Mike HeapMike is an experienced Product Manager who focuses on all the “non-development” areas of My AskAI, from finance and customer success to product design, copywriting, testing and more.
Table of Contents
- HubSpot Breeze vs Forethought at a glance
- What are HubSpot Breeze and Forethought, and who owns each one?
- How does an AI support agent work inside a CRM versus as a standalone platform?
- How can I use these AI agents: replies, copilot, or triage?
- Direct replies
- Copilot replies
- Triage and routing
- Voice
- Which is easier to set up: HubSpot Breeze or Forethought?
- How do HubSpot Breeze and Forethought differ in what they can be trained on?
- 'Static' content
- 'Dynamic' content
- Which has better answer quality, Forethought or HubSpot Breeze?
- Is HubSpot Breeze or Forethought easier to improve?
- Which has more features: Forethought or HubSpot Breeze?
- How easy is it to customize HubSpot Breeze and Forethought?
- What about vendor lock-in?
- Do they have any other AI features?
- Tagging
- Agent translation
- Sentiment and CSAT
- What about security, is Forethought more secure than HubSpot Breeze?
- Which costs more, HubSpot Breeze or Forethought?
- The pricing model
- The overall cost
- Conclusion - should I choose HubSpot Breeze or Forethought?
- FAQs
Created time
Aug 28, 2026 05:50 AM
Title length (<60)
Author
Last optimised
Ecomm?
Tags
Image
Publish date
Aug 9, 2026
Video
Slug
hubspot-breeze-forethought-comparison-2026
Featured
Type
Article
Ready to Publish
Ready to Publish
Forethought takes seven of the eleven scored categories, HubSpot two, with two level. HubSpot's agent is what most readers will end up buying, on the two categories it wins: $0.50 a resolution, and a setup you can do yourself.
HubSpot's customer agent is the one most support teams can reach: $0.50 per resolution, metered against a paid Professional or Enterprise hub you likely already hold, plus credits and a seat assigned to it. Forethought is the deeper platform (triage and QA ship as products of their own) and the harder purchase: no self-serve signup, a 20,000-ticket history recommended behind it, and an implementation measured in months. The choice is between a metered line item and a five-figure annual contract.
So the first question is which one you can actually buy.
Most support leaders I talk to are already on HubSpot. That one fact settles more of this decision than any feature comparison.
HubSpot's agent switches on inside a CRM you may already be paying for (Professional or Enterprise, on one of seven hubs). Forethought is bought, installed and tuned on top of the desk you have.
Forethought's own integration catalog lists HubSpot. A HubSpot shop can therefore run either product.
HubSpot's customer agent deploys into HubSpot's own channels and nowhere else. So a Zendesk, Intercom or Gorgias shop can only run one of the two (Forethought).
Disclosure: I co-founded My AskAI, so I obviously have a horse in this race. We run AI customer service for 200+ ecommerce and SaaS businesses, and our agents have resolved >1,000,000 tickets.
Across that customer base we hold a 72%+ resolution rate on a rolling 30-day basis. Neither product here is ours, and both are scored on their own published pages and documentation.
I scored every feature on its own merits first, then asked separately whether you can actually deploy it inside your stack.
HubSpot wins on setup and on price. In my experience those are the two things most buyers decide on first.
HubSpot Breeze vs Forethought at a glance
HubSpot Breeze (the customer agent) | Forethought (Solve / Triage / Assist) | |
What it is | CRM-native autonomous support agent | Standalone multi-agent AI support platform |
Headline price | $0.50 per resolution, on top of a plan floor | No published price; quote-only |
How you buy it | Self-serve, once three gates are cleared | Sales-gated; bought from Zendesk Sales |
Best for | Teams already on HubSpot Professional or Enterprise | Enterprises with 20,000+ historic tickets and a CX ops team |
Not a fit for | Any team not on a paid HubSpot Professional or Enterprise hub | Small teams, thin ticket history, anyone needing a price before a call |
Biggest limitation | Runs only inside HubSpot, and cannot be deleted once created | Its pricing page publishes no price or outcome definition |
Scorecard | 2 of 11, 2 tied | 7 of 11, 2 tied |
What are HubSpot Breeze and Forethought, and who owns each one?
TL;DR: HubSpot's agent belongs to HubSpot and arrives as part of Agent Hub. Forethought belongs to Zendesk, which bought it and now sells it through its own sales team, while the product still installs on competing helpdesks.
The two products have different owners, and one of those owners (Zendesk) sells a competing helpdesk.
Forethought, the customer-service AI vendor behind Solve, Triage and Assist, belongs to Zendesk. The acquisition closed on 26 March 2026. Buying it took longer: purchase availability for Forethought AI agents did not open until about ten weeks after the deal closed.
