9 |
| Approve a discount | 1 | 3 | 1 | 3 |
In our practice the thresholds are: 18 points and above - first wave, 8 to 12 - second wave. If rule clarity is scored 1, the scenario is not automated whatever the total: the system can suggest, but the decision stays with a person.
Your team’s scores may differ. In B2B with a long sales cycle, follow-ups weigh more than in high-volume sales. What matters is not the exact number, but that priority is discussed on one scale rather than by how loud the complaints are.
Seven Scenarios Worth Starting With
Below are the scenarios that consistently score high in Exceltic.dev projects. The first five go into the first wave, the last two into the second.
Automatic Lead Distribution
A new lead should get an owner the moment it is created, not when the sales leader checks the group chat. The rule is built on source, language, region or product, and within a group leads are handed out in turn or based on current workload.
The main trap is vacations and sick leave. A queue that does not know a rep is away gives leads to someone who will not see them for a week. That is why the rule needs an availability check and a fallback recipient.
First Response Control With Escalation
In our projects, first response time is more visibly linked than other metrics to whether a lead gets to a conversation. It is also the easiest one to control automatically. An example rule: no outgoing call or message after 15 minutes of working time - a reminder to the rep; after 30 minutes - to the team lead; after an hour the lead is reassigned.
Thresholds are tuned per channel. For inbound chat 15 minutes is a lot; for a request about a corporate plan it is fine.
Tasks on Stage Change
Every pipeline stage should leave behind a next task with an owner and a due date. A deal moves to “Proposal sent” - a task “Call about the proposal” is created, due in two business days.
This closes the most common gap - deals with no next step that quietly go stale in the middle of the pipeline. The stage logic has to be thought through in advance, though: otherwise the tasks will lock in chaos rather than order.
Required Fields on Stage Transitions
Required fields stop a deal from moving forward until the data the next stage or a report needs is filled in. Without an amount and a close date a deal does not move to negotiation; without a loss reason it does not move to lost.
Require only what is actually used. Reps will fill ten required fields on one transition with dots and dashes, and the data will end up dirtier than without the check.
Handing a Won Deal Over to Onboarding and Billing
The moment a deal is won is where context gets lost most often. The rep switches to new clients, and the implementation team learns about the deal from an email three days later.
A rule on the move to won creates a task for the implementation team, carries over the key fields and sends a notification to that team’s channel, not to direct messages. If onboarding and billing live in another system, the data handoff is already an integration matter, not a rule inside the CRM.
Duplicate Control
One client messages you, submits a form on your website and calls, and the CRM creates three records for three different reps. The client gets called twice, and in reports one lead turns into three.
Built-in mechanisms cover only part of the cases. For example, according to the documentation as of Q4 2026, duplicate control in Kommo works for leads from communication channels and does not check imported leads automatically.
Merging happens only if one of the duplicates is in the Incoming leads stage, and third-party integrations are not always supported. That is why we usually build a separate check by phone and email for the remaining sources, before the record is created.
Follow-up Reminders
The follow-up after a sent proposal or a pause in the conversation is what reps forget most often. The rule “no activity on a deal in negotiation for five business days - a task for the owner” works better than a chain of automated emails sent in the rep’s name.
Automated emails fit simple sales. In a B2B deal the client quickly recognizes the template, and in our experience that hurts trust more than a pause does.
What Not to Hand Over to Automation
Automate actions, not decisions: if a rule cannot be written without the words “usually” and “depending on the situation”, it is a judgment call, and it is better left to a person.
Qualification decisions. The system can collect data for lead scoring - company size, source, form answers - and highlight good fits. But if it discards leads as unqualified on its own, the mistake is invisible: nobody re-checks a discarded lead. A sensible compromise is automatic scoring as a hint and a manual decision with a required rejection reason.
Discounts and non-standard terms. A discount depends on context that is not in CRM fields: the client’s strategic value, competitors in the deal, the history of the relationship. An automatic “minus 10% when the amount is above X” quickly becomes a condition every client expects. What is worth automating here is the approval route - who confirms and by when - not the amount itself.
