Monday pipeline reviews have a familiar shape. Someone pulls up the deals board. The team debates gut feelings about which deals are real. An hour passes. Nothing materially changes.

The problem isn't the meeting. It's that nobody walked in with a clear, criteria-based read on which deals are actually at risk - before the opinions started flying.

This hack fixes that in five minutes using a CSV export and a prompt.

 

What the Hack Does

You export your open deals from HubSpot, drop the file into Claude or ChatGPT, and ask it to score each deal against three criteria that genuinely predict pipeline health:

  1. Is the next step defined or just a vague "follow up"?
  2. Is the close date realistic, or has it been pushed and is now overdue or suspiciously far out?
  3. Is the deal amount a specific number that suggests a real quote, or a round number that looks like a guess?

The output is a ranked list of your worst-scoring deals with a one-sentence explanation of each risk. Walk into Monday's standup with that list and the conversation is immediately more useful.

 

How to Set it Up

Step 1: Set Up Your Deal View

Go to Sales > Deals in HubSpot. Switch to the list (table) view if you're not already there.

Filter for open deals only. Depending on how your pipeline is set up, that means filtering out Closed Won and Closed Lost stages. If you want to scope it further - specific owner, specific pipeline, deals above a certain value - apply those filters now.

Before exporting, customise your columns to include the properties the AI needs to score meaningfully. The default deal view will not have everything. Add these at minimum:

  • Deal name
  • Deal owner
  • Deal amount
  • Close date
  • Deal stage
  • Next activity date
  • Associated company name
  • Pipeline (if you run multiple)

One important note: HubSpot's deal export doesn't include notes or engagements. Only the properties visible as columns in your view come through in the CSV. This matters for the "next step" criterion - if your team logs next steps as notes on the deal record rather than in a dedicated property, that information will not be in the export.

If your portal has a custom Next Step property - a text field where reps log what they have committed to doing next - add that column now. It's the single most useful field for this exercise. If you don't have one, this hack is a good reason to create it. More on that below.

 

Step 2: Export as CSV

With your columns set and filters applied, click Export in the top right of the deals view. Select CSV as your file format and choose to export the columns in your current view.

HubSpot will email you a download link. It usually arrives within a couple of minutes.

 

Before You Upload: A Quick Note on Data and Confidentiality

Deal data sits in a different privacy category to contact data - but it's not risk-free to share externally, and it is worth taking 30 seconds to think about before you upload.

Check your deal names. If your team names deals using individual contact names - "Jane Smith - Renewal 2026" rather than "Corp - Renewal 2026" - your export contains personal identifiers. Under GDPR and similar privacy frameworks, a named individual is personal data regardless of the context. Either rename deals to company-based formats before uploading, or remove the deal name column from the export and replace it with a company name column instead. The AI can score deals without knowing the individual's name.

Treat deal amounts and pipeline data as commercially sensitive. Your live pipeline, revenue figures, deal stages, and close date forecasts aren't personal data. But they are confidential business information. You're sharing your sales strategy with an external tool, and that's worth knowing. The enterprise and team versions of both Claude and ChatGPT explicitly don't use submitted data to train their AI models. A free or personal account may. For anything involving real pipeline numbers, use an enterprise or team account and verify the data handling policy before uploading.

Deal owner names are internal employee data. Minor in most jurisdictions for this kind of internal analysis, but worth noting if your organisation has strict internal data handling policies.

The short version: swap individual names out of deal names if they're there, use an enterprise AI account, and you're good to go.

 

Step 3: Run the Prompt

Open Claude or ChatGPT. Upload the CSV and use this prompt:

"Here are my open deals. For each, score 1 to 5 on three criteria:

(1) Is the next step defined or just 'follow up'?

(2) Is the close date realistic or overdue?

(3) Is the amount a specific number backed by a quote, or a round-number guess? List the 5 worst-scoring deals with one sentence each on why they are at risk."

What comes back is a table: deal name, scores across each criterion, and a plain-language risk summary for the bottom five. That's your agenda for the pipeline review.

 

What the AI is Actually Doing

It's worth understanding what each criterion is measuring, and what the AI can and can't infer from a CSV.

Next step quality. If you have a Next Step property in the export, the AI reads the text and assesses whether it is specific ("send revised proposal by Thursday") or vague ("check in," "follow up," "TBC"). Vague next steps are a reliable signal that the deal has no real momentum. If you don't have a Next Step property, the AI will use deal stage and next activity date as proxies, less precise but still useful.

Close date realism. The AI looks at whether the close date has already passed, whether it's the last day of a month or quarter (a common sign of optimistic padding), and whether the deal stage suggests the timeline is credible. It can't see how many times the date has been pushed. HubSpot doesn't include close date history in a standard export. But a deal in early-stage Discovery with a close date three weeks out, or a deal overdue by 45 days still showing as open, both score poorly on this criterion without needing the history.

Amount credibility. Round numbers ($5,000, $10,000, $50,000) suggest the amount was entered as a placeholder rather than derived from a real quote or proposal. Specific numbers ($7,400, $12,750) suggest someone has done the actual calculation. The AI flags round amounts as lower confidence. This is a heuristic, not a rule, but it's a surprisingly reliable signal for deals that aren't yet grounded in a real commercial conversation.

 

The One Property Worth Adding First

If your team doesn't have a dedicated Next Step text property on deals, add one before running this hack. It takes two minutes in Settings > Properties > Deals > Create property.

A Next Step property does two things: it makes this hack significantly more accurate, and it changes rep behaviour. When there's a visible, reportable field that says "what happens next," reps fill it in because they know it'll be reviewed. Deals with empty Next Step fields become immediately visible as unworked pipeline - which is exactly what you want to see before a standup.

Once it exists, add it to your deals board as a column. Make it part of your pipeline review criteria. The data quality compounds fast.

 

What to Do With the Output

The five worst-scoring deals are your standup agenda. For each one, the question isn't "Is this deal going to close?", it's "What specifically needs to change for this deal to be real?"

That reframe matters. A deal with a vague next step needs a committed action, not a forecast update. A deal with an overdue close date needs a conversation with the prospect, not a date push. A deal with a round-number amount needs a quote sent. The AI's one-line risk summary makes the right question obvious before anyone has to debate it.

Run this every week before your pipeline review. The hour-long gut-feel debate becomes a 20-minute triage with clear owners and actions.

 

Want a Pipeline that Scores Itself?

If this makes you think about building a more systematic pipeline hygiene process - custom scoring properties, deal health workflows, or a pipeline review dashboard that surfaces at-risk deals automatically - that's the kind of setup we build at Neighbourhood.

We're a Diamond HubSpot Partner. If your pipeline data is there but your reviews are still running on instinct, get in touch and we will help you fix that.

Talk to us about your HubSpot setup.

 

Get More Hacks Like This

Ops Intel is our newsletter for HubSpot admins and sales ops teams -practical hacks and no filler, straight to your inbox every week. Subscribe to Ops Intel.

We're also on Facebook and YouTube with shorter, visual breakdowns of hacks like this one. Come find us there.

Happy HubSpotting!