The best AI chatbot with human handoff is the one you rarely need to trigger. A content-trained bot should resolve most questions on its own, and when it does escalate, the entire conversation must transfer to the agent so the customