← ALL BLUEPRINTS / AGENT BLUEPRINT
Website chat agent for a small B2B SaaS team on HubSpot that books demos and routes trial users
A HubSpot-based website chat agent that answers product questions from approved sources, qualifies visitors with a few questions, and books meetings or points self-serve buyers to the trial. It turns heavy pre-sales research traffic into booked meetings and clean CRM records so a 1 to 10 person team gets more pipeline without staffing live chat.
| ROLE | SALES | PLATFORM | HUBSPOT |
|---|---|---|---|
| INDUSTRY | B2B SOFTWARE / SAAS | ||
| DRAWN BY | HACKING DEMAND | ISSUED | OCTOBER 26, 2026 |
What The Concierge does
A small B2B SaaS company (under ten people, founder still closing most deals) gets steady traffic to its pricing, docs and comparison pages, but most visitors leave without talking to anyone. Some should just start a trial. Others are evaluating for a team and would buy faster with a call, and nobody is watching chat to tell them apart. The Concierge works for the sales team (in practice the founder and one or two reps). It owns the first conversation on the website. It answers product, pricing-tier and integration questions from approved content. It asks three to four qualification questions. It sends small or self-serve fits to the trial or signup, and books qualified buyers straight onto the right person's calendar. It writes every conversation, answer and qualification field into HubSpot so nothing gets lost. It does not negotiate, discount, promise roadmap items or handle support for existing accounts. Those go to a person.
Built on HubSpot
The team already runs HubSpot and has no one to staff chat, so the HubSpot Breeze Customer Agent is the right core. It reads the HubSpot knowledge base and site, writes directly to contact and deal records, and hands off to the shared inbox without another system to maintain. That matters for a team this small that will run it themselves. HubSpot falls short in two places for a SaaS motion. It lacks native product usage data and offers only basic round-robin scheduling. A light n8n or Make sync brings trial and activation events into HubSpot properties. Chili Piper handles instant in-chat booking with owner-aware routing. Warmly adds company identification on anonymous research visits. A custom AI build is not needed unless the product questions require live lookups into the app, and that can be added later through an API connector.
What starts it
- A visitor opens the chat widget on any page, or is invited to chat after time on the pricing, comparison or integrations page
- A known contact with an active trial (identified via HubSpot cookie or login) returns to the pricing or plans page
- A visitor from a target-fit company (revealed by visitor identification) lands on a high-intent page
- A visitor submits the demo or contact form and abandons the scheduling step
- Outside business hours, any chat that would otherwise wait for a human reply
What it reads
- HubSpot knowledge base, website pages and product docs, indexed as the agent's only answer source
- HubSpot CRM contacts, companies and deals, including lifecycle stage, owner and past conversations
- Product usage and trial data (signup date, seats, key activation events) synced into HubSpot contact or company properties
- Approved pricing and plan sheet, plus a short list of approved competitor and security answers
- Visitor identification and intent signals from Warmly or RB2B
- Rep calendars and routing rules in Chili Piper or HubSpot Meetings
How it works, step by step
- Lay the foundations in HubSpotBecause nothing exists yet, start with the CRM plumbing. Create the properties the agent will fill (use case, team size, current tool, timeline, trial status, chat qualification result), a lifecycle and lead status scheme, an owner assignment rule, and one deal pipeline stage for chat-booked meetings. Connect the product database or billing system so trial signup and activation events land on the contact and company record, using n8n or Make if there is no native sync.
- Build the knowledge base the agent answers fromCollect the top 30 to 50 questions buyers actually ask: pricing tiers, integrations, security and data handling, setup time, trial limits, and how you compare to the two or three tools people mention most. Write approved answers into the HubSpot knowledge base and site pages, and point the Breeze Customer Agent only at those sources. Mark anything the agent may not answer (custom pricing, contract terms, roadmap) as escalate-only.
- Greet and identifyOn chat open, the agent checks HubSpot for a known contact and Warmly or RB2B for company context. Returning trial users get a greeting that acknowledges their trial. Unknown visitors get a short question about what brought them in, not a form. Email is requested only once the visitor wants something (a booking, a doc, a follow-up).
- Answer questions, then qualifyThe agent answers the visitor's question first, citing the source page, then asks up to four qualification questions woven into the conversation: team size or seats, current tool or process, timeline, and whether they are evaluating for themselves or a team. Each answer is written to the matching HubSpot property as the chat runs.
- Route to trial or meetingUsing rules the founder sets (for example, seats above a threshold, an enterprise need such as SSO, or an active trial with strong usage), qualified visitors are offered a meeting booked inside the chat via Chili Piper or HubSpot Meetings and routed to the right owner. If an account already has an owner, it goes to that owner. Self-serve fits get a direct trial or signup link and a relevant setup guide. Visitors who are not ready get a resource and are enrolled in a short nurture.
