What: Using Breeze Assistant to pressure-test your deal pipeline - stage definitions, exit criteria, skip patterns, zombie deals, and the custom property surface that Smart Deal Progression now reads from - and turn the findings into a Pipeline Sharpening Plan: a configuration-first action list that makes the suggestions from Smart Deal Progression trustworthy on day one, instead of a rolling source of rep frustration.
Prompt of the week
This prompt asks Breeze to look at your actual stages, your actual exit criteria, and the skip patterns your conversion reports are quietly hiding - and produce a plan that gets the whole thing structurally fit for an AI agent to read meeting transcripts and propose deal updates without making the reports worse.
Smart Deal Progression launched at the Spring 2026 Spotlight as one of the headline features of the new Sales Hub. It listens to meeting transcripts via Notetaker, then suggests CRM updates, drafts follow-up emails, and recommends next steps after every customer call - including custom property updates powered by Data Agent. Reps review the suggestions, approve or edit, and apply with one click.
On paper, it solves the most evergreen complaint in sales operations: reps don’t update the CRM after meetings, deal stages go stale, and pipeline reports drift further from reality every week. In practice, the feature is only as good as the pipeline structure it is reasoning against. If your stages are vague - “Qualified”, “Engaged”, “In Progress” - the AI will confidently propose stage transitions based on its best guess, and your reps will lose faith in the suggestions on day one.
Strip the platform names off the threads in the HubSpot and Salesforce communities and you genuinely cannot tell which is which. The same questions cycle every few weeks: “how do I stop reps skipping deal stages?”, “why do my conversion rates make no sense?”, “what should the exit criteria for [stage X] actually be?”, and the evergreen “our pipeline data is a mess and I don’t trust any of the reports.”
The problem is not that the tools don’t exist. HubSpot has pipeline rules, deal stage probabilities, required properties, Smart Data Capture settings, and now Smart Deal Progression sitting over the top of all of it. The problem is that nobody has sat down and rewritten the stage exit criteria into language an AI agent - or, frankly, a new sales rep - can actually evaluate against.
There is also a quieter, more recent reason this prompt is timely. The Smart Data Capture configuration that controls which deal properties Smart Deal Progression can update was reset earlier this month, and any custom properties previously in there were removed. If anyone on your team set this up months ago and assumed it was still working, the answer is: probably not in the way they think. Worth checking before you go any further.
This prompt is the forcing function for the work everyone agrees is overdue. It does the analysis your team does not have a free three days for, and hands back something specific enough to act on this week.
Prompt structure
Paste this into Breeze Assistant and make sure CRM data access is enabled in your AI settings so Breeze can reference your deal records, pipeline configuration, historical stage transitions, and Smart Data Capture settings:
Role: You are a HubSpot Sales Pipeline Architect specialising in
deal pipeline design, exit criteria, and the configuration work
required before deploying AI features like Smart Deal Progression
across a sales team.
Task: Audit our current deal pipeline and produce a Pipeline
Sharpening Plan that rewrites stage exit criteria into observable
language, flags overlapping or redundant stages, recommends
pipeline rules to prevent stage-skipping, and produces a Smart
Data Capture configuration plan covering which deal properties
Smart Deal Progression should and should not be allowed to update
- including the credit-cost implications of every custom property
added.
Context:
- Company: [COMPANY NAME]
- Industry: [INDUSTRY]
- HubSpot tier: [must be Sales Hub Pro/Enterprise or Service Hub
Pro/Enterprise to run Smart Deal Progression]
- Pipelines in scope: [e.g., "New Business only" or "New Business
+ Renewal + Expansion"]
- Approx active open deal volume: [NUMBER]
- Sales motion: [TRANSACTIONAL / CONSULTATIVE / ENTERPRISE / MIXED]
- Typical sales cycle length: [e.g., "21 days SMB, 90 days
enterprise"]
- Notetaker adoption: [NOT ENABLED / ENABLED, PATCHY USE /
ENABLED, CONSISTENT USE]
- Approx monthly customer-facing meeting volume per rep: [NUMBER]
- HubSpot Credit budget for AI features: [MONTHLY NUMBER, or
"tight / moderate / generous"]
- Current exit criteria documentation: [NOWHERE / IN HEADS /
INFORMAL DOC / FORMAL PLAYBOOK]
Audit the following areas:
1. STAGE DEFINITION & EXIT CRITERIA
For every stage in scope:
- Is there a written exit criterion?
