Lower cost custom agents

Smart CRM & Sales Hub Breeze & Other AI-Powered Features
,
1

Yesterday HubSpot announced that they are lowering the cost of running custom agents by 70%. I’m not sure if this is a result of internal optimization or a straight price cut. There is also a cost estimation tool now in the agent builder.

Any early thoughts on where this would change the cost vs. value equation for you wrt using custom agents to automate a workflow?

One area I am looking at is a ‘new lead pre-call briefing agent’ that does company and buyer research for the rep.

3

Hi @Jignesh-at-Klarion,

Welcome to the Community!

Thanks for sharing this here! The builder lets you test an agent without using credits and review estimated credits per run before you publish, so that could be a good way to get a feel for the numbers. :slight_smile:

This article walks through that process: Create and customize agents in the agent builder

For teams deploying custom AI agents: what framework are you using to decide if a workflow is worth the credits?

Hey @danmoyle, @hannahclee, @chawkins – do you have any experience with the above?

Thanks in advance!! :x_ray: :fallen_leaf: :maple_leaf:

Sam, Community Manager

4

Hey @Jignesh-at-Klarion, I’ve been setting up custom agents for clients since they started using credits in July. The 70% drop changes things for one type of use case in particular. Would love to share the framework i use…

  1. Cost per run vs. rep time. Test the agent in the builder first (it’s free) and note the estimated credits per run. A pre-call brief saves a rep 15-20 mins of research. Even before the 70% drop that was worth it, and now it’s an easy yes.
  2. Control how often it runs. This is where the real cost sits. Don’t trigger it on every new contact. Trigger it on a booked meeting or a new deal. Then set a monthly run limit on the agent (Edit run limit) and compare actual usage against the estimate after the first month.
  3. Pick the right tool. For a simple summary of one record, a workflow AI action or the data agent costs less. Custom agents make sense for multi-step work like your briefing use case (company research + buyer context + CRM history > talking points).
  4. Check that reps actually use it. After 30 days, see if reps open the briefs and if those calls move faster. If they don’t, the agent isn’t worth it at any price.

For your use case: trigger on meeting booked, keep the brief to 5-6 bullets, write it to the contact record, and start with a capped run limit. That gives you real cost and value numbers within a month.

5

The pre-call briefing is a strong test case because the value is easy to measure in time saved per meeting. A useful next step is to compare the estimated credits per run with the minutes reps actually save, then set a monthly cap so costs stay predictable. If reps consistently use the brief, that result makes a clear case for expanding the agent to other high-value moments.

6

Building on what hubspotfan2 said about comparing estimated credits against the minutes reps actually save — one thing I’d add is to measure the variance, not just the average. Briefings that pull in a lot of company data can swing widely in credits per run, so track the 90th-percentile cost alongside the median before you scale. If the expensive outliers happen to be the meetings that matter most, the average will mislead you.

Another angle worth testing first: run the same brief through a workflow AI action for a week and compare output quality head-to-head. For a single-record summary, the workflow step often lands close enough at a fraction of the credits, and the comparison gives you a defensible answer when someone asks why this workflow deserves a full custom agent.