Zendesk documents the product as one that can run away from Zendesk's own platform, which I did not expect from an acquirer. From Zendesk's help center:
"operate independent of the Zendesk platform to help any team automate customer service"
The same article says "Forethought plans are available for purchase for any customer", and the route it points you at is Zendesk Sales.
You buy this from Zendesk's sales team. It is a Zendesk-owned product rather than a marketplace listing you install yourself.
On the HubSpot side, the naming has moved on. Breeze is no longer the umbrella name.
HubSpot's own knowledge base now calls the layer Agent Hub, and the old Breeze explainer URLs redirect there (Breeze now survives only as the assistant's name). The deployable support product is called the customer agent, and Breeze Assistant survives as the rep-facing surface.
Buyers still search for "Breeze". The thing you actually deploy is HubSpot's customer agent.
Both products review well on G2:
HubSpot | Forethought | |
Rating | 4.4/5, on the HubSpot Service Hub listing rather than a Breeze-specific page | 4.3/5 |
Reviews-page link |
How does an AI support agent work inside a CRM versus as a standalone platform?
TL;DR: HubSpot's agent runs on CRM records and reads deal stage, lifecycle stage and ticket history with no connector. Forethought sits above whichever helpdesk you already run and trains a model on your own ticket history. That structural difference decides which of the two you can run at all.
HubSpot's agent lives where your data already is. It reads deal stage, lifecycle stage, ticket history, custom properties and record owner, with no connector to configure.
It also inherits the CRM's permission model. What a user cannot see, the agent cannot see either (which saves you a whole conversation with your security reviewer).
Forethought trains a separate model per customer, on that customer's own data. Your model never sees anyone else's tickets.
Personal and health information is stripped out before training starts. At answer time it uses retrieval, which just means it looks up your own articles and tickets before it writes anything.
Both handle uncertainty in much the same way. The exit path is the first thing I check on any agent. HubSpot's agent either gives a sourced answer, asks a clarifying question, or hands the conversation to a human.
Forethought's integration catalog lists HubSpot among 23 helpdesks and CRMs, so it installs on top of the CRM.
HubSpot's customer agent installs into HubSpot. So the shortlist you can actually write depends on which desk you already run.
How can I use these AI agents: replies, copilot, or triage?
TL;DR: Both run an autonomous mode and a copilot mode. Forethought adds a third, a triage agent that classifies and routes tickets before anyone answers them, and its copilot runs as a browser extension over whichever helpdesk you use.
Direct replies
Autonomous replies are what both products are bought for. HubSpot's deployment documentation is specific about where the agent can go:
"You can assign the agent to WhatsApp, Facebook, calling (BETA), live chat, forms, email channels, or custom channels."
HubSpot's marketing pages describe the same thing more loosely as chat, email, voice and social. Where the two disagree, I take the documentation.
Forethought's Solve covers chat, email, voice, SMS, Slack, mobile and a headless API. The channels are tier-gated.
Team gets chat and mobile. Professional adds email, voice and Slack, and the Solve API is Enterprise only.
HubSpot has one rollout control I like a lot. You can set the agent to take a stated percentage of conversations, so you can start it on a slice of the queue.
Copilot replies
Breeze Assistant is the HubSpot side of this, and it sits inside the HubSpot UI your reps already work in (no second tab, nothing to install).
Forethought Assist is a Chrome browser extension that overlays the helpdesk. Working across desks is what the extension buys you.
For a team that lives in a native inbox panel, an overlay is a different habit to learn, and in my experience that training cost never makes it into the business case.
Triage and routing
Only one side has this as a product. Forethought's Triage classifies incoming tickets by sentiment, intent, spam, urgency, language and product type, then routes on the result.
Ready-made triage models are included from the Team tier. Custom handoff models are capped at 3 on Team and 6 on Professional (check that ceiling against your own ticket taxonomy before you sign).
HubSpot covers pieces of the same ground. It generates ticket names automatically and detects the language of an incoming ticket for routing.
Voice
HubSpot's calling channel is marked BETA in the documentation, so I treat it as early. Forethought's voice channel sits inside Solve and is gated to Professional and above.
If voice is on your roadmap, Forethought's voice channel is further along and costs more to reach. On the calls I sit in, voice is usually last on the list.
Which is easier to set up: HubSpot Breeze or Forethought?
TL;DR: HubSpot, decisively. Its wizard takes minutes once you reach it, though three gates stand in front of it. Forethought has no self-serve signup at all and quotes 30 to 90 days.
There are three gates in front of HubSpot's wizard, and the setup documentation is clear about all of them. First, a plan: Professional or Enterprise on any of seven hubs, covering Marketing, Sales, Service, Data, Content, Smart CRM and Revenue.
Free and Starter cannot run it on any hub at all. That gate disqualifies more of the teams I talk to than anything else on the list.
Second, credits. "HubSpot Credits are required to deploy the customer agent to channels." Third, a seat: "An assigned seat is required to create, edit, and manage the customer agent."
