Check your data before switching on Nurture Agent
A Breeze prompt that audits one HubSpot list before Nurture Agent goes live and returns a branded report on which fields are worth personalising with.
This one is for the email marketers, the HubSpot admins, and anyone about to let Nurture Agent rewrite their emails without knowing what it has to work with.
What: Using Breeze Assistant to run a Nurture Agent data readiness audit on a single list. It runs eight checks against the email you're planning to send, gives the list a readiness rating, picks the three to five fields most worth personalising with, and returns the lot as a report styled with your own brand kit. It's read-only, so nothing in your portal changes.
Prompt of the week:
Nurture Agent changes what your data is for. It doesn't write new emails. It takes an automated email you've already published and rewrites the subject line, preview text, body and CTA for each recipient, based on what your CRM knows about them. Personalisation tokens and smart content keep working alongside it.
That puts every record on the list to work. A contact with a current job title, a clear industry and a pricing-page visit last week gives the agent something useful to say. A contact with "Mgr - mktg", a free-text industry and no activity since 2023 gets the generic version. Or worse, a confident line built on something that stopped being true a year ago.
Most lists are a mix, and the list view won't show you the split. A property report tells you how full a field is across your database. It won't tell you how many people on this list will get something genuinely personal from this email, or which fields would help this particular email do its job.
Some fields shouldn't be used at all. Records collect things over time: notes from a support call, health details from an event booking form, a personal email someone gave a rep. An agent hunting for relevant detail won't always know which of those would feel helpful and which would feel like you've been reading over someone's shoulder.
This prompt checks all of that for one list at a time, measured against the email you're about to send, and hands it back as a branded report you can share before anyone switches the agent on.
Prompt structure
Paste this into Breeze Assistant and make sure CRM data access is enabled in your AI settings. Fill in the four brackets at the top: the list, the type of email it will receive, what the email should do, and how the recipient should feel. For the email type, pick the closest of: form follow-up, event invite or reminder, weekly newsletter, mass or promotional send, or re-engagement email.
It works best on a single list rather than your whole database. Before you run it, check your brand kit is set up in HubSpot, since the report pulls its logo, colours and fonts from there.
You are auditing CRM data quality before we switch on Nurture Agent
to personalise marketing emails.
Audience to audit: the contacts in the list [LIST NAME].
Email type this list will receive: [EMAIL TYPE].
What this email should do: [e.g. get them to register for our
event].
How the recipient should feel: [e.g. personally invited, not
marketed at].
Keep the email's goal and feeling in mind throughout. The point of
this audit is to find out whether our data can support that email,
and which data would help most.
PART 1: THE AUDIT
For this list, check and report on the following:
1. Core property completeness
For each of these properties, give the number and % of contacts with
a value: job title, company name, industry, lifecycle stage,
country.
2. Data freshness
What % of contacts have had no property updates or activity in the
last 12 months? Flag any properties where values look outdated or
inconsistent (e.g. job titles in mixed formats, free-text
industries).
3. Engagement history
What % of contacts have at least one form submission, page view or
email open in the last 6 months?
4. Associations
What % of contacts are associated with a company? With an open or
closed deal? With a ticket?
5. Custom objects and custom properties
List any custom objects associated with these contacts and the % of
contacts with at least one association. List any custom contact
properties that have no description, an unclear label, or are less
than 30% filled.
6. Duplicates and hygiene
Estimate how many likely duplicate contacts are in this list (same
name and company, or same email domain and name).
7. Sensitive data
Flag any properties on these contacts that look like they hold
sensitive or personal information that shouldn't be used to
personalise an email.
8. Personalisation opportunities
Think about what this email should do and how the recipient should
feel. Then look across every record type available to you (contact,
company, deal and ticket properties, custom objects, and engagement
history such as pages viewed, forms submitted, emails clicked and
events attended) for fields that could make this email more relevant
and help it achieve its goal. Include useful fields that aren't
obvious, not just the standard ones.
For each useful field, give: the field name, the record it sits on,
the number and % of contacts in this list with a value, and one
sentence on how it could be used in this email to support the goal
or feeling. Flag any field that could feel intrusive or overly
familiar rather than helpful.
Then work out:
- A readiness rating for this list: Ready / Ready with gaps / Not
ready
- The 3 to 5 fields most worth using for personalisation, chosen for
how well they support the email's goal and feeling as well as how
complete they are
- The 3 biggest gaps to fix before switching on, in priority order
- An estimate of how many contacts would receive meaningfully
personalised content versus generic content, based on how many
have values for those best fields
Don't guess. If you can't access or calculate something, mark it as
"Not available" rather than estimating.
