Pipedrive Nova:
Now, after a:
Sales Meeting:
it can directly suggest:
CRM:
updates.
For example:
Deal Stage.
Deal Value.
Expected Close Date.
Contact Details.
And even:
Custom Fields.
This is very convenient.
However:
The worst approach is to approve everything just because it looks reasonable.
Today’s lesson is:
A very simple:
30-second check.
First, review three things
Customer’s exact words.
↓
Current CRM value.
↓
AI suggestion.
In this order:
Don’t reverse it.
Why must you check the "customer’s exact words" first?
Because the easiest thing for AI is:
To take:
A vague statement from a person.
And organize it into:
A very clean, structured field.
For example, if a customer says:
"If the budget is finally approved, we hope to close by the end of the month, roughly around 300,000."
A human can understand that:
This contains a lot of uncertainty:
"If the budget is finally approved."
"Hope to close by the end of the month."
"Roughly."
"Around."
But CRM prefers:
Clean entries for:
Value.
Date.
Stage.
At this point:
AI might naturally want to:
Convert fuzzy information
into
structured data.
The problem isn’t that AI is making things up
It’s that:
The conversation
and
CRM fields
are inherently different.
Conversations allow for:
Possibilities.
Approximations.
Hopes.
Verify and double-check.
CRM, on the other hand:
requires you to
enter a definite value.
So the first step:
is not to see if the AI suggestion looks accurate,
but to go back and check:
What did the customer actually say?
For example, if Nova suggests changing the Deal Value to 300,000
Don’t approve immediately.
Instead, ask:
Did the customer say:
"The official quote is 300,000"?
Or:
"The budget is roughly 300,000"?
These two sentences imply very different things.
The first suggests:
A confirmed Deal Value.
The second might mean:
Just a Budget Range.
If your company’s:
CRM Deal Value
is defined as:
The official estimated deal amount,
then whether you can update depends on:
Your company’s rules.
Step two: Check current CRM value
Assuming:
Current Deal Value:
Was:
250,000.
And Nova:
Suggests:
300,000.
Don’t just ask:
"Is 300,000 correct?"
Also ask:
"Why was 250,000 set before?"
Maybe it was:
The last official proposal.
Maybe:
A sales rep’s initial estimate.
Or maybe:
The budget the customer confirmed before.
Or even:
Outdated data from three weeks ago.
Current value isn’t always correct
But it provides:
Context.
If a customer just casually said,
"Budget is roughly 300,000,"
and CRM currently shows:
250,000,
which is:
Yesterday’s official quote,
you shouldn’t
overwrite the official field
just because AI heard "300,000."
Pipedrive has already built this comparison feature into Nova
Nova displays:
Current Pipedrive value
side-by-side with:
AI suggested value
so you can
compare directly.
It’s not:
AI saying,
"Change to 300,000,"
and you must automatically:
Accept.
You can look at:
Before
and
After,
then decide whether to:
Approve.
Edit.
Or
Skip.
Even more importantly: If the CRM has been updated by someone else after Nova’s suggestion
Pipedrive also:
flags when
a field:
has changed.
This detail is crucial
because in teams, it’s common for:
A meeting ends in the morning,
Nova generates suggestions,
But in the afternoon, another sales rep or manager
updates the CRM,
and then you review Nova’s suggestions in the evening.
If there’s no warning,
you might accidentally
overwrite newer info with old AI suggestions.
So step two is not just about "what Nova wants to change"
It’s about seeing:
What the official current record really is.
This is:
The Current Value.
Step three: Then check AI suggestions
Only now do you ask:
What does Nova want to change?
Then compare it with:
The customer’s exact words,
and the current CRM value.
There are usually three possible outcomes:
Approve.
When it’s clear and ready to be recorded.
Edit.
The direction is right, but
numbers, dates, or
field content
need adjustments.
Skip.
The conversation hasn’t provided enough info yet
to update the official record.
Deal Stage is most often misjudged
For example, a customer says:
"The content looks fine, but I still need to check with the CFO."
AI might think:
The deal
has moved to the next stage.
But your company’s pipeline rules might state:
Only after
the Decision Maker
formally approves
can you move to the next stage.
So,
"content looks fine"
isn't enough.
Stage doesn’t reflect moods
A customer’s enthusiasm
is not a stage.
A customer saying:
"I’m interested"
is not a stage either.
True stages should correspond to your company’s
verifiable criteria, such as:
Needs confirmed.
Official proposal sent.
Decision maker involved.
Budget approved.
Contract signed.
Each company’s stages differ.
You can write field rules directly into Nova
Pipedrive lets users:
set specific instructions
for various CRM fields.
For example, tell Nova how to judge:
Expected Close Date,
how to interpret currency,
and which phrases
are enough
to update a certain field.
This feature is very useful.
Don’t let each sales rep guess "what counts as a deal stage" on their own
If you have five salespeople,
each may define stages differently.
AI also struggles
to maintain consistency for you.
The best practice isn’t to
ask Nova to
"be more accurate,"
but to first
clearly define your company’s rules.
For example, define Expected Close Date like this
Don’t just enter:
"Next month,"
as the last day of next month.
Only update it when
the customer clearly specifies:
a target date,
or when your company has a formal project timeline
with a clear estimated finish date.
Otherwise,
keep the current value.