Clear those three and setup is fast. It is a named wizard step or two, with a free preview so you can test answers before anything reaches a customer.
HubSpot's own partner customers report setting it up in under an hour, against a week for the chatbot it replaced (vendor-published numbers, so read them as a best case). One agency reports being live inside 30 days.
There is a third route, which is to choose the AI layer separately from the desk. My AskAI installs into the HubSpot you already run and the AI layer stays a decision of its own.
Forethought is a different kind of project. There is no self-serve signup, and its FAQ quotes 30 to 90 days to full setup.
It recommends 20,000 or more historic tickets plus 2,000 monthly email or chat tickets for the model to perform (a floor that rules out a lot of teams reading this). Every customer is paired with a CSM, and Grammarly is the published fast outlier at around a week and a half.
There is no free trial either. From Forethought's pricing page:
"Yes, we offer a Proof of Value (POV) instead of a traditional free trial."
The free preview is HubSpot's real counterweight here. In my experience it is what puts real answers in front of a stakeholder before anyone signs anything.
Forethought's equivalent is a Proof of Value run alongside a CSM. That is a scheduled project with a start date on it.
How do HubSpot Breeze and Forethought differ in what they can be trained on?
TL;DR: Forethought, by a wide margin. It connects to Notion, Confluence, Google Drive, SharePoint, Snowflake and an MCP connector. HubSpot's agent reads HubSpot-hosted content, uploaded files and a capped public URL crawl.
Source type | HubSpot's customer agent | Forethought |
Help center and KB articles | ✅ HubSpot-hosted, public and private | ✅ 18 knowledge and LMS systems |
Historic tickets | — Auto-drafts articles via the separate Knowledge Base Agent | ✅ Fed directly into training the resolution model |
Public URL crawl | ✅ Up to 5,000 URLs per domain | — Content arrives through connectors |
File uploads | ✅ Around 20 formats | — Through Drive, SharePoint and S3 |
Notion | ❌ | ✅ |
Google Drive, OneDrive, SharePoint | ❌ | ✅ Google Drive and SharePoint |
Confluence, Guru, Glean | ❌ | ✅ |
Shopify and commerce data | ❌ | ✅ |
Snowflake and Amazon S3 | ❌ | ✅ |
MCP connector | ❌ | ✅ |
Mark a source internal-only | ✅ Private sources answer without appearing in citations | — |
'Static' content
HubSpot's content-sources documentation sets out what the agent reads. It covers HubSpot-hosted knowledge base articles, both public and private, plus website pages, landing pages and blogs.
Add around twenty upload formats and short question-and-answer pairs (the Q&A pairs are the quickest win on a thin knowledge base). The public URL crawl carries a stated cap:
"Up to 5,000 URLs within the domain will be automatically crawled and synced"
Uploads are static in the literal sense. Change the PDF and you re-upload the PDF.
HubSpot's customer-agent documentation names no connector for Notion, Google Drive, OneDrive, Confluence, SharePoint or any third-party help center.
That is the customer agent's own list. The wider App Marketplace is a separate question.
If your answers live in a wiki, that gap is the thing to solve first. We built My AskAI to connect natively to Notion, Google Drive, OneDrive and Confluence, and any source can be marked internal-only so it powers your team's answers without ever appearing to a customer.
'Dynamic' content
Forethought's catalog is where its lead comes from. Eighteen knowledge and LMS systems are in there, including Notion, Confluence, Guru, Glean, Document360, ReadMe and Shelf (if your wiki is on that list, this is where Forethought pulls ahead).
Its connectors reach Google Drive, SharePoint, Amazon S3, Snowflake, Shopify, Airtable and WordPress, plus an MCP connector. Historic tickets, macros and canned responses all count as training inputs.
That breadth has a cost attached. The model needs volume to work, which is where the 20,000 historic ticket recommendation comes from.
A thin ticket history is where that architecture underperforms.
If you have neither the wiki nor the ticket volume, that is the gap our Train on Historic Tickets feature closes. My AskAI backfills the last 5,000 historic tickets by default and drafts starter knowledge out of them, so you can start with nothing written down.
Sync cadence differs too. On HubSpot, knowledge base updates sync immediately, and "Other content sources are automatically re-synced on a weekly basis."
A week is a long time if you are shipping pricing changes (we push knowledge updates most days).
Which has better answer quality, Forethought or HubSpot Breeze?
TL;DR: Both publish named-customer numbers in a similar band, both define the metric themselves, and both headline figures sit at the top of their own range. Neither set of rates is measured the same way, so there is nothing to score one against the other with.
HubSpot's customer agent page puts one aggregate above the case studies: 70% of conversations resolved automatically, with top teams hitting 90%. That is HubSpot's own figure for its own customer base, self-reported and self-defined.
Underneath it sit named customers. Sticos, a Visma company, is quoted as "handling 91% of all initiated chats, with a 75% resolution rate", attributed to its customer success manager.