PART 2: THE REPORT
Create the findings as a single, self-contained HTML file: all CSS
in one <style> tag, no JavaScript, and no external resources apart
from Google Fonts if our brand font needs it (with a web-safe
fallback).
Styling
- Use the brand styling stored in this HubSpot portal: the logo,
colours and fonts from our brand kit. If any of these aren't
available, choose a clean, neutral style that sits well with what
is.
- Header: a dark band in our primary brand colour, with a subtle
gradient using our secondary colours.
- Page: light grey background, white cards with a thin border,
rounded corners and generous spacing.
- Progress bar colours: a positive brand colour (or green) at 70% or
above, an accent or amber colour from 30% to 69%, and red below
30%.
Layout, in this order
1. Header: our logo, a small uppercase label reading "NURTURE AGENT
· DATA READINESS AUDIT", the list name as the headline, and the
email type, the email's goal and today's date underneath.
2. Verdict card: a large coloured badge showing the rating, next to
a one- or two-sentence summary explaining why.
3. Headline numbers: four cards in a row showing contacts audited,
average core property completeness, % with recent engagement, and
contacts likely to receive meaningful personalisation. Each card has
a small uppercase label, a large number, and a one-line note
explaining how it was calculated.
4. Personalisation split: a prominent full-width card with two large
bars comparing contacts likely to receive meaningful personalisation
with those likely to receive generic content (number and % for
each), plus a short note explaining which fields the split is based
on.
5. Best fields to personalise with: a prominent full-width card
(positive colour top border) showing the 3 to 5 best fields from
section 8. For each, show the field name in bold, the record it sits
on, a progress bar with the number and % of contacts filled, and one
sentence on how to use it to support the email's goal or feeling.
6. Audit sections: one card per section (1 to 8), with the section
number in the accent colour before the heading. In every card:
- Show each metric as a row: the property or metric name in bold
on the left, the number of contacts on the right, then a
horizontal progress bar with the % at the end.
- State the date window used where relevant (e.g. "13 April – 13
October 2026").
- Show flagged issues as small tags with contact counts (e.g.
"Mixed-format job titles · 140 contacts"), and give a few real
examples of the inconsistent values.
- Show anything you couldn't access or calculate as "Not
available" in grey italics, with the reason.
Section-specific details:
- Section 1: add a small note under the bars explaining the
colour thresholds.
- Section 4: show company, any deal, open deal, closed deal and
ticket as separate bars, and note whether low coverage matters
for this email type.
- Section 5: show custom object associations as bars, then any
custom properties that are undocumented, unclearly labelled or
under 30% filled as bars, with the issue named beside the
contact count.
- Section 6: show the number of likely duplicate pairs and a
small table of up to 5 examples (name, company, matching
signal), with a note to review before merging.
- Section 7: show flagged sensitive properties as bars and tags,
followed by a highlighted callout recommending which fields to
exclude from personalisation.
- Section 8: show every personalisation opportunity in a table
(field, record, % filled with a small bar, how it could be
used), with any intrusive fields tagged as "Use with care".
7. Fix before switching on: a card (accent colour top border) with 3
numbered items in priority order, each a bold heading with its key
figure followed by one sentence explaining why.
8. Footer: small grey text noting this was generated by Breeze from
live CRM data on today's date and should be checked against
HubSpot's Data Quality tools.
Make it responsive so the cards stack on mobile, and add print
styles so cards don't split across pages and colours are kept when
printed or saved as a PDF.
Breeze returns the report as an HTML file. Save it with a .html ending and open it in your browser, or print it to PDF to share it.
Why this prompt works, and how to adapt it
Most data audits ask whether your records are complete. This one asks whether they're complete enough for a specific email, sent to a specific list, with a specific job to do.
A few things to note about how it is constructed:
The goal and the feeling come first. Telling Breeze what the email should do and how it should land changes what counts as useful. A job title matters for a webinar invite. For a re-engagement email, the last page someone viewed might matter more.
Section 8 looks past the obvious fields. Job title and company are where most personalisation stops. Asking Breeze to search deals, tickets, custom objects and engagement history turns up the fields you'd forgotten you had, like the event someone attended in spring or the product they asked support about.
