Audit, score, and fix your CRM so the Prospecting Agent actually has something to work with
This one is for the Sales Hub admins, the RevOps leads, and anyone who has ever watched the Prospecting Agent generate an outreach sequence for a contact whose job title is "Unknown" and whose company has no industry, no employee count, and a website field that reads "n/a".
What: Using Breeze Assistant to produce a Prospecting Agent Readiness Audit — a structured analysis of your HubSpot contact and company records that identifies which accounts are enriched enough to support reliable AI-generated outreach, which are not, and exactly what needs fixing before you ask the agent to write a personalised sequence on your behalf.
Prompt of the week:
The Prospecting Agent is one of the most visible parts of HubSpot's current agentic push. The Q1 2026 product narrative put it front and centre: an AI that researches a prospect, drafts a personalised outreach sequence, and hands it to the rep for review and send. In a demo, on clean data, it is genuinely impressive.
In most live portals, the reality is different. The Prospecting Agent's output is only as good as the record it reads. If the company record has no industry, the personalisation defaults to something generic. If the job title field is blank or inconsistent — "VP Sales", "vp of sales", "VP, Sales", "sales vp", all meaning the same thing but entered differently by four different reps — the agent's ability to calibrate tone and angle collapses. If the company has no employee count, the agent cannot judge whether an enterprise pitch or an SMB pitch is appropriate. If the contact's last activity was fourteen months ago and nobody has ever marked that deal as closed-lost, the agent will happily draft a first-touch sequence for someone who told you no over a year ago.
The problem is not the agent. The problem is that most portals were not built with an AI reasoning over the records in mind. They were built for humans who could fill in the gaps mentally, skip the stale records by instinct, and recognise that "Director - Acme Co." means the same thing as "Director of Operations, Acme Corp." The Prospecting Agent cannot do that. It takes the record at face value and produces output that reflects it — which, when the record is poor, is output that embarrasses the rep who sends it.
The community questions have followed a predictable pattern since the agent launched. "Why does the Prospecting Agent keep producing generic sequences even though I told it to personalise?" "How do I know which contacts are ready for AI outreach and which aren't?" "The agent keeps including company information that's wrong — old headcount, the wrong industry." "Is there a way to filter for records the agent will actually do a good job on?" The answer to all of these is the same: you need an honest picture of your CRM's enrichment state before you can deploy the agent with any confidence, and most portals have never had that picture.
This prompt produces it.
It goes further than a standard data quality report. A data quality report tells you what percentage of records have a job title. This prompt tells you which records are Prospecting Agent-ready right now, which need a specific property fixed before they qualify, and which need to be archived, enriched via Breeze Intelligence, or routed back to a rep for manual review before any AI touches them. The output is an action register, not a statistic.
Prompt structure
Paste this into Breeze Assistant and make sure CRM data access is enabled in your AI settings so Breeze can reference your contact and company records, property definitions, deal history, and recent activity:
Role: You are a HubSpot CRM Architect specialising in AI-readiness.
You understand exactly which contact and company properties the
Prospecting Agent relies on to produce personalised, accurate
outreach, and you know the difference between a record that will
produce a strong AI sequence and one that will produce an
embarrassing one.
Task: Audit our HubSpot contact and company records against the
property requirements of the Prospecting Agent and produce a
Prospecting Agent Readiness Audit — a structured report that
classifies records by AI-readiness tier, identifies the specific
enrichment gaps driving poor agent output, and delivers a
prioritised remediation plan with the correct fix for each gap type.
Context:
- Company: [COMPANY NAME]
- Industry: [INDUSTRY]
- HubSpot tier: [Sales Hub Pro/Enterprise, Breeze Intelligence
add-on: YES/NO]
- Approximate total contact records: [NUMBER]
- Approximate total company records: [NUMBER]
- Primary ICP (Ideal Customer Profile):
[e.g., "B2B SaaS, 50-500 employees, UK/EU, VP Sales or
Head of RevOps persona"]
- Prospecting Agent deployment status:
[NOT YET DEPLOYED / DEPLOYED, LOW USE / DEPLOYED, ACTIVE USE]
- Current data enrichment tools in use:
[Breeze Intelligence / ZoomInfo / Cognism / Clearbit / NONE]
- Top three persona types the agent will be outreaching to:
[LIST job title patterns]
- Average sales cycle: [WEEKS or MONTHS]
- Deal stages where Prospecting Agent outreach is intended:
[e.g., "pre-MQL, early MQL, re-engagement of stale leads"]
Audit the following areas:
1. PROSPECTING AGENT PROPERTY DEPENDENCY MAP
Produce an inventory of every HubSpot contact and company
property the Prospecting Agent actively draws on when
constructing a sequence. For each property, specify:
- Whether it is CRITICAL (agent output degrades significantly
without it), SUPPORTING (contributes to personalisation
quality but absence produces generic rather than wrong
output), or CONTEXTUAL (useful when present, not relied upon)
- The specific way a missing or malformed value manifests in
agent output — what does a blank job title actually produce?
