Pipedrive and CodeWords: keeping the pipeline honest
Pipelines that reflect optimism rather than reality, activities as the leading indicator, custom fields as an interface, and automating around a CRM salespeople will actually update.
On this page
- What we'll cover
- Find out what already runs
- The pipeline is an opinion
- What the Pipedrive API reaches
- Connecting it to CodeWords
- Seven automations worth building
- Activities, the leading indicator
- Custom fields, which are an interface
- Handoff at the point of winning
- Building it so it survives
- Limits worth knowing about
- What to build first
- Frequently asked questions
- Related reading
Pipedrive's design premise is that a salesperson should be able to see what to do next, which is why it is organised around activities rather than records. Teams choose it because it is simple enough that people actually use it.
That premise should govern what you automate around it. The risk with CRM automation is always the same — you make the system require more, the people using it comply less, and the data gets worse. The automation worth building removes entry work and reports on what the pipeline is really doing, rather than enforcing process on people who will route around it.
What we'll cover
- Find out what already runs
- The pipeline is an opinion
- What the Pipedrive API reaches
- Connecting it to CodeWords
- Seven automations worth building
- Activities, the leading indicator
- Custom fields, which are an interface
- Handoff at the point of winning
- Building it so it survives
- Limits worth knowing about
- What to build first
- Frequently asked questions
Find out what already runs
Pipedrive has its own automation, and it is usually already doing some of what you are about to build.
Workflow automations create activities, move deals, and send emails on triggers.
Required fields per stage enforce data entry at stage transitions.
Email sync attaches correspondence to deals automatically for users who enabled it.
Installed apps write to deals and contacts, sometimes to the same fields you intend to use.
List what exists before adding anything. The characteristic failure is an external automation setting a field that a built-in workflow also sets, producing a value that depends on ordering and a bug nobody can reproduce.
If a native workflow can do the job, use it. External automation earns its place when the logic needs data Pipedrive does not hold.
The pipeline is an opinion
Any pipeline report is a summary of what salespeople believe, entered at moments when they were optimistic. A few automations make it closer to a description of reality.
Stage age is the honest metric. A deal in negotiation for ninety days is not in negotiation.
Expected close dates in the past are the clearest signal a deal has stalled, and they accumulate silently.
Deals with no activity scheduled are, in Pipedrive's own terms, not being worked.
Value that has never changed since creation often means the number was a placeholder.
Win rates by stage entered, rather than by deal created, tell you where deals actually die.
Reporting these is more useful than any forecast adjustment, and it is information the sales manager wants and cannot easily assemble.
What the Pipedrive API reaches
Deals can be created, read, updated, and searched, with their stage, value, and custom fields.
Persons and organisations are the contact records, with their own custom fields and relationships.
Activities — calls, meetings, tasks — can be created and read, which is how automation participates in the daily workflow rather than sitting beside it.
Notes can be attached to any record, which is where automated context belongs.
Pipelines and stages can be read, which anything setting a stage needs since identifiers differ per pipeline.
Files can be attached to deals.
Webhooks fire on record changes.
Products can be attached to deals where you use them, which matters for anything touching what was actually sold.
Connecting it to CodeWords
CodeWords connects to more than 3,000 integrations, and the connection is made once and reused.
- Open CodeWords and start a new automation.
- Describe what should happen in plain language to Cody, the automation builder: which records, what triggers it, and what should result.
- Authorize the connection with an API token belonging to a dedicated integration user rather than a salesperson.
- Describe the exceptions: a duplicate person, an organisation that cannot be matched, a stage that does not exist in that pipeline.
- Test against one pipeline before anything runs across the account.
You describe the outcome; Cody builds it, connects it, and deploys it. The free plan covers light use, with Pro at $39 per month and Business at $100 per month as usage grows; details are on the pricing page.
Seven automations worth building
Inbound lead creation with deduplication. A form, a demo request, or a trial signup becomes a person, an organisation, and a deal, checked against what already exists.
Enrichment from your own systems. Trial usage, support history, and billing status written onto the record, so the salesperson sees what is true without asking anybody.
Stalled deal reporting. Deals with no activity, past their close date, or aged in stage, reported to the owner rather than to a channel.
Activity creation on real triggers. A trial ending, a usage threshold crossed, a support escalation — each creating a task for the owner with the reason attached.
Post-close handoff. When a deal is won, create whatever the delivery or onboarding process needs, with the context from the deal carried across.
Data quality reporting. Missing fields that matter, deals with no contact, organisations duplicated, values that look like placeholders.
Pipeline reporting that includes the honest metrics. Stage age, activity coverage, and win rate by stage entered, delivered weekly.
Activities, the leading indicator
Pipedrive organises around activities for a reason, and it makes them the most useful thing to automate around.
A deal with no scheduled activity is not being worked. That is the cleanest health metric available and it needs no interpretation.
Create activities rather than sending notifications. A task in Pipedrive appears in the salesperson's normal workflow; a message in a chat channel does not.