Sensitive data gets its own check. Section 7 and the "Use with care" tag deal with the fields that are technically available but would make an email feel intrusive. That's a judgement you want made before the agent starts writing, not after a complaint.
The personalisation split is the number that matters. Fill rates are useful, but the headline is how many people will get a genuinely personal email and how many will get the generic one. That's the figure to take to whoever is signing off the switch-on.
"Not available" stops guessing. Breeze can only report on what it can access. Where it can't see something, it says so in grey italics rather than inventing a plausible number.
The report is part of the prompt. Specifying the layout, colours and print styles means you get the same structure every time you run it, which makes before-and-after comparisons easy. No JavaScript means it opens anywhere and prints cleanly.
Adapting it for your portal:
Already know which properties matter? If your team relies on fields the standard checks won't cover, name them. Add: "Also check these properties in sections 1, 2 and 8: [property names]. Treat them as candidates for the best fields to personalise with."
Using custom objects? Breeze will look for them anyway, but naming them saves it guessing. Add: "Include these custom objects in sections 5 and 8: [object names]. For each, tell me which of its properties could support this email."
Auditing a newsletter list rather than a triggered email? A weekly send goes to people at very different stages. Add: "This list gets a weekly newsletter. Break the personalisation split down by lifecycle stage, and tell me which fields help most for each stage."
Sending in more than one language? A well-personalised line still falls flat in the wrong language. Add: "We send in [languages]. Add preferred language and country to section 1, and flag contacts where the two don't match."
Re-running after a clean-up? Save the first report and come back to it. Add: "Compare with the audit from [date]: [paste the headline numbers]. Show which gaps closed, which didn't, and how the personalisation split has moved."
Just want the findings, not the report? Add: "Skip Part 2. Give me the findings from Part 1 as tables in this chat." Useful for a quick check before you commit to a full run.
Beyond the prompt:
The report tells you where the list stands. What you do first matters as much as what it finds.
Start with the sensitive data callout from section 7. Agree with whoever owns data protection which fields stay out of personalisation, and take them out of any tokens or smart content in the email before the agent goes live. It's the quickest fix on the list and the one that prevents the worst outcome.
Next, work through the three gaps to fix, in the order the report gives them. Inconsistent job titles, industries and countries are usually a workflow job rather than a manual one, so standardise the values once and let automation keep them tidy.
Then look hard at the personalisation split. If most of the list sits on the generic side, switching on now mostly gets you the email you already had. Either close the gaps first or start the agent on a smaller list where the best fields are well filled.
Once it's live, read a sample of what it sends. Check the personalised versions against the best fields the report picked, and look for anything that edges into the "Use with care" territory.
Then re-run the prompt on the same list a few weeks later. Watching the split move is the clearest sign the clean-up is working.
Nurture Agent is only as personal as the records behind it. Running this audit before every new list goes live turns that from a hope into something you can measure, and gives you a straight answer when someone asks whether the agent is ready.
FAQs
What does a Nurture Agent data readiness audit check?
It checks whether the contacts on one HubSpot list have the data Nurture Agent needs to personalise a specific email. The prompt in this article runs eight checks: core property completeness, data freshness, engagement history, associations, custom objects and properties, duplicates, sensitive data, and personalisation opportunities. It then rates the list as Ready, Ready with gaps or Not ready.
Does running this Breeze prompt change anything in my HubSpot portal?
No. The prompt only reads your CRM data and returns a report as an HTML file. It doesn't edit, merge or delete any records, so you can run it as often as you like.
What do I need in HubSpot to run this prompt?
You need access to Breeze Assistant, with CRM data access switched on in your AI settings, and a brand kit set up if you want the report in your own branding. Nurture Agent itself has separate subscription and HubSpot Credits requirements, so check HubSpot's Nurture Agent Knowledge Base article for the current list before you plan around it. [Tier and credit claims held for human sign-off.]
Why does my report show "Not available" for some checks?
Breeze can only report on data it can access. If a property, object or activity type isn't visible to it, the report marks that check as "Not available" and gives the reason, rather than estimating a number. Check those areas manually or with HubSpot's Data Quality tools.
Is this better than using HubSpot's Data Quality tools?
They do different jobs. HubSpot's Data Quality tools show the health of your properties and records across the whole database. This prompt judges one list against one email, picks the fields most worth personalising with, and estimates how many contacts will get a personalised version, so the two work best together.