- Is the criterion observable - i.e., could a meeting transcript
or activity record objectively confirm it has been met?
- Is it testable, or is it vague seller-internal language like
"qualified", "engaged", "ready to buy"?
For each stage, output a REWRITE in the form:
"Stage [X] exits when [specific observable thing] is true,
confirmable via [meeting transcript / property value / activity]."
2. STAGE OVERLAPS & REDUNDANCY
Identify:
- Stages with overlapping or near-identical exit criteria
- Adjacent stages with no meaningful difference in conversion
rate (statistical noise rather than real signal)
- Stages that are really sub-states of another (e.g., "Demo
Scheduled" + "Demo Completed" + "Demo Followed Up" when one
stage with a sub-status would be cleaner)
For each, recommend: MERGE / SPLIT / KEEP, with a note on the
impact on twelve months of historical reporting.
3. SKIP PATTERNS & PROCESS INTEGRITY
Surface:
- Which stages get skipped most often, and the volume
- Backwards stage moves (a deal moving from later to earlier)
- Deals closing won or lost without ever passing through
mid-funnel stages
- Stages systematically bypassed by certain deal types
(partner-sourced, renewal, expansion)
Recommend pipeline rules to either ENFORCE the progression or
adjust the pipeline structure to reflect the real flow.
4. TIME-IN-STAGE & ZOMBIE DEALS
Assess:
- Average time per stage versus a healthy benchmark
- Stages where deals routinely age out (60+ days with no activity)
- Deals open with close dates already in the past
- Stages where the close date keeps getting pushed without a
stage change (a classic sign Smart Deal Progression will
amplify if left unaddressed)
Recommend automated rules and a RE-ENGAGE / NURTURE / ARCHIVE
decision framework.
5. SMART DATA CAPTURE CONFIGURATION
Smart Deal Progression updates deal properties from meeting
transcripts via the Smart Data Capture settings under Data
Management. Every custom property added to that configuration
consumes HubSpot Credits per transcript run; standard properties
do not. Classify each deal property in scope into one of three
buckets:
- AI-CAN-UPDATE (standard property, free): low-risk, easily
verifiable from a transcript - e.g., next steps, decision-maker
name, competitor mentioned, meeting outcome
- AI-CAN-UPDATE (custom property, credit-costed): medium-risk,
specific to your sales motion, worth the credits - e.g.,
MEDDIC fields, qualification answers, pain-point tags
- NEVER-AI: high-risk, locked at the property level - e.g.,
contracted value, signed close date, legal terms, anything
that drives commission or invoicing
For each custom property in the AI-CAN-UPDATE bucket, estimate
the monthly credit cost based on meeting volume so the team can
plan budget.
6. NOTETAKER READINESS
Smart Deal Progression depends on Notetaker (or a third-party
transcript integration) capturing meetings. Assess:
- What % of customer-facing meetings currently get captured?
- Which meeting types are being missed (in-person, phone calls
not run through HubSpot Calling, video platforms not connected)?
- Are reps consenting to recording consistently?
- Are external attendees giving consent in line with regional
compliance (UK/EU GDPR, CCPA)?
Recommend a minimum viable threshold of meeting coverage before
SDP should be enabled, and a rollout plan if coverage is patchy.