Attach the reason. An activity saying "follow up" is ignored; one saying "trial ends Friday and they have not invited anybody" is acted on.
Do not flood. An automation creating several activities a day per salesperson gets its activities ignored along with everything else. Be selective about what is worth interrupting for.
Report coverage, not completion. What share of open deals have a next step scheduled is a better management number than how many activities were ticked off.
Custom fields, which are an interface
Anything automation reads or writes becomes a dependency, and Pipedrive's custom fields are edited casually.
Reference fields by their identifier, not their label, since renaming a field breaks anything matching on the name.
Use the right type. A dropdown for fixed values, a date for dates. Text fields holding dates in four formats cannot be reported on by anybody.
Name the automation's own fields distinctly, so it is obvious to a person which fields are written by a machine.
Record the source identifier from whatever system created the record, which is what makes automation idempotent.
Read field definitions rather than hard-coding options, since dropdown values change and writing one that no longer exists fails quietly.
Handoff at the point of winning
The most valuable moment in a CRM is the one that gets the least attention: a deal closing and the customer becoming somebody else's responsibility.
Carry the context, not just the record. What was promised, what the concerns were, what timeline was discussed. That information is in the deal's notes and emails and it does not travel on its own.
Create the onboarding work where onboarding happens, in whatever system that team uses, rather than expecting them to read a CRM they do not open.
Record what was actually sold, including products and any non-standard terms, since this is where delivery and billing disagreements begin.
Introduce the people. An automated handoff that does not tell the customer who they are now dealing with leaves them wondering whether they have been abandoned.
Close the loop back to the salesperson when onboarding completes or stalls, since they will hear about it from the customer otherwise.
The handoff is the point where a good sales process most often produces a poor customer experience, and almost all of it is information transfer that nobody enjoys doing by hand.
Building it so it survives
Search before creating. Persons and organisations duplicate rapidly otherwise, and duplicates in a small CRM are noticed by customers.
Resolve stage identifiers per pipeline, since they differ and an automation that works on one pipeline fails on the next.
Handle at-least-once webhooks by keying on the record and change rather than reacting blindly.
Prefer notes and activities over silent edits. An automated note is visible; a changed field appears to be somebody's own work.
Report the outcome. Records created, updated, and flagged as possible duplicates.
Limits worth knowing about
Rate limits apply and depend on your plan, with a token bucket that bulk work can exhaust quickly.
Search is not immediately consistent, so a record created and searched for immediately may not be found — use the identifier returned instead.
Custom field identifiers are opaque strings, which is inconvenient to read and is what makes them stable.
Pagination applies to listings, and a report built from the first page understates everything.
Some features are plan-dependent, including parts of the automation and reporting capabilities, so confirm before designing around one.
What to build first
Stalled deal reporting: deals with no scheduled activity, past their expected close date, or aged unusually long in stage, sent weekly to each owner rather than to a shared channel. It writes nothing, it takes little building, and it surfaces the deals that are quietly dead while the forecast still counts them.
Two habits make the difference. Send each owner their own list, since a combined list belongs to nobody. And lead with the count, because "you have six deals with no next step" is read and a table of forty rows is not.
Frequently asked questions
Will automation make salespeople update the CRM more?
Only if it removes work. Automation that enriches records and creates tasks helps; automation that adds required fields and enforcement gets routed around, and the data gets worse rather than better.
Why does a field keep reverting?
Something else is writing it — usually a built-in workflow or an installed app. List what writes to that field, decide which one owns it, and make everything else read rather than write.
How do I stop duplicate contacts?
Search before creating, on normalised email and on name-plus-organisation, and store the source identifier on the record. Automations that create without looking are the main source of duplicates in any CRM.
What is the best measure of pipeline health?
The share of open deals with a next activity scheduled. It is simple, it is hard to game without doing the actual work, and it predicts far better than the stage distribution does.
Should automation move deals between stages?
Rarely. A stage reflects a salesperson's judgement about a relationship, and automated movement produces a tidy board that misrepresents reality. Report that a deal looks stalled and let the owner decide.
Why does setting a stage fail on one pipeline?
Stage identifiers are per pipeline. Resolve the stage for the specific pipeline rather than reusing an identifier, which is the usual cause of an automation that worked in testing and fails on the second pipeline.
What enrichment is most valuable?
Whatever the salesperson currently asks somebody else for — trial usage, open tickets, payment status. It already exists in your systems, it is invisible in the CRM, and putting it on the record removes a recurring interruption for both people.
What should happen when a deal is won?
Create the onboarding work where that team actually works, carry the promises and concerns from the deal's notes across, record what was sold including any non-standard terms, and introduce the new contact to the customer. Most of it is information transfer nobody enjoys doing by hand.
Should the salesperson hear about onboarding problems?
Yes, automatically. They will hear about it from the customer otherwise, without context and usually at an awkward moment. Closing the loop back is a short automation and it protects the relationship somebody spent months building.