Nutribees reports "We have lowered the number of tickets handled by customer support by 77%". HubSpot's dedicated Sticos case study frames the same deployment as automating 41% of inquiries and deflecting 91% of chats, which are two more measures again.
Forethought's published numbers sit in a similar place. Grammarly is written up at 87% deflection with 4.2 CSAT, and Upwork reports 52% to 65% self-serve on chat plus 90% classification accuracy.
Its Solve page headlines a resolution rate of up to 98%. That figure is a marketing ceiling.
Both vendors define the metric themselves (HubSpot in writing, Forethought not at all), and the two definitions are not measuring the same event. I read both as vendor claims, because the number moves with whoever writes the definition.
HubSpot's is written down, on the same page as the rate:
"A resolution means support was provided by the agent, and the conversation was not handed to a human rep for 72 hours."
Forethought publishes deflection figures per customer without publishing what qualifies. So a like-for-like comparison of the two rates is not available, and anyone offering you one is guessing (I include the analyst round-ups in that).
For context on where the field sits, our AI resolution rate benchmark study at My AskAI puts the published median at 70% across 195 rated deployments from 38 vendors. Read that as a directional aggregate.
The match with HubSpot's own 70% is a coincidence of arithmetic and nothing more. One is a single vendor's average over its own customers; the other is the midpoint of 195 deployments published by 38 different vendors.
Vendors self-select what they publish. A billing queue and a technical-support queue score differently on the same model, so neither figure tells you what yours will be.
Is HubSpot Breeze or Forethought easier to improve?
TL;DR: Forethought includes more improvement machinery, including gap detection, auto-generated articles and QA scoring across every interaction. The best of it sits on the top tier. HubSpot's Knowledge Base Agent is narrower, and gated to Service Hub Professional and Enterprise.
Forethought's Discover analyzes historic tickets and help center content to find where answers are missing (the gap-hunting most teams still do by hand). It then drafts both articles and Autoflows to close them.
Agent QA scores 100% of interactions for empathy, grammar, relevance and resolution quality, against rubrics you write yourself.
Both sit behind tier limits that change the budget. Rubrics run 1 on Team, 5 on Professional and 20 on Enterprise (one rubric is not a QA program).
The detection-and-drafting half of Discover is Enterprise-only, listed under that tier's inclusions.
HubSpot's answer is narrower and simpler. The Knowledge Base Agent drafts articles from resolved tickets, and from questions the customer agent could not answer (the second source is the useful one).
It runs a detection, analysis, generation and validation pipeline, and scans for updates once a week. Setting it up is two dropdowns, which I mean as praise.
HubSpot's quality tooling is built around calls. The Service Hub price list carries conversation intelligence with tracked terms, AI transcript enrichment and Coaching Playlists, which is a review surface built around recorded calls.
What it does not put next to Forethought is rubric scoring on every ticket (yes, even the ones nobody escalates), and that is the thing a support QA program gets built on.
Forethought wins this category on capability. HubSpot's narrower version needs nobody to own it.
The cost sits on the Forethought side. Every part of that machinery is something a person has to own, from the rubrics you write yourself to the intents that have to be trained before a workflow can sit on top of them.
The first thing I check is who owns the rubrics and the trained intents. Both go stale fast without a name against them.
Which has more features: Forethought or HubSpot Breeze?
TL;DR: Forethought has the broader support-specific surface, covering triage, QA, gap detection, a workflow canvas and browser agents. HubSpot's breadth runs sideways into marketing and sales agents a support team will not use.
Capability | HubSpot's customer agent | Forethought |
Autonomous replies | ✅ | ✅ |
Rep copilot | ✅ Breeze Assistant, native | ✅ Assist, a Chrome extension |
Ticket triage and routing | — AI ticket names, language detection | ✅ A dedicated Triage agent |
QA scoring on interactions | — Call-focused conversation intelligence, transcript enrichment and Coaching Playlists | ✅ Every ticket, on 1, 5 or 20 rubrics by tier |
Gap detection and article drafting | ✅ Knowledge Base Agent, on Service Hub Professional and Enterprise | ✅ Discover, on Enterprise |
Workflow canvas | ✅ No-code agent builder | ✅ Orchestrator |
API-triggered actions | ✅ | ✅ Action Builder, any public API |
Browser-based actions | ❌ | ✅ Browser Agents |
Headless or API embed | — Custom channels | ✅ Solve API, on Enterprise |
Multi-brand | ✅ Independent tone and knowledge | ✅ 2 on Team, 20 on Professional |
Voice | ✅ Calling, marked BETA | ✅ Professional and above |
CRM record context without a connector | ✅ | — Reads CRM data through connectors |
Marketing and sales agents | ✅ | ❌ |
Forethought's Browser Agents are the most unusual capability either vendor offers. They act inside browser-based systems that have no API at all, working the interface the way a rep would.