Constraints:
- Do NOT recommend pipeline restructures that would invalidate more
than 12 months of historical reporting without flagging the impact
and proposing a parallel-pipeline migration path
- Every rewritten exit criterion must be expressed in language a
meeting transcript could plausibly confirm - if a criterion
depends on an offline signal (handshake at a trade show, internal
procurement decision the AI cannot see), flag it as
NOT SDP-OBSERVABLE
- The Smart Data Capture configuration must explicitly call out the
credit cost per custom property per transcript run and estimate
monthly total based on meeting volume
- NEVER-AI properties must be locked at the property level, with
user permissions restricting who can edit them, before SDP is
enabled
- If usage, fill rate, or pattern data for a stage or property is
not visible from the current context, state:
"SIGNAL MISSING: [what needs checking manually]"
- Recommend SDP is enabled in REVIEW-AND-APPROVE mode for at least
4 weeks before any change to the rep workflow is considered
Output format:
### I. PIPELINE HEALTH SUMMARY
{3-sentence overview of current pipeline integrity, the single
biggest risk to Smart Deal Progression giving reliable suggestions,
and overall AI readiness rating}
### II. STAGE DEFINITION REWRITES
| Stage | Current Criterion | Rewritten Observable Criterion |
| Confirmable Via |
### III. STAGE OVERLAP & REDUNDANCY
| Stages | Issue | Recommendation | Reporting Impact |
### IV. SKIP PATTERN & PIPELINE RULES
| Pattern | Volume | Cause | Recommended Rule |
### V. TIME-IN-STAGE CLEANUP
| Issue | Volume | Recommended Rule |
### VI. SMART DATA CAPTURE CONFIGURATION
| Property | Type (Standard/Custom) | Bucket | Est. Credit Cost / Month |
| Lockdown Action |
### VII. NOTETAKER READINESS PLAN
{Coverage assessment, gaps, minimum viable threshold, rollout
sequence}
### VIII. 30 / 60 / 90 DAY CONFIGURATION ROADMAP
{Prioritised actions with owner roles: Sales Manager / RevOps /
CRM Admin / Finance for credit budget}
Why this prompt works - and how to adapt it
There is a particular kind of work that sits in every sales operations backlog without ever quite making it to the top: pipeline cleanup. It is unglamorous, the wins are slow to appear in any single report, and the political cost of telling a senior rep their stage definitions are vague is high. So nothing happens, the conversion rates get a little less reliable each quarter, and everyone learns to read pipeline reports with a healthy scepticism that should not be necessary.
The prompt is what gets it onto the calendar - by doing the analysis for you. Staring at every stage and deciding what would actually have to be true for a deal to move into it: that is the bit you cannot delegate to a junior, the bit a consultant charges three days for, and the bit Breeze is genuinely good at when you give it the right context.
A few things to note about how it's constructed
Sales motion changes everything. A 21-day transactional cycle and a nine-month enterprise cycle are not just longer or shorter versions of each other - they are different sales processes, with different stages and different signals. Smart Deal Progression’s suggestions will only feel right if the stages reflect the actual buying behaviour, which means Breeze needs to know what that behaviour looks like. Pump the Context section full: motion, cycle length, average meeting volume per rep, Notetaker adoption. The audit it produces is a function of what you put in.
The “observable language” requirement is the high-leverage move. Most stage exit criteria are written in seller-internal language - “qualified”, “engaged”, “ready” - that means whatever the rep wants it to mean on a Friday afternoon. The prompt forces every criterion into language a meeting transcript could plausibly confirm: “budget figure stated”, “decision-maker named”, “success criteria agreed”. That is the bit that makes Smart Deal Progression’s suggestions reliable rather than hopeful - and, as a side effect, it also makes onboarding new reps faster.
The Smart Data Capture configuration is where the credit costs hide. Every custom property added to the configuration consumes HubSpot Credits each time a transcript runs against it. Standard properties don’t. The prompt forces explicit credit estimation per property so the cost is visible up front, not at the end of the month when finance asks.
The NEVER-AI bucket is the trust layer. The fastest way to lose sales-team trust in any AI feature is to let it propose changes to fields that drive commission or invoicing. The matrix forces every deal property into one of three buckets - what the AI can update for free, what it can update with a credit cost, and what it must never touch - and then locks the never-touch fields at the property level before the feature is enabled. Boring. Critical.
The “SIGNAL MISSING” flag is Breeze’s permission to say “I don’t know”. Pipeline configuration and historical data are visible to it. Rep tribal knowledge about why a stage exists, or what “qualified” has come to mean for your team this quarter, is not. When a recommendation depends on something invisible to the agent, you want the gap flagged rather than papered over with a confident-sounding guess. This single instruction is one of the more reliable safeguards against the AI generating fluent nonsense.
The historical-reporting constraint exists to stop a useful instinct turning into a destructive one. A pipeline cleanup that invalidates twelve months of conversion data is not a cleanup; it is an act of vandalism on your reporting. The prompt is explicit that any restructure with reporting impact must be flagged with a parallel-pipeline migration path, not just executed.
