AI systems4 min readAugust 2026
Build a CRM pipeline that matches how you actually sell
Template CRM stages produce forecasts nobody believes. How to map the stages your jobs actually move through and keep the pipeline honest.
Open your CRM and read the stage names out loud. If nobody in the office can say what qualified means without hedging, your forecast is a guess wearing a nice interface. This is worth fixing before you spend another dollar on software, because every report, reminder, and automation downstream inherits whatever the stages really mean.
Template stages describe someone else's business
Most CRMs ship with stages borrowed from software sales. Discovery, evaluation, negotiation, procurement. A trade business does not sell that way. An HVAC replacement moves through a call, a site visit, a quote, a follow up or three, and a yes or a no. Forcing that path into a template built for enterprise deals produces stages that mean nothing, and stages that mean nothing produce numbers nobody trusts.
The tell is a forecast meeting where every number needs a verbal footnote. That one is basically closed. Those two are dead but I have not moved them yet. Once the stages need translation, the CRM has stopped being a source of truth and become a chore someone performs before meetings.
Map the path your jobs actually take
Pull up your last twenty closed jobs, won and lost, and write down what happened to each one. Where it came from, what the steps were, where it stalled, how it ended. Patterns show up fast. Most service businesses land on something close to:
- 01New inquiry
- 02Contacted
- 03Visit scheduled
- 04Visit done
- 05Quote sent
- 06In follow up
- 07Won or lost
Name the stages in your own words. If your team says estimate instead of proposal, the stage is called estimate sent. And keep the count low: five to seven working stages is plenty. Every extra stage adds another judgment call about where a job belongs, and judgment calls are where pipelines rot.
Every stage needs an exit condition and an owner
A stage earns its place when you can say exactly what moves a job out of it. Quote sent is a good stage because the exit is an event: the customer said yes, said no, or went quiet past your follow up window. Qualified is a bad stage until you define it, for example: we confirmed the service area, the scope, and that they want the work done this season. Write the exit condition into the stage description where everyone can see it.
Each stage also needs a normal amount of time. If a quote usually gets an answer within a week, a quote sitting for three weeks is a flag, and your CRM should raise it without anyone remembering to check.
Record why jobs are lost
When a job dies, log the reason in plain terms:
- Price
- Timing
- Went with someone else
- Never answered
- Out of area
Six months of honest loss reasons will tell you more about your pricing and your follow up than any dashboard. Skip the twelve option dropdown. Five reasons your team will actually use beat a taxonomy nobody fills in.
Add automation only after the stages are honest
Once the stages mean something, automation attaches cleanly. A task fires when a quote goes out. A reminder fires when a job sits past its normal stage time. An AI summary of the last call attaches to the record when the job changes stage, so whoever picks it up next has context without doing archaeology. AI is useful here as a drafter and a summarizer. Stage decisions stay with whoever talked to the customer.
Automating a pipeline where stages mean nothing just moves fiction around on a schedule, and now the fiction sends reminders.
Ten minutes a week keeps it alive
A pipeline stays honest because someone looks at it. Pick a fixed ten minutes each week and ask three questions:
- What moved
- What stalled past its normal time
- What has no next step with a date on it
If you want the reporting side built for you, the CRM and reporting page at growthkali.com/services/crm-reporting covers what we install and what the weekly report looks like.