- Log, summarize and alertEvery chat creates or updates the contact and company, logs the transcript, and writes a three-line summary (need, fit, next step) to the timeline. Booked meetings create a deal in the chat-booked stage and send the rep a Slack or email alert with the summary and trial usage, so they walk in prepared.
- Weekly review and tuningEach week someone on the team reviews unanswered or escalated questions, wrong answers flagged by reps, and drop-off points in the qualification flow. They add or fix knowledge articles and adjust routing thresholds.
When a person takes over
- Pricing beyond the published tiers, discounts, annual contract terms or procurement questions go to the founder or account owner through a live handoff in HubSpot inbox, or a booked call if no one is online
- Security reviews, questionnaires, data residency or legal questions are escalated to the named owner with the transcript, because a wrong answer can kill or bind a deal
- Existing paying customers asking about bugs, billing or account changes are recognized via CRM lifecycle stage and moved to the support inbox, not the sales flow
- Any question the agent cannot answer from approved sources with confidence is handed to a person instead of guessed at, and the visitor is told when to expect a reply
Rules it follows
- Answer only from the approved knowledge base and published pages. If the answer is not there, say so and offer a person. Never invent features, integrations or limits
- Never quote custom pricing, discounts or roadmap dates, and never imply a feature is coming unless it is on the public roadmap
- Respect the product-led path: never push a meeting on a visitor who clearly wants to self-serve, and never gate the trial link behind qualification questions
- Use trial and usage data only to tailor routing and context. Do not recite a visitor's in-product activity back to them in chat, and only use identification data in line with the site's consent settings (managed with CookieYes)
- Keep a fixed tone guide (plain, short answers, no hype) and cap qualification at four questions so research-heavy buyers are not interrogated
How you know it works
- Meetings booked from chat per week and the share that become qualified opportunities
- Pipeline created from chat-sourced deals, tracked by original source in HubSpot
- Trial signups started from chat and trial-to-paid conversion for chat-assisted trials versus others
- Share of chats resolved without human help, and the escalation rate on questions outside the knowledge base
- No-show rate on chat-booked meetings and rep-flagged answer accuracy issues
From pilot to live
- Phase 1: Foundations and shadow mode (weeks 1 to 3)Set up HubSpot properties, routing, the trial data sync and the knowledge base. Run the agent on the pricing and demo pages only, with every booking and summary reviewed by the founder daily. Fix wrong or missing answers before widening scope.
- Phase 2: Site-wide live with routing (weeks 4 to 8)Turn the agent on across the site, including after-hours coverage. Enable in-chat booking through Chili Piper, visitor identification and returning-trial greetings. Hand day-to-day ownership to a named person on the team, with a weekly 30-minute review of transcripts and escalations.
- Phase 3: Expansion signals and handoff completeAdd rules for trial and existing-customer expansion (seat limits reached, upgrade-page visits) that route to the account owner. Document the routing rules, knowledge update process and escalation list in a short runbook so the team runs and tunes it without outside help.
Tools that fit
HubSpot
CRM and marketing software
Chili Piper
Demand conversion platform
Warmly
Reveal site visitors and buying signals
n8n
Workflow automation platform
CookieYes
Cookies management
Some tool links are partner links. Hacking Demand may earn a commission at no extra cost to you.
This blueprint is a version of The Concierge (website chat agent) ↗
Want The Concierge built for you?
Our experts build it inside your stack. Then we hand it off, maintain it, or operate it with you. Agents and programs get better with tuning, so most teams keep us on to run them. Bring this blueprint to a scoping call.
Scope this agent ↗ Get your own blueprint ↗Good questions.
Straight answers.
How does the agent decide between pushing a trial and booking a demo when we run both motions?
You set explicit rules in HubSpot, such as seat count above a threshold, a need like SSO or a custom integration, or an active trial with strong usage. Visitors who meet them are offered a meeting. Everyone else gets the trial link with no friction. The trial link is never hidden behind questions, and the rules are reviewed weekly against which chats actually turned into revenue.
Our buyers read docs and comparison pages for weeks before talking to us. Won't a chat bot annoy them?
It is built for that buyer. It answers the question asked first, from your own docs, and only then asks up to four qualification questions. It asks for an email only when the visitor wants something, such as a booking or a follow-up. Proactive pop-ups are limited to high-intent pages like pricing and comparisons, not every blog post.
We are a tiny team. What does it take to keep this running after it is handed off?
Plan on one owner spending roughly half an hour a week. They read escalated and unanswered chats, add or correct knowledge articles, and adjust routing rules. The runbook from Phase 3 covers how to update answers, change who gets which meetings and add new escalation topics. Most upkeep happens in the HubSpot knowledge base, not in the agent configuration.
Every good build
starts with a conversation.
New builds, remodels, and additions, powered by AI. Let’s plan the work.