AI CRM
CRM automation for sales teams: ten workflows worth building
By Viktorio Shopov, co-founder of Sellinger. Reviewed by Asen Levov. Published . Last updated . 8 minute read.
Ten CRM workflows are worth building: capture and de duplicate, assign an owner, send the first touch, follow up on silence, stop on a reply, honour an opt out, alert a person to a warm reply, log a booked meeting, sync with your main CRM, and audit stale leads. Each is a trigger and an action. Automate the repetitive, time sensitive and rule based ones, and keep a person on exceptions.
CRM automation sounds big. In practice it is a short list of if this, then that rules. The skill is choosing which rules, writing them so they cannot do harm, and checking them. Here are ten, in the order we would build them.
What are the ten workflows?
1. Capture and de duplicate
Trigger: a lead arrives from an import, a form, an ad or a signal. Action: record the source, match on a stable identifier and merge instead of creating a second record.
Risk: bad matching merges two different people. Match on the LinkedIn profile or the email address, not on a name. See duplicate leads: merge and prevent.
In Sellinger, Facebook lead form leads are polled every 60 seconds and Google Ads leads every 300 seconds, then created as leads and contacted by the campaign.
2. Assign an owner and a campaign
Trigger: a new lead passes the fit check. Action: set the owner, the campaign and the first channel.
Risk: leads with no owner. Make the rule total: every lead gets an owner or a reason it does not.
3. Send the first touch
Trigger: the lead is assigned and the account has capacity. Action: send the first message within the daily limits and the sending schedule.
Risk: a burst that looks like abuse. Sellinger's defaults are 20 LinkedIn invitations a day and 150 a week, and email defaults to 50 sends a day per mailbox. The agent cannot exceed the limit set per account.
4. Follow up on silence
Trigger: no reply after a set number of days. Action: send the next follow up, with a new angle.
Defaults: LinkedIn 4 follow ups, then email; email 5, then WhatsApp; gaps from 5 days on LinkedIn and 1 to 3 days on email. Every value can be changed per campaign. Our data suggests using fewer than the maximum: see how many follow ups.
5. Stop on a reply
Trigger: a reply on any channel. Action: stop the sequence on the other channels.
Risk: an email arriving after the person answered on LinkedIn. This is the rule that most protects the prospect.
6. Honour an opt out
Trigger: a reply such as STOP or unsubscribe. Action: put the contact on the suppression list, stop the agent and close the lead.
A weak or unclear phrase should not write the permanent record: stop and let a person decide. Before every send on email, LinkedIn and WhatsApp the platform checks the list. See opt outs and suppression.
7. Alert a person to a warm reply
Trigger: a positive reply, a question about price, or a request for a call. Action: notify the owner in Slack or email and pause the agent on that lead if you want a person to answer.
Risk: noise. Alert only on the replies a person must act on. Sellinger's agent stops on its own when it cannot answer, for example a price it has no answer for.
8. Log a booked meeting
Trigger: a slot is booked on a connected Cal.com or Calendly calendar. Action: move the lead to a meeting stage and record held or no show after the call.
Our own numbers show why to record held: 351 showcase meetings were booked and 33 are confirmed held. See meeting booking with Cal.com and Calendly.
9. Sync with your main CRM
Trigger: a lead changes stage or a conversation is added. Action: write it to HubSpot, or to another system through the API and webhooks.
Sellinger has a two way sync with HubSpot, and a REST API with webhooks from the Full Stack bundle up. Salesforce, Pipedrive and Zoho are coming soon, not live. The HubSpot connection guide has the steps.
10. Audit stale leads
Trigger: weekly. Action: list leads with no activity for a set time and decide: revive, change channel, or close.
If you use Claude with Sellinger's MCP connector, the audit stale leads recipe does this in one prompt. Do not delete the leads: contacts stay in the database, and you choose their status.
Which workflows come first?
Build in this order: 1, 5, 6, then 3 and 4, then the rest. The first three protect the prospect and your account. The next two create the volume. The last five make the process easier to run.
| Workflow | Time saved | Risk if wrong |
|---|---|---|
| Capture and de duplicate | High | Merged or doubled leads |
| Stop on a reply | Medium | A message after a reply |
| Honour an opt out | Low | A complaint or a legal problem |
| Follow up on silence | High | Fatigue, refusals |
| Alert on a warm reply | Medium | Alert fatigue |
How do you test a workflow?
Run it on 20 leads in approval mode. Read what it would have done. Check three cases: a normal lead, a lead that replied, and a lead that opted out. Only then turn it on for the list.
Keep a log of changes, one at a time. When a number moves, you will know why.
What should you not automate?
A fit decision on thin data, a sensitive reply, a price beyond the published plans, anything with a legal consequence. On our own account 93% of replied conversations since May 2026 ran with no human message, but only after a long period of rules and reading threads. The exceptions still went to a person.
How do you keep automation honest over time?
Automations drift. A rule written in January fires on a list that no longer looks like the one it was written for. Three habits keep that in check.
- A monthly read of the log. Pick ten leads that the automation touched and follow each one from trigger to outcome. If you would not have done it by hand, change the rule.
- A kill switch you have tested. Know where to pause an agent or a campaign, and try it before you need it. A person can take over any conversation in Sellinger, and a campaign can run in approval mode.
- One change at a time. Config optimization in Sellinger proposes changes to a campaign, up to 3 per cycle by default, with auto approve off, for the same reason: changes you can see are changes you can undo.
The numbers worth watching after any change are meetings held per person messaged, not only replies. On our own account replies stayed near 40% while meetings drifted below 1% from June 2026, which no single automation caused and no dashboard of activity would have caught.
A checklist
- [ ] Capture records the source and matches on a stable identifier.
- [ ] A reply on any channel stops the others.
- [ ] An opt out closes the lead and writes to the suppression list.
- [ ] Limits and schedules are set per account.
- [ ] A person is alerted only to replies that need one.
- [ ] I record booked, held and no show.
- [ ] I test every workflow on 20 leads first.
Common questions
What is CRM automation?
Which workflow saves the most time?
Can I build these with webhooks?
Is automating a CRM risky?
Do I need Salesforce for this?
How Sellinger helps
Read how it works on See the agentic CRM, or see a real campaign in Read the Sellinger case study.
