AI only moves the needle when it’s close to your customer data and your daily workflows. Breeze by HubSpot brings three pieces together on one platform: Customer Agent for front‑of‑house conversations, Breeze Assistant for collaborative work, and the Marketplace + Studio to deploy and tailor AI safely at scale.
Customer Agent turns inbound chats and emails into pipeline. It reads and writes to the CRM, recognizes returning visitors, and routes conversations with context so reps don’t start cold.
What changes
Qualify and book directly from chat, email, or messaging
Voice support for people who prefer to talk (voice in HubSpot chat)
Live CRM updates as the conversation unfolds
Smart routing and visitor authentication
Coaching mode to test and refine responses
Where to deploy first
Pricing and comparison pages for high‑intent traffic
Contact and demo pages to reduce form friction
Post‑webinar pages to capture momentum
Knowledge base and support entry points for deflection + expansion
Assistants work with your team on analysis, prep, and content. Breeze Assistant understands HubSpot context and pulls from tools your teams already use (Google Workspace, Microsoft Teams, Slack).
High‑value uses
Meeting prep: roll up account activity, notes, and next steps
Content drafting with brand and product context
BI‑style Q&A on pipeline, accounts, and campaigns
Memory that adapts to preferences over time
Desktop and mobile parity for on‑the‑go work
When to choose assistants vs. agents
Assistants are collaborative and help humans do work
Agents run tasks autonomously and hand off to humans when needed
The Breeze Marketplace is the discovery layer for ready‑to‑run agents and assistants. Breeze Studio is where you configure them, add knowledge, and connect triggers—no code required.
How it comes together
Discover and install agents from the HubSpot agents marketplace
Customize instructions, add tools and knowledge, and set limits in Studio
Automate actions with triggers, similar to workflows
Monitor runs and review outputs to improve quality
Starter agents to consider
Company Research Agent for pre‑call prep
Closing/Handoff/Health agents for late‑stage velocity and CS
RFP or onboarding assistants to standardize complex workflows
Knowledge sources: document what content each agent/assistant can reference
Roles and permissions: who can publish, who can edit, who can run
Review cycles: define when humans must approve outputs
Logging: capture runs and outcomes for QA and training
Credits: plan for features that consume credits (for example, Customer Agent); confirm current policies at launch
Phase 1 — Foundation
Define two charters: one Customer Agent use case and one Assistant use case
Draft instructions, tone, and guardrails
Connect approved knowledge sources; restrict anything sensitive
Pilot on a single page (pricing) and a single team (SDR/BDR)
Phase 2 — Activation
Add voice; enable visitor authentication and routing
Create Studio triggers tied to form submissions and deal stage changes
Ship two marketplace agents and configure in Studio
Add dashboards for booked meetings, deflections, and agent impact on pipeline
Phase 3 — Scale
Extend Customer Agent to support and post‑purchase pages
Add assistant routines for weekly digests and QBR prep
Expand triggers to lifecycle stages and CS milestones
Review transcripts and adjust guardrails; publish a governance guide
Meetings booked from chat and messaging
Lead qualification rate and time to first response
Support deflection and CSAT for service entry points
Pipeline created and influenced by agents
Human review rate and revision time (quality proxy)
Let’s design an AI operating model that actually works, for marketing, sales, and support. Let’s Chat