The first personal reply in a complex deal. In high-volume sales an auto-reply in chat saves time for both sides. But if the head of a 200-person company comes to you with a question about implementation, a template “Your request has been received” tells them they have landed on a conveyor belt. Here automation should speed up the person: instantly assign an owner and show them the context, while the first message should come from them.
Built-in Triggers, Connectors or Custom Code
Each rule can be implemented in one of three places, and this choice determines how easy it will be to find, check and fix later.
Built-in CRM tools are the first choice for rules within one system. In Kommo this is the Digital Pipeline (Leads > Automate).
Triggers are attached to each stage: create a task, change the responsible user or the stage, send an email or a webhook, launch a Salesbot. The full list is in Kommo’s documentation on Digital Pipeline triggers.
According to Kommo’s pricing table as of Q4 2026, Salesbot, rule-based automation, required fields and webhooks are available starting from the Advanced plan; Base does not have them. Additional triggers on tags and field changes are moved into a separate Add-on trigger pack, which is marked only for Pro and Enterprise. If your sales team works in Kommo and rules have piled up for years, we start Kommo implementation and customization with an audit of what is already set up, not with new triggers.
In HubSpot the equivalent tool is workflows, under Automation > Workflows. According to the HubSpot Knowledge Base as of Q4 2026, they are available on the Professional and Enterprise tiers of Marketing Hub, Sales Hub, Service Hub, Data Hub, Smart CRM and Revenue Hub.
For distribution there is the Rotate record to owner action in Sales Hub and Service Hub Professional and above. It hands out records in turn (Round Robin), by workload (Load Balanced) or randomly, and for leads and deals it is marked as beta.
Custom code actions in JavaScript or Python (Python in beta) inside a HubSpot workflow are available in Data Hub Professional and Enterprise. Each run of such code has to finish within 20 seconds. For teams on HubSpot the order is the same: during HubSpot implementation we first review the existing workflows and only then add new ones.
No-code connectors - Zapier, Make, n8n - are needed when a rule touches another system: create an invoice, write a row to a spreadsheet, send an event to the team messenger. Their downside is that the logic leaves the CRM: the admin does not see that the deal is being changed from outside, and when something breaks they look for the cause in the wrong place. We compared what to choose for Kommo in Zapier vs Make vs n8n.
Custom code is justified when a rule needs state, retries and a log: a distribution queue that accounts for absences, escalation based on a working calendar, a duplicate check across several systems at once. Ongoing data sync between systems is a separate architectural task; we covered it in our article on CRM integration.
The short selection rule: one system and a rule without memory - built-in tools; two systems and simple logic - a connector; state, retries or many systems - code.
Why CRM Automation Stops Working Over Time
Automation rarely breaks loudly. More often it degrades quietly, and the team notices through indirect signs - leads get lost again, reports stop adding up again.
- The rule has no owner. It was set up by an admin who left a year ago, and nobody knows whether it can be turned off. Such rules live forever and get in the way of new ones.
- Rules conflict. One assigns an owner by region, another by product, and the result depends on which one fired last. Reps see leads “jumping” and stop trusting distribution.
- A flood of tasks. Each rule creates its own task, and by lunchtime a rep has 40 automated reminders. They learn to close them in bulk without reading, and miss what matters along with the noise.
- Loops. Rule A changes a field, the field change triggers rule B, and it moves the deal to a stage where A fires again. The deal bounces between stages, and the activity feed fills up with entries.
- Silent breakage after changes. Someone deleted a stage, replaced a value in a list or created a new field instead of the old one, while the condition referenced the previous one. There is no error, the rule just stops firing, and you find out a month later.
- Notification fatigue. All notifications go to one shared channel or personally to the sales leader. A month later the channel is muted, and the escalation about an unanswered lead drowns along with everything else.
How to Audit Your Automations Once a Quarter
A short audit once a quarter takes a few hours and removes most of the problems above.
- Inventory. Put all rules into one table: trigger, condition, action, where it is configured (CRM, connector, code), date of the last change.
- An owner for every rule. This is a specific person who knows why the rule is needed and decides whether to change it. A rule without an owner is turned off two weeks after a warning.
- Metrics. Watch four indicators: median first response time, the share of open deals with no next task, the share of overdue automated tasks and the number of rules with zero runs in 90 days.