What does an inconsistently formatted company name do to
the firmographic section?
- Whether the property is best populated by human input,
enrichment tool, workflow automation, or rep review
At minimum cover: Job Title, Job Function, Seniority Level,
First Name, Last Name, Email, Company Name, Company Domain,
Industry, Number of Employees, Annual Revenue, City/Country,
HubSpot Score or equivalent engagement score, Last Activity Date,
Associated Deal Stage, Lifecycle Stage, and any custom ICP-fit
properties in use.
2. RECORD-TIER CLASSIFICATION
Classify all contact records (and their associated company
records) into four readiness tiers:
TIER 1 — AGENT READY: all CRITICAL properties populated,
values consistent with ICP, no contradictions between contact
and associated company record, last activity within 180 days
or a valid re-engagement reason exists
TIER 2 — ONE FIX AWAY: one CRITICAL property is missing or
malformed; fixable in under five minutes per record with
enrichment or rep input; agent output will be strong once
resolved
TIER 3 — ENRICHMENT REQUIRED: two or more CRITICAL properties
missing; suitable candidate for bulk Breeze Intelligence
enrichment or third-party data append before agent deployment
TIER 4 — DO NOT ROUTE TO AGENT: record is stale (last activity
over 180 days, no open deal, no re-engagement signal), a
known competitor contact, a closed-lost deal with no re-
engagement trigger, or a duplicate. Must be reviewed by a
human before any outreach.
Report: total count and percentage per tier, and the top five
property gaps driving Tier 2 and Tier 3 classifications.
3. FIELD CONSISTENCY AND CONTAMINATION AUDIT
Identify properties where inconsistent formatting is silently
degrading agent output even when a value exists:
- Job title variants that represent the same role differently
(flag patterns across the database, not individual records)
- Company name inconsistencies between the contact's company
name field and the associated company record name
- Industry values that mix standard HubSpot picklist options
with freetext entries from imports
- Phone and country fields formatted in ways that will produce
incorrect regionalisation in agent sequences
- Any property used as a personalisation token in existing
sequence templates where the fill rate is below 70%
For each consistency issue, propose: the standardisation rule,
the workflow or property that enforces it going forward, and
whether a one-time bulk correction is needed or whether a
cadenced cleanup is more practical.
4. STALE RECORD MANAGEMENT
The Prospecting Agent cannot distinguish between a contact who
has never been contacted and one who said no eighteen months ago.
That job belongs to the data model. Identify:
- Contacts with no activity in 180+ days and no associated
open deal: volume, typical age, likely source
- Closed-lost deals where the contact has not been re-enrolled
in a nurture sequence and no suppression flag exists
- Contacts marked as a lifecycle stage that no longer reflects
reality (e.g., "Lead" with no engagement activity in 12+ months)
- Duplicate contact records where the agent might sequence
the same individual twice from two different records
For each pattern, recommend: ARCHIVE / NURTURE SEQUENCE
ENROLMENT / MANUAL REP REVIEW / MERGE / SUPPRESS FROM AGENT.
Do not recommend mass deletion without a retention policy check.
5. ICP-FIT SCORING GAPS
If the Prospecting Agent is being asked to prioritise who to
outreach to, it needs a reliable ICP-fit signal on the record.
Assess:
- Whether a HubSpot Contact Score, Lead Score, or custom ICP
score property exists and is actively calculated
- Whether the scoring criteria reflect the ICP described in
Context or are based on outdated criteria from a previous
strategy
- How many Tier 1 records have no score at all
- Whether the score property is surfaced in the Prospecting
Agent's queue view so reps can see it before reviewing a
drafted sequence
If scoring is absent or outdated, propose a scoring property
rebuild with the specific criteria and weights appropriate
for the ICP described in Context.
6. ENRICHMENT ECONOMICS
If Breeze Intelligence is available, or a third-party enrichment
tool is connected, calculate:
- The estimated number of Tier 3 records that could be promoted
to Tier 1 or Tier 2 with a single enrichment pass
- The credit or cost implication of enriching those records
- Whether the Tier 3 records are worth enriching (are they
ICP-fit targets, or low-priority contacts that happened to
come from an under-qualified list import?)