The policy is written in plain language, and you get a reasoning trace you can inspect on a test run (the trace is what your ops lead will want before it touches a refund). If a chunk of your workflow lives in an old internal tool, that is a real capability, and HubSpot has no equivalent.
Orchestrator is a workflow canvas: an action, then agent instructions, then if-otherwise branching.
Typed outputs, run counts and live success percentages sit against each automation (visible success rates are rarer than they should be).
HubSpot's breadth runs into other departments. Agent Hub includes pre-built agents for AEO, data, prospecting and deal progression alongside the customer agent (none of which answer a support ticket).
The list runs on through Breeze Assistant, the Knowledge Base Agent, a Customer Success Agent, Breeze Intelligence enrichment, and multi-brand agents with their own tone and knowledge. API-triggered actions, guardrails and a no-code agent builder come with it.
Most of that a support team never opens.
HubSpot's own setup page carries one line to read before you switch anything on:
"once a customer agent is created, it can’t be deleted."
You can pause it. You cannot remove it, so I create the first one in a sandbox portal.
A custom agent built in the agent builder can be deleted permanently. Those are two different objects, and it is easy to read one as the other.
Forethought and Zendesk publish different add-on lists for the same product.
Forethought's pricing page lists three add-ons: Multibrand, Analytics API and Discover. Zendesk's help center labels seven, adding Assist, Triage, Agent QA and AI QA.
Both are official. Either way, ask which list applies to the contract you are being quoted.
How easy is it to customize HubSpot Breeze and Forethought?
TL;DR: Forethought. Autoflows are business logic written in plain language, and the Action Builder reaches any public API. HubSpot's Guidelines govern how the agent talks.
I think Guidelines, HubSpot's customization surface, gets less credit than it deserves. From the documentation:
"Guidelines provide instructions for how the customer agent communicates."
Behind that sit four free-text fields the admin writes into: Tone, Response style, Scripted responses and Guardrails.
Five personality presets give you a starting point you can write straight over, and there are templates if you want a scaffold. The fields are free text, so you write your own instructions into them.
Guidelines govern communication: how the agent sounds, what it will and will not say, and the wording it must use verbatim.
HubSpot's Guidelines page stops short of per-scenario policy routing. I hear that request from teams as soon as the first policy exception comes up.
Forethought's customization reaches into business logic. Autoflows are written in natural language, with no decision tree to build, and Forethought claims you can write one in minutes.
The Action Builder connects to any public API, so order lookups and refunds are in scope (refunds are the one I get asked about most). The Workflow Builder needs intents trained first, and Orchestrator's node graph sits above the lot.
Forethought has no simulation mode and no live preview for workflow builders, which is exactly the job HubSpot's free preview does.
So you write the business logic, then watch it run on live conversations, which is not where I like to find out that a policy was wrong.
What about vendor lock-in?
TL;DR: HubSpot's customer agent runs only inside HubSpot. Forethought installs on 23 helpdesks and CRMs, HubSpot among them, and Zendesk's own documentation says it can operate independent of the Zendesk platform.
HubSpot's deployment page enumerates where the agent can go, and every destination on it is a HubSpot surface:
"Once you've set up the customer agent, you can assign it to Facebook, WhatsApp, calling (BETA), live chat, form, or email channels."
No third-party helpdesk is named as a destination anywhere in HubSpot's customer-agent documentation. HubSpot's helpdesk integrations sync data; they do not deploy the agent into another desk.
Contracts are annual, and credits do not roll over from month to month. HubSpot's deployment page adds that if you have bought extra credit capacity packs, your credits "may be automatically upgraded if your usage exceeds your current allotment".
Add the fact that the agent cannot be deleted, and the commitment runs to the end of the contract.
If portability is what you are protecting, that is the case for keeping the agent layer separate from the desk. We run My AskAI inside Zendesk, Intercom, Freshdesk, Freshchat, Gorgias and HubSpot, and the trained agent moves with you, so changing helpdesk does not mean rebuilding the AI.
Forethought's catalog covers the competing desks. It still lists Intercom, Gorgias, Freshworks and Freshchat, Front, Gladly, Help Scout, Kustomer, ServiceNow, Salesforce and HubSpot, alongside its owner's product.
That is how the catalog reads now, with no promise about the future (catalogs move after an acquisition).
One risk is still open. The acquisition completion release says nothing about standalone availability, or about how the product will be packaged going forward.
The line about remaining available to non-Zendesk customers came from the original announcement and has not been restated since. Payment terms are annual only, and procurement data shows buyers routinely negotiating renewal uplifts out of the contract.
Do they have any other AI features?
TL;DR: Forethought's extras are support-shaped, covering triage classification, QA scoring on every interaction and content insights. HubSpot's extras are platform-shaped, covering ticket summaries, enrichment, and agents for marketing and sales.