- Cleanup. Dead rules are deleted, similar ones are merged, and notifications are split across channels on the principle “one event - one recipient”.
What This Looks Like in a Typical Project
A typical Exceltic.dev project to bring order to automation looks like this: a sales team of 8-12 reps, 500-1,000 new leads a month and a CRM with two to three years of history.
Before the work starts, the picture is usually similar. The account has 40-70 rules, about a third of which have not fired in the last quarter. Median first response time is two to five business hours, 25-40% of open deals have no next task, and the sales leader hands out leads manually at the start of the day.
The work takes 4-6 weeks. First comes the inventory and turning off dead rules, then the first wave: distribution that accounts for absences, first response escalation, tasks on stages and required fields on three key transitions. We usually move distribution and escalation into a custom webhook handler; the rest stays in the CRM’s built-in tools.
6-8 weeks after launch, in projects like these we see the following ranges.
| Metric | Before | After |
|---|---|---|
| Active rules | 40-70 | 15-25 |
| Median first response time | 2-5 business hours | 10-30 minutes |
| Open deals with no next task | 25-40% | under 5% |
| Leads distributed manually | the majority | rare exceptions |
Exact numbers depend on the channel and the team’s discipline, but the direction is consistent. The main effect is not even speed: the sales leader stops checking the pipeline by hand every morning.
Who Needs This Approach Most
The approach is designed for companies of 15+ people where the sales team already works in Kommo or HubSpot, but the sales leader still hands out leads, reminds people about calls and cleans up data before reports.
It is especially useful if the business processes in your CRM were set up by several admins at different times and nobody can say which of them work. If the pipeline is not documented, though, start with the process: sales pipeline automation without clear stages only speeds up chaos.
Frequently asked questions
Which automation rules should you set up first?
Start with an inventory: which rules already exist, who set them up and whether they fire. Then score the candidates on three criteria - action frequency, cost of error and rule clarity - and put the top scorers into the first wave. In most sales teams that means distributing new leads, first response control with escalation and tasks on stage change. CRM automation in this order delivers a noticeable effect in the first weeks and does not create a flood of tasks that reps start to ignore.
Which Kommo or HubSpot plan do you need for automation?
As of Q4 2026, in Kommo Salesbot, rule-based automation, required fields and webhooks are available from the Advanced plan; Base does not include them. In HubSpot, workflows are included in the Professional and Enterprise tiers of Marketing Hub, Sales Hub, Service Hub, Data Hub, Smart CRM and Revenue Hub. Round-robin distribution through Rotate record to owner requires Sales Hub or Service Hub Professional or Enterprise. Custom code actions in workflows are available in Data Hub Professional or Enterprise. Check the current pricing page before you buy: both vendors change what is included in their plans.
Do you need Zapier or Make if your CRM has its own triggers?
For rules inside one CRM, no. Built-in triggers are easier to find, check and maintain, because the logic lives where the data is. Connectors are needed when a rule touches another system: accounting, spreadsheets, the team messenger, an onboarding service. If a rule needs to remember state - whose turn is next, how many times to retry a failed call - a custom webhook handler is more reliable. You can test it and keep a log of its runs.
How many automation rules is normal for a sales team?
There is no universal number. In Exceltic.dev projects, working automation for a team of 8-12 reps usually fits into 15-25 rules. If there are more than 50 rules, some of them most likely duplicate each other, conflict or do not fire at all. What matters more than the count: every rule should have an owner, a clear goal and a metric that shows it is working. A rule that has not fired once in a quarter should be turned off.
Key Takeaways
- Start with an audit: turn off rules that have not fired in a quarter and assign an owner to the rest.
- Choose your first scenarios by frequency, cost of error and rule clarity. The first wave usually includes lead distribution, first response escalation and tasks on stages.
- Automate actions, not decisions: qualification, discounts and the first reply in a complex deal stay with people.
- Once a quarter, check four metrics: first response time, deals with no task, overdue automated tasks and rules with no runs.
If your CRM has dozens of rules but reps still hand out leads and remind each other about calls by hand, describe your situation to the Exceltic.dev team. We will work out which rules work, what is worth turning off and what to move into code, and estimate the scope of work.