This isn’t a Pipedrive rule,
but a rule you design based on your sales process.
Why is it especially important not to alter Expected Close Date casually?
Because it’s more than
just a note.
In Pipedrive’s
Forecast View,
Expected Close Date
is used to schedule
Revenue Projections.
If AI turns
"Hope to finish in October"
into
a definite date,
management’s forecast
might shift accordingly.
So even a small AI update
can ultimately affect
how the entire pipeline
is understood.
That’s why
high-impact fields
should not be reviewed in just half a second.
Deal Value is similar
When customers talk about:
300,000,
it doesn’t necessarily mean
the Deal Value is exactly 300,000.
It might be:
Budget.
Ceiling.
Part of a package.
Or exclude
another service.
Without internal rules,
AI can’t know
what you really want a field to represent.
So this 3-step method really asks: "Is the evidence strong enough?"
First:
Customer’s exact words.
What did they really say?
Second:
Current CRM value.
What is the current official record?
Third:
AI suggestion.
Why should it change?
If after seeing all three
it’s still unclear
why the field should update now,
then don’t change it yet.
One more important Nova limitation: CRM Update Suggestions are mainly based on Meeting Transcripts
Pipedrive explains:
Nova’s CRM Update Suggestions
are based on:
the meeting transcript.
Emails synced into
Pipedrive
are NOT used as a basis
for these CRM update suggestions.
This is a limitation often overlooked.
For example, at a meeting the customer says: "We are considering 300,000"
But two hours later,
they send an email
formally confirming
the purchase is only 220,000.
If you only review Nova later in the evening,
don’t assume that
because that email synced into Pipedrive,
Nova included it
in the CRM update suggestion.
It didn’t.
So even if AI has data, it doesn’t mean it uses all data to inform reasoning
This is a crucial concept across AI workflows.
Pipedrive may have access to:
CRM,
Email,
Calendar,
Meeting information,
but different features use
different data sources.
Pre-call Briefs
use more CRM context.
CRM Update Suggestions
base their info solely on
Meeting Transcripts.
Don’t mix them all up.
If major changes happen after the meeting, process the latest info first
For instance:
The customer sends a post-meeting email
changing prices,
timelines,
cancelling needs,
or formally confirming details.
The CRM should be updated based on
the latest confirmed info,
not just
accept all Nova suggestions blindly
to clear them out.
Nova’s suggestions are NOT a to-do list
Seeing five suggested updates
doesn’t mean
you have to approve all five today.
The right answer may be:
Approve two,
Edit one,
and skip two.
This is where
human review truly matters.
You can even classify fields as "low impact" or "high impact"
For example:
If a customer clearly provides
a new phone number during a meeting,
it’s usually easier to verify.
But changes to
Deal Stage,
Deal Value,
or Expected Close Date
affect:
Pipeline,
Forecasts,
and Sales Management,
and should be reviewed more carefully.
You don’t have to apply the same review standards to every field
This is another important AI workflow concept.
Not all AI suggestions need word-for-word human verification,
nor should they all be accepted blindly.
The risk level
and review depth
vary by field.
For low-risk fields, verify quickly and pass
For high-impact fields,
check the
transcript,
original words,
current value,
then decide.
This approach
balances
efficiency
and control.
You can even test your review rules with 10 meetings
Don’t assume on day one
that Nova is very accurate.
Select 10 typical sales meetings,
record how often it suggests changes to
Deal Stage,
Value,
and Close Date,
and how many times you
approve, edit, or skip.
This is much more useful than asking "Is Nova accurate?"
You may find
Contact Details
are almost always
good to accept as is,
but Deal Value
often requires manual correction,
and Expected Close Date
often turns a "hope" into
a definite date.
You can adjust
your review process accordingly.
Then write the often-missed field rules into Nova
For example, if phrases like:
"roughly,"
"hope,"
"maybe,"
"expected"
are used,
they shouldn’t
alone
justify updating
the official Close Date.
Write these rules
into the
Field Instructions,
so AI better understands
how your company interprets information.
But Field Instructions aren’t foolproof
They only
increase
consistency.
Meeting transcripts
may still be
inaccurate.
Speakers might be
misidentified.
Customers often
speak unclearly.
So crucial fields
still require
a final human check.
Remember just one key action today
When Nova pops up after a meeting with CRM update suggestions,
don’t first look to:
approve.
First look at:
the customer’s exact words.
Then review:
the current CRM value.
Finally,
check:
the AI suggestion.
Because AI is best at tidying up fuzzy info
The reality is:
The real world
is not so neat.
"Interested"
does not equal
a closed deal.
"Roughly 300,000"
does not equal
an official Deal Value.
"Hope for the end of the month"
does not equal
a confirmed Close Date.
"I need to confirm"
does not mean
the stage has advanced.
The most important thing about CRM isn’t quick data entry
It’s:
That the data truly represents reality.
AI can help
by gathering info,
organizing meetings,
and suggesting updates—
taking these repetitive tasks off your plate.
In the end,
humans only need to
do one truly valuable thing:
Judge whether a suggested change is actually valid now.
If you want to know which step in your workflow is best to hand off to AI first, comment "process" and I can help you get started.
Today, let’s progress with AI a bit more.
Learn one AI tip daily.
Save a little time every day.
Improve a little more each day.
SasaDaily, growing with you.
