HubSpot Breeze vs Salesforce Agentforce: which AI layer fits your revenue team

Pricing and prerequisites, side by side

HubSpot Breeze: you pay when the agent succeeds

In 2026 HubSpot moved its two flagship agents to outcome-based pricing. The Customer Agent charges 50 HubSpot Credits, about $0.50, per conversation it resolves. The Prospecting Agent charges 100 credits, about $1, per lead it qualifies and recommends for outreach. Credits cost $10 per 1,000, and the Starter, Professional and Enterprise tiers include 500, 3,000 and 5,000 credits respectively. Both agents ship with a 28-day trial. One correction worth knowing, because it changes the model: HubSpot’s knowledge base was updated in July 2026 to state that data enrichment through Breeze Intelligence does not consume credits; smart properties do.

The consequence is that an unresolved conversation costs nothing. If the Customer Agent hands 28% of tickets to a human, as HubSpot’s own figure implies, you pay for the 72%. That aligns the vendor with your outcome in a way consumption pricing does not, and it makes the bill easy to forecast from a number you already track: resolved tickets and qualified leads per month.

Salesforce Agentforce: you pay when the agent acts

Agentforce prices consumption. Flex Credits cost $500 per 100,000, a standard action consumes 20 credits, or $0.10, and a voice action 30. Customer-facing conversations are metered at $2 each. Per-user add-ons run $125 per user per month, an Agentforce User License is $5 per user per month, and Agentforce 1 Editions bundle the platform from $550 per user per month with 2.5 million credits a year included. Salesforce Foundations gives Enterprise Edition and above Agent Builder, Prompt Builder, 1,000 Agentforce for Service conversations and 100,000 credits at no cost, which is about 5,000 standard actions. Employee-facing agents draw down the same credit pool.

Data 360, formerly Data Cloud, is the line item that surprises people. It is not required for an agent that reads and writes CRM records, but it becomes necessary when an agent must reason over large unstructured content or needs one resolved identity across systems with no shared key, and it is priced separately at $500 per 100,000 credits plus $240 per 1,000 profiles a year. The consequence of consumption pricing is the mirror image of HubSpot’s: an agent that fails to resolve still consumed actions, and a subagent that takes nine actions to do a three-action job costs three times as much, on every conversation, until someone redesigns it.

What each one needs from you before day one

Breeze’s prerequisites fit on an index card. A HubSpot database without gross duplicates, a knowledge base the Customer Agent can answer from, a written ideal-customer profile for the Prospecting Agent, and a person who reviews what the agents did in the first month. The trade-off is scope: the agents do what HubSpot designed them to do. If your support process has a step the Customer Agent does not model, the step does not happen.

Agentforce’s prerequisites are an engagement. The objects the agent touches need a measured duplicate rate and fill rate. The object model needs the unused fields and three-spelling picklists cleaned out, because the agent reads metadata as vocabulary. The agent user needs permission sets matched to a human in the role, and the Manage AI Agents permission set, which Testing Center requires alongside System Administrator, needs a short named list of holders. Each subagent needs its decision rules written as test cases. Every integration the workflow crosses needs a Flow or Apex action with an owner. A third-party comparison from Vantage Point put the setup gap at weeks to months for Agentforce against minutes to hours for Breeze; we would not stand behind their figures, but the direction matches what we see.

How the bill behaves as usage grows

A worked example, on published list prices before the discounts every Salesforce customer negotiates. A support team resolves 1,000 conversations a month through an agent. On HubSpot Professional that is 50,000 credits, less the 3,000 included, or about $470 a month, and the figure scales with resolutions. On Agentforce the same 1,000 customer-facing conversations are $2,000 a month at list, with the first 1,000 conversations covered by Foundations, and the figure scales with conversations whether or not they resolve. A prospecting motion that qualifies 300 leads a month is about $300 on Breeze; on Agentforce it is an employee-facing agent metered by action, so the cost is 300 times the actions per lead times $0.10, which is why Check 7 of any Agentforce pre-flight is actions per outcome.

The pattern generalizes. Breeze is cheaper and more predictable at the volumes a mid-market team runs, because the unit is the outcome. Agentforce is more expensive per outcome at list, and it buys control over what the agent does on the way there, which is worth paying for when the process is genuinely yours and worth nothing when it is not. The side-by-side above is the version we draw on a whiteboard in the first hour of these conversations.