Tagging
Forethought's Triage is the tagging layer, and it classifies on sentiment, intent, spam, urgency, language and product type. Forethought claims 90% classification accuracy, and Upwork's case study reports hitting that number (a vendor's own case study, so read it as the ceiling).
HubSpot's equivalents are lighter: AI-generated ticket names, conversation and ticket summaries, automatic language detection for routing, and call summaries.
That is useful admin. HubSpot has no dedicated classification agent behind it.
Agent translation
Forethought's pricing page lists multilingual support in the Team tier. Older write-ups place it higher up the ladder, and I had it filed a tier up myself. The Solve chat widget detects the language a customer opens in and can switch mid-conversation.
The same page also carries a separate Multilingual Communications block alongside the tiers, so ask which of the two you are being quoted.
Forethought's Solve page headlines 100 or more languages supported. Which languages that headline covers is the thing to get confirmed in writing.
HubSpot supports every language HubSpot itself supports, and it detects the language of an incoming ticket for routing purposes.
Sentiment and CSAT
Forethought collects CSAT inside Solve and scores every interaction in Agent QA, so sentiment comes out as a measured score. HubSpot's version is shallower, surfacing inside call summaries and ticket enrichment (context, with no scored CSAT behind it).
What about security, is Forethought more secure than HubSpot Breeze?
TL;DR: Forethought carries more certifications on paper, including HIPAA and SOC 2 Type II, with redaction on by default. HubSpot's advantage is architectural, because the agent inherits the permission model of the CRM it lives in.
Forethought's security page is the longer list. SOC 2 Type II with annual audits, HIPAA certification, GDPR and CCPA, and alignment with the NIST cybersecurity framework and 800-53 (HIPAA is the one that shortens a healthcare review).
It encrypts with AES-256 at rest and TLS in transit. Role-based access control, customer-specific data segmentation and single sign-on are on the same list.
There is a real-time trust report with audit logs, plus a bug bounty. Its data processing agreement and security policy are both public.
Two qualifiers your reviewer will want. Forethought's materials say it aligns with ISO 27001, which is a different claim from holding the certification (alignment is self-asserted; certification is audited).
Forethought's Enterprise tier lists "Enterprise security and governance controls" inside its everything-in-professional block. What those add over the standard set is not published, so ask before you assume.
Redaction is the feature to put in front of a security reviewer. I have seen it close a review on its own. Personal, health and financial data is stripped automatically before it reaches the model, with nothing to switch on.
HubSpot answers a narrower question, and its architectural point is permission inheritance. The agent sees exactly what the user it is acting for can see, and nothing else.
That removes a whole class of exposure without anybody configuring anything. HubSpot's AI trust page publishes model cards and a zero-data-retention commitment (a checklist item, with no detail behind it).
HubSpot's trust page runs to model cards, a retention commitment and little else.
Which costs more, HubSpot Breeze or Forethought?
TL;DR: For almost every reader, Forethought. HubSpot publishes $0.50 per resolution on top of a seat and onboarding floor. Forethought publishes nothing, and third-party procurement data puts the median annual contract in the mid-seventy-thousands.
The pricing model
HubSpot charges $0.50 per resolution, billed as 50 credits, and it publishes what a resolution is:
"A resolution means support was provided by the agent, and the conversation was not handed to a human rep for 72 hours."
A resolution can also be counted when the agent qualifies a lead. Under that headline sits the Service Hub price list.
Professional starts at $90 per seat per month if you commit annually, or $100 if you pay monthly, plus a one-time $1,500 onboarding fee. The Professional figure renders with promotional strikethrough markup on the page, so treat it as a live promotion that can move.
Enterprise starts at $150 per seat per month with a one-time $3,500 onboarding fee (that fee is a year-one line of its own, and it is easy to leave out of a first model).
Credit pools run 500, 3,000 or 5,000 a month, and they do not roll over. HubSpot prices credits themselves at "$9.00 per 1,000 credits when you pay annually".
Two structural traps hide under that clean per-unit number. Credits expire monthly.
And if you have already bought extra credit capacity packs, HubSpot says going over your allotment can upgrade those credits automatically.
The 14-day free window has a condition attached. It is a credit holiday on a paid seat, offered on the purchase of at least one Service Professional or Enterprise seat (a trial you pay for, in other words).
Forethought publishes no dollar figure. Every tier and every add-on on the pricing page carries a Get a Quote button where a number would be, and the model gets one sentence:
"Our pricing model is a blend of platform access fees and an outcome-based pricing cost."
Forethought markets outcome-based pricing while publishing no rate, no definition of what qualifies as an outcome, and no method for verifying one.
The one unit figure in circulation is negotiation guidance from a procurement marketplace, which says per-conversation pricing for Solve often ranges from $0.50 to $2.00 and moves with volume. That is a different denominator entirely (a conversation can contain no resolution at all), and it is guidance for a negotiation, never a published rate.