- The ongoing enrichment trigger that should fire on new
contacts at the point of creation to prevent Tier 3
accumulation recurring
If no enrichment tool is connected, flag the gap and estimate
the manual effort required to promote the highest-value Tier 3
records to Tier 1 by rep research alone.
Constraints:
- Do NOT recommend deleting records without explicit confirmation
that a data retention review has been completed. Recommend
ARCHIVE or SUPPRESS flags as the default
- Every field consistency recommendation must include one concrete
before/after example showing how the current value manifests in
Prospecting Agent output versus the corrected value
- If the volume, property fill rates, or activity history needed
to populate the tier counts are not visible from the current
context, state: "SIGNAL MISSING: [what needs checking manually]"
- Do NOT recommend enriching records that fall into Tier 4 before
human review. Enriching stale or suppression-worthy records
wastes budget and risks outreaching to contacts who opted out
- Any recommendation touching GDPR-relevant data (email address,
phone, company contact details) must be flagged as requiring
a data privacy review before bulk modification
- Distinguish clearly between fixes that a RevOps admin can
execute in a workflow and fixes that require rep judgment
on individual records. Do not conflate the two
Output format:
### I. EXECUTIVE SUMMARY
{3-sentence overview: current AI-readiness state of the CRM,
the single property gap with the highest impact on Prospecting
Agent output quality, and the recommended first action}
### II. PROPERTY DEPENDENCY MAP
| Property | Dependency Level | Impact if Missing | Best Population Method |
### III. RECORD-TIER CLASSIFICATION
| Tier | Count | % of Total | Primary Gap Driving Classification |
{Plus: top five property gaps by volume across Tier 2 and Tier 3}
### IV. FIELD CONSISTENCY ISSUES
| Property | Issue Type | Example (Before → After) | Fix Method | Priority |
### V. STALE RECORD REGISTER
| Pattern | Volume | Recommended Action | Owner | Urgency |
### VI. ICP SCORING ASSESSMENT
{Current state of scoring, gap analysis, and — if rebuild is
needed — proposed scoring criteria with weights for the ICP
described in Context}
### VII. ENRICHMENT ECONOMICS
| Tier 3 Segment | ICP Fit | Records | Estimated Enrichment Cost |
Recommended Action |
{Plus: recommended ongoing enrichment trigger and property}
### VIII. 30 / 60 / 90 DAY REMEDIATION PLAN
{Sequenced actions with owners across RevOps Admin / Sales Ops /
Rep Team / Data Privacy / Sales Leadership}
Why this prompt works — and how to adapt it
Most CRM audits answer the wrong question. They count blank fields and produce a completion percentage. The completeness percentage is not the question the Prospecting Agent cares about. The question it cares about is: can I reason over this record well enough to write a sequence that a prospect will not immediately dismiss as a mail merge? That is a different bar, and this prompt sets it explicitly.
A few things to note about how it is constructed:
The tier system is the core insight. The difference between a Tier 1 record and a Tier 3 record is not just data quality — it is a routing decision. Tier 1 records go to the Prospecting Agent now. Tier 2 records go to an enrichment workflow tonight and come back tomorrow ready to go. Tier 3 records go to a Breeze Intelligence enrichment batch this week. Tier 4 records go to a human before anything else happens. The prompt produces a number for each tier, which turns the audit output into a resourcing conversation rather than a vague aspiration to "clean the data."
Field consistency is the silent killer. A contact whose job title field reads "VP Sales" will produce a different agent output to one whose title reads "VP of Sales" — not because those are different roles, but because the agent interprets the phrasing as a data signal. Multiply that inconsistency across hundreds of records and the Prospecting Agent's personalisation logic becomes unreliable in ways that are hard to diagnose because each individual record looks fine. The consistency audit section forces the problem into the open before it produces embarrassing sequences.
Stale record management is a compliance issue, not just a quality one. An AI-generated cold outreach to a contact who is a known competitor, a closed-lost deal from two years ago, or someone who has previously unsubscribed is not a data quality problem — it is a potential GDPR breach and a relationship problem. The prompt treats Tier 4 classification as a hard gate that requires human review before the agent is allowed near the record. That is not being overcautious; it is being professionally responsible with the tool.