Adapting it for your portal
Long sales cycles deserve a tighter focus. If yours runs six months or more, add this to the Context: “Our typical deal involves [N] decision-makers and [N] meetings before close. Weight the audit towards the mid-funnel stages where deals stall.” Breeze will then focus the rewriting effort on the stages where Smart Deal Progression has the most to offer - the long, ambiguous middle where deals quietly die.
Running multiple pipelines? Name each one explicitly in Context (New Business, Renewal, Expansion, Partner-Sourced). Breeze will produce stage-definition rewrites for each pipeline and flag any that should run on a separate SDP configuration. Renewal pipelines in particular almost always need different treatment from new business.
Running HubSpot alongside Salesforce changes the constraint set. Stages and properties that participate in the sync cannot be renamed unilaterally, so flag those up front:
Our HubSpot pipeline syncs with Salesforce via [HubSpot native /
third-party connector]. The following stages or properties are part
of the sync and cannot be renamed without coordinated change:
[LIST].
Breeze will respect those constraints and route the rewrites through additive properties rather than breaking changes.
Be honest about patchy Notetaker adoption. Smart Deal Progression at thirty per cent meeting coverage will look broken for reasons that have nothing to do with the AI itself. State the current coverage level in the Context and Breeze will produce a rollout plan with a minimum viable adoption threshold before SDP should be enabled - usually somewhere around seventy-five per cent of customer-facing meetings.
On a tight credit budget? Add this constraint:
“Treat HubSpot Credits as a constrained resource. Recommend the smallest Smart Data Capture configuration that delivers the highest leverage, prioritising standard properties and only including custom properties where the credit cost is justified by the operational value.”
Breeze will then produce a leaner configuration optimised for cost, not coverage.
A quarterly cadence is where the compounding value sits. Save the output, then 90 days later re-run with this addition:
“Compare against the output from [DATE] and report on which stages have been rewritten, which conversion rates have stabilised, which new skip patterns have emerged, and which Smart Data Capture properties are pulling their weight against their credit cost.”
That gives you a rolling pipeline-integrity dashboard in prompt form, and a very easy progress update for sales leadership.
Beyond the prompt
The plan is the strategy. Everything past it is execution - and unlike most cleanup work, the order of operations here is non-negotiable.
Start with rewriting exit criteria. For each stage, get a single sentence written in observable language, agreed by the sales manager who owns that stage and a senior rep who works it. Publish the new definitions as a one-page playbook. This step alone, before any AI is involved, will tighten your pipeline reports within a quarter.
Then run the Smart Data Capture configuration. Lock down the NEVER-AI properties at the field level - anything that drives commission, anything that drives invoicing, anything with a contractual implication, gets explicitly flagged as not editable by Smart Deal Progression. Do this before you turn the feature on, not after a rep finds the AI has rewritten a contracted value. Then add the standard properties (free) you want SDP to populate from transcripts. Only then add the custom properties you can justify the credit cost on.
Then enable Smart Deal Progression in review-and-approve mode for at least four weeks. Track the approval rate per rep and per stage as your trust metric - if approvals on a particular stage hover below sixty per cent, the exit criterion for that stage probably needs another rewrite. Treat the review rate as feedback on your pipeline definitions, not just on the AI.
Pair that with pipeline rules preventing stage-skipping for the stages where skipping has been a real problem, and time-in-stage automation rules for zombie deals. Now the loop is closed: clean stage definitions, tight property-level controls, AI suggestions calibrated against real conversations, credit costs predictable, and rules preventing the worst of the manual workarounds.
And if you are also running Prospecting Agent or planning to roll out the new Plays framework, none of this is wasted effort. Prospecting Agent feeding leads into a sharpened pipeline, with Smart Deal Progression keeping the data accurate as deals progress - that is the unified Sales Hub story HubSpot has been telling since the Spotlight. Skip this work and you get the inverse: an AI confidently rewriting a pipeline that was never reliable in the first place, at a credit cost that quietly grows every month.
Sharpen the stages first. Once they are written in language a transcript can confirm, the rest - trustworthy AI suggestions, honest conversion reports, and a sales team that actually trusts the pipeline data - comes along for the ride.