The unit being marketed and the unit being billed do not reconcile without a sales call.
The overall cost
Take 10,000 tickets a month, half of them resolved by the AI, on a 20-seat support team.
Monthly tickets | AI resolutions at 50% | Resolution charge | Professional seats (20 at the $90 annual rate) | Monthly total | Effective cost per resolution |
10,000 | 5,000 | $2,500 | $1,800 | $4,300 | $0.86 |
The $0.50 headline lands at $0.86 once the seats it requires are in the calculation, and that is on the $90 annual-commitment rate. Pay monthly at $100 a seat and the same month is $4,500, or $0.90.
Add a one-time $1,500 onboarding fee in year one, and $0.86 is the number I take into a budget conversation.
$0.86 is still a competitive number for a native CRM agent.
On the Forethought side there is no arithmetic to do, because there is no published rate to multiply. The only evidence is procurement data (observed purchases, gathered after the fact).
Vendr's marketplace listing puts the median annual contract at around $74,483 across 69 observed purchases, in a range from $35,670 to $151,400. Payment terms are annual, with an average negotiated saving of 22.4%.
That is a median contract value, and Forethought publishes no list price to hold it against. Vendr's page was also stamped before the acquisition closed, so the figure predates Zendesk's ownership (and Zendesk's pricing team).
HubSpot's agent adds a metered line item to a bill you already receive.
Forethought is an annual contract that lands in the mid-five figures after a negotiation.
On the HubSpot side, budget the total cost. The seat floor, the onboarding fee and the credit auto-upgrade all sit under that $0.50.
We charge per ticket for the AI itself, whether or not it resolves: $0.10 a ticket on Scale and $0.12 on Pro, each on top of that plan's monthly base. The AI line then stays put as the resolution rate climbs. You can run it for 30 days with every feature unlocked, unlimited tickets and no card, on the HubSpot seats you already pay for.
We have full pricing explainers for HubSpot's customer agent and for Forethought if you want the arithmetic in detail, plus an ROI calculator for HubSpot.
Conclusion - should I choose HubSpot Breeze or Forethought?
TL;DR: Forethought wins seven of eleven categories and is the stronger platform. It is also the one most readers cannot buy. If you are on HubSpot Professional or Enterprise and want an agent running this week, HubSpot's is the right answer anyway.
Forethought is the better product, and HubSpot's agent is the better answer for most people reading this. The condition that flips it is ticket volume with a CX ops team behind it.
Above that line, Forethought's machinery starts paying for the sales cycle it costs you.
The eleven categories are modes, ease of setup, training and integrations, answer quality, improving, features, price, customization, lock-in, other AI features and security (voice is scored under modes). Evidence comes from each vendor's own documentation and published pricing pages, Zendesk's help center, named customer stories on both vendors' sites, G2 ratings and procurement data on Vendr.
HubSpot Breeze | Forethought | ||
Modes | 7/10 | 7/10 | Tie |
Ease of setup | 8/10 | 4/10 | HubSpot win |
Training/Integrations | 4/10 | 9/10 | Forethought win |
Answer Quality | 7/10 | 7/10 | Tie |
Improving | 7/10 | 8/10 | Forethought win |
Features | 7/10 | 8/10 | Forethought win |
Price | 8/10 | 3/10 | HubSpot win |
Customization | 6/10 | 9/10 | Forethought win |
Lock-in | 4/10 | 6/10 | Forethought win |
Other AI features | 6/10 | 8/10 | Forethought win |
Security | 7/10 | 8/10 | Forethought win |
That is 7 category wins of 11 for Forethought, 2 for HubSpot and 2 tied.
Capability and reachability are not the same axis. Forethought's seven wins are capability wins, and they are real.
HubSpot's two are ease of setup and price, which is to say: can I get it, and what does it cost. Those are the two questions a buyer answers first.
For most teams reading this, they are the only two that get answered at all.
A quote-only platform with a 20,000-ticket floor and a three-month implementation is a different purchase from a $0.50 line item.
So your ticket history, your team and your desk decide this.
Choose HubSpot's customer agent if:
- You already pay for Professional or Enterprise on one of the seven qualifying hubs
- You want the agent answering customers within days
- Your answers already live in HubSpot's knowledge base and website content
- You want CRM context in every reply with nothing to configure
- You want a published per-unit price you can model before you commit
Choose Forethought if:
- You have 20,000 or more historic tickets and 2,000-plus a month coming in
- You run a helpdesk that is not HubSpot, or you want the agent to outlive the desk
- Your knowledge lives in Notion, Confluence, Google Drive, SharePoint or Snowflake
- You need ticket triage and QA scoring as products in their own right
- You have a CX ops team who will own tuning and a security reviewer asking for HIPAA
There is a third route. My AskAI installs into the HubSpot you already run, and charges per ticket on top of a monthly plan base.
