You set up HubSpot's Prospecting Agent, connected a selling profile, and let it loose on your database - but it isn't working the way it should. It's flagging contacts who unsubscribed a year ago, recommending outreach to people at companies that already churned, and drafting emails that miss the mark on tone.
The instinct is to blame the AI. In most implementations we've seen, the agent is doing exactly what it was told to do - the problem is what it was told, and what data it was told it with.
INPUTS
Selling Profile
too generic
CRM Data
duplicates, gaps
Buying Signals
no provider connected
OUTPUT
Lead Recommendations
Incomplete inputs don't produce a wrong answer by accident - the agent builds an accurate recommendation from inaccurate data.
Quick Answer
HubSpot's Prospecting Agent recommends bad leads when its selling profile is vague, the CRM data it reads (engagement history, bounce status, company fit) is stale or incomplete, or it isn't connected to a contact-sourcing provider for buying signals. The agent is only as accurate as the CRM data and configuration behind it - fixing the underlying setup, not the AI itself, is what resolves bad recommendations.
What HubSpot's Prospecting Agent Is Actually Built to Do
Before diagnosing what's going wrong, it helps to be precise about what the HubSpot Prospecting Agent is supposed to do. It researches contacts and companies, drafts personalized outreach, and can run semi-autonomously (with rep review) or fully autonomously.
It draws on a mix of sources: your CRM engagement history, public data (LinkedIn, company websites, news), and - if you're on the buying signals beta - connected providers like ZoomInfo, Apollo, or Surfe for contact sourcing and in-market signals.
How Much Does the HubSpot Prospecting Agent Cost?
Pricing is outcome-based: HubSpot charges per lead the agent recommends for outreach, billed through HubSpot Credits, with a free trial period for new activations. There's no flat monthly fee for the agent itself - the cost scales with how many leads it actually surfaces, which means a poorly configured setup doesn't just produce bad recommendations, it burns credits on recommendations you'll never use.
Who Can Use It
The Prospecting Agent is available on the Starter, Professional, and Enterprise editions of HubSpot, and it runs on HubSpot Credits. To set it up or manage it, a user needs to be a Super Admin or have Access prospecting agent permissions specifically granted - worth checking before you assume everyone on the sales team can configure it themselves.
Is Your CRM Data Ready for AI Agents?
See Our Data Cleaning ServicesThe Real Reason It Recommends the Wrong Leads
The agent considers recent engagement - up to the past year - associated with a contact record. If that engagement history is incomplete because of duplicate records, unsynced integrations, or properties that were never mapped correctly during a migration, the agent is working from a partial picture. It isn't guessing wrong; it's building an accurate recommendation from inaccurate inputs.
The same applies to your selling profile. HubSpot lets you create up to 85 selling profiles, each defining the value proposition, tone, and target segment the agent writes for. A single generic selling profile covering every product and every persona will generate generic - and often misdirected - outreach, even with perfect CRM data behind it.
And if you're relying on buying signals to prioritize accounts, the agent needs a connected data provider to source verified contacts. Without one, it's limited to whatever contact data already exists in your CRM - which, for a lot of B2B databases, means outdated titles, dead email addresses, and gaps at the accounts that matter most. This is the same sync problem behind most CRM integration issues we get called in to fix.
Well-Configured vs. Poorly-Configured Prospecting Agent
One mistake doesn't show up in the table above because it's less about the agent's settings and more about migration history: custom properties from a past CRM migration left unmapped, so the agent's view of company fit is missing fields your sales team relies on manually without realizing the agent can't see them.
A Quick Self-Check
If more than a handful of the agent's recommendations feel off, ask: when was the CRM data behind this contact last cleaned up, and does the selling profile it's using actually describe the segment this contact belongs to? Most of the time, the answer points straight at the fix.
How to Fix It: A Practical Checklist
Getting the HubSpot Prospecting Agent to recommend the right leads isn't about tweaking AI settings - it's about the same CRM fundamentals that make any sales tool work: clean data, clear segmentation, and the right integrations feeding it.
|
1
Split selling profiles In Agent setup > selling profiles, create one per product line, market, or persona - HubSpot allows up to 85. |
2
Deduplicate contact records Run a merge pass on contacts and companies before enrollment, not after - retroactive cleanup won't fix history already read. |
|
|
3
Connect a data provider Link ZoomInfo, Apollo, or Surfe under the buying signals settings before enabling that beta, not after. |
4
Maintain suppression lists Upload competitors, existing customers, and opted-out contacts, and review the list quarterly as your customer base grows. |
|
|
5
Start semi-autonomous Assign one owner to review flagged recommendations weekly for the first month before switching to fully autonomous. |
6
Re-check property mappings Pull up the field mapping from your last migration and confirm the properties the agent reads are still populated. |
|
What to Expect After You Fix It
Cleaning up the inputs doesn't change recommendations retroactively - the agent doesn't re-score contacts it already flagged. What it does change is the next enrollment cycle. Once selling profiles are segmented and duplicate records are merged, expect the improvement to show up gradually as the agent works through your database again, not overnight.
A few things worth tracking in the weeks after: the rate at which reps mark recommendations as irrelevant (this should drop first, since it's the most direct signal of a mismatched selling profile), how many recommended leads actually convert to a booked meeting, and how many credits you're spending per accepted lead versus per rejected one. If the reject rate stays flat after a clean data pass, the issue usually isn't the CRM data anymore - it's the selling profile's positioning or the segment definition itself.
If you've just connected a contact-sourcing provider for buying signals, give it longer. Signal coverage builds up as the provider tracks account activity over time, so a newly connected integration will look thin for the first few weeks even when it's configured correctly.
"The agent isn't the bottleneck we usually find - it's whatever was already broken in the CRM before the agent got turned on," says Stefan Loncar, CEO of Loncom Consulting.
If the Prospecting Agent is surfacing bad leads, it's worth asking what else in your CRM runs on the same incomplete data - reporting, forecasting, and other AI agents and automations all read from the same records. We've written before about how dirty CRM data undermines AI automation more broadly - the Prospecting Agent is just one visible symptom of that same pattern.
Ready to Fix Your Prospecting Agent Setup?
Book a Free ConsultationFAQ
What is HubSpot's Prospecting Agent?
It's an AI-powered sales tool built into HubSpot's CRM that researches contacts and companies, then drafts and can send personalized outreach on a rep's behalf, running semi- or fully autonomously depending on the settings chosen.
Why does the Prospecting Agent recommend contacts who shouldn't be enrolled?
Usually because the underlying CRM data is incomplete or outdated - duplicate records, missing bounce checks, or unmerged engagement history - or because suppression lists for competitors and existing customers were never uploaded. It's the same root cause we've seen behind HubSpot Buyer Intent not working as expected.
How do I know if my HubSpot data is holding the agent back?
Check how recently your contact and company records were deduplicated, whether every selling profile matches a specific segment rather than covering everyone, and whether a contact-sourcing provider is connected if you're using buying signals.
Does this apply if I'm only using the agent for a few contacts at a time?
Yes - even manual, low-volume enrollment pulls from the same engagement history and selling profile settings, so a data or configuration gap will show up in recommendation quality regardless of scale.
Can Loncom help fix this?
Yes. This is the kind of setup work we do for HubSpot implementations - cleaning and deduplicating CRM data, structuring selling profiles by segment, and connecting the right data providers so tools like the Prospecting Agent work from accurate information instead of around it.
Leave a comment