Enrichment economics make the plan defensible. The question "should we enrich these records?" is not a data quality question; it is a commercial one. If the 800 Tier 3 records are ICP-fit target accounts that came from a conference list last year, enrichment is cheap compared to the pipeline value of getting them agent-ready. If the 800 records are a cold-purchased list from three years ago that never produced a single meeting, enrichment is waste. The prompt forces both the volume and the ICP-fit judgment to sit next to each other so the decision can be made commercially rather than by default.
"SIGNAL MISSING" is the anti-confabulation clause — again. Breeze can read property fill rates, deal associations, and activity dates from the CRM context. It cannot read the informal suppression list in the SDR team's shared spreadsheet, the outreach history from a tool that was not integrated into HubSpot, or the compliance team's view on which imported lists are legally actionable. Where the audit reaches the edge of what the CRM can tell it, the prompt flags the gap rather than guesses past it.
Adapting it for your portal:
High-velocity, low-ACV motion? If you are running hundreds of sequences a week against SMB prospects, the per-record manual review suggested for Tier 4 is not practical. Add: "Our motion is high-velocity. Tier 4 records should be classified by rule rather than individual review. Propose the workflow criteria that automatically archive or suppress a record based on activity date, lifecycle stage, and deal history, without requiring a rep to touch each one." The output shifts from a review cadence to an automated suppression workflow.
Named-account ABM motion? If you are running a tight account list where every contact matters, add: "Our motion is ABM. We are working a defined target account list of [NUMBER] companies. Weight the audit entirely towards completeness and enrichment quality on those accounts, and flag any account where the buying committee coverage — the number of contacts associated with the account — is below [NUMBER] stakeholders." The tier classification becomes account-level rather than contact-level.
No Breeze Intelligence in your tier? If enrichment credits are not available, the Tier 3 recommendations will default to manual research. Add: "We do not have Breeze Intelligence. For Tier 3 records that are ICP-fit, propose the specific properties a rep can populate in under three minutes using public sources — LinkedIn, company website, Companies House — and build the data entry into the existing prospecting workflow rather than treating it as a separate admin task." The plan becomes rep-executable rather than tool-dependent.
Starting fresh after a bad import? If the root cause of your enrichment gaps is a single bulk import that came in under-enriched, add: "A significant portion of our records originated from a bulk import in [MONTH/YEAR]. Treat that cohort as a separate remediation workstream, assess whether the source data is good enough to enrich from, and recommend whether to process them in bulk or retire the cohort and rebuild from a cleaner source." Import debt needs a different treatment to organic accumulation.
Quarterly cadence? Save the output and re-run the prompt 90 days later with: "Compare against the output from [DATE]. Report the change in tier distribution — how many records have moved from Tier 3 to Tier 1 or 2, whether the Tier 4 backlog has reduced, and which field consistency issues from the previous audit have been resolved and which have re-emerged from new imports." That turns the audit into a living CRM health metric rather than a one-off project.
Beyond the prompt:
The Prospecting Agent Readiness Audit tells you the state of the data. What you do with the output is where the real work begins.
Start with Tier 4 suppression. Before anything else is fixed, Tier 4 records need a human gate. The commercial cost of the Prospecting Agent outreaching to a contact who opted out, or who told a rep no six months ago, is disproportionate to the effort of routing them correctly. A simple active list — all contacts meeting Tier 4 criteria — that acts as a suppression check against Prospecting Agent queues costs an afternoon to build and prevents the kind of incident that lands in a sales leadership meeting.
Then move to field consistency, not volume. It is tempting to start the remediation by mass-enriching the Tier 3 records — the numbers are big and the fix feels impactful. Resist. A hundred Tier 1 records producing inconsistent agent output because job title formatting is chaotic will undermine rep confidence in the tool faster than the Tier 3 backlog will. Fix the consistency issues first. Run the agent on clean Tier 1 records. Let the reps see what good output looks like before you scale.
Once the reps trust the output, work the Tier 2 records systematically. These are the quickest wins: one property, one fix, one record promoted to agent-ready. Build the workflow that auto-enriches on creation, and set the SDR team a weekly target of clearing the manual Tier 2 queue. Track the tier distribution monthly. When Tier 1 crosses sixty per cent of active contacts, the Prospecting Agent is running on a CRM it can actually work with.
The Tier 3 enrichment conversation should happen after the Tier 2 rhythm is established, not before. By the time you get there, you will have a clear answer to the commercial question — is this cohort worth the enrichment spend? — because you will have watched what happens to sequences when the data underneath them is good.
The Prospecting Agent is not a shortcut to skipping CRM hygiene. It is a reason, finally, to take it seriously.