If your shortlist is wider than two, we have a roundup of HubSpot Breeze alternatives and a matching one for Forethought. We also compared HubSpot's agent against Freshdesk's Freddy, and Forethought against Decagon, if you want another pair to triangulate from.
FAQs
Who owns Forethought?
Zendesk. The acquisition closed on 26 March 2026, and purchase availability for Forethought AI agents opened about ten weeks after that.
Zendesk's help center documents the product as one that can "operate independent of the Zendesk platform to help any team automate customer service". It also states that "Forethought plans are available for purchase for any customer."
The buying route is Zendesk Sales, so you buy it the way you buy any other Zendesk product (there is no self-serve checkout).
Can Forethought replace HubSpot's customer agent inside HubSpot?
Yes. Forethought's live catalog lists HubSpot among 23 helpdesks and CRMs, so it installs on top of the CRM you already run.
HubSpot's customer agent stays home. Its deployment documentation enumerates Facebook, WhatsApp, calling in beta, live chat, forms and email channels, and no third-party helpdesk is named as a destination.
You can choose the agent layer separately from the desk. We run inside HubSpot as that layer (chosen on its own merits), so the AI is a separate decision from the desk you run.
What can each AI agent be trained on?
HubSpot's agent reads HubSpot-hosted knowledge base articles, website pages, landing pages and blogs. Add around twenty upload formats and short question-and-answer pairs (uploads do not refresh themselves).
There is also a public URL crawl capped at 5,000 URLs per domain, plus CRM data for personalization.
Forethought trains on historic tickets, macros and canned responses. Add 18 knowledge and LMS systems, including Notion, Confluence, Guru and Glean. Connectors cover Google Drive, SharePoint, Amazon S3, Snowflake, Shopify and MCP.
How long does setup take and do we need developers?
HubSpot is a no-code wizard, and fast once three gates are cleared: a Professional or Enterprise plan on a qualifying hub, HubSpot Credits, and an assigned seat (Free and Starter clear none of them). Its named customers report setting the agent up in under an hour, and there is a free preview so you can test answers before deploying.
Forethought quotes 30 to 90 days to full setup. There is no self-serve signup, every customer is paired with a CSM, and 20,000 or more historic tickets are recommended for the model to perform (that ticket floor is what rules most teams out).
No developers are required on either. In my experience Forethought's 30-to-90-day window is what makes it a project.
How much does Forethought AI cost, and how does that compare to HubSpot Breeze?
HubSpot publishes its rate, while Forethought quotes on request, so the comparison is between a modeled number and a negotiated one.
HubSpot's customer agent | Forethought | |
Pricing model | Per resolution, on top of seats | Platform access fee plus an outcome-based cost |
Headline rate | $0.50 per resolution (50 credits) | Not published; quote only |
What counts as a resolution | Published: the agent handled it and no human took it within 72 hours | Not published, and no verification method is stated |
Free trial | 14 days of credits, on a paid Professional or Enterprise seat | A Proof of Value instead of a trial |
At 10,000 tickets a month with half resolved by the AI, HubSpot's resolution charge is $2,500. Add 20 Professional seats at the $90 annual-commitment rate and the monthly total is $4,300, which works out at $0.86 per resolved conversation (at the $100 monthly rate it is $4,500, or $0.90).
There is also a one-time $1,500 onboarding fee, which I fold into the first-year number.
On the Forethought side, procurement data puts the median annual contract at around $74,483, in a range from $35,670 to $151,400. That is a median contract value across observed purchases, and Forethought publishes no list price to hold it against.
Vendr's page was also stamped before the acquisition closed, so the figure predates Zendesk's ownership and Zendesk's pricing team.
Can they handle multilingual support and agent translations?
Forethought's pricing page lists multilingual support in its entry Team tier, and carries a separate Multilingual Communications block alongside the tiers, so confirm which one your quote covers. The Solve widget detects the language a customer opens in and can switch mid-conversation.
Its Solve page headlines 100 or more languages supported, which is a count (ask for the actual list in writing).
HubSpot supports every language the platform itself supports. It detects the language of an incoming ticket so it can be routed to the right team.
How do they compare on security and compliance?
Forethought carries the longer list: SOC 2 Type II with annual audits, HIPAA certification, GDPR and CCPA, NIST alignment, AES-256 at rest and TLS in transit.
Add role-based access control, data segmentation and single sign-on, plus automatic redaction of personal, health and financial data before it reaches the model (no add-on, no configuration). Its materials say it aligns with ISO 27001, which is a different claim from holding it.
HubSpot's strength is architectural: the agent inherits the CRM's permission model, so it can only see what the user it acts for can see. HubSpot publishes model cards and a zero-data-retention commitment on its AI trust page.
Forethought's Enterprise tier adds what its pricing page calls "Enterprise security and governance controls". What those add over the standard set is not published, so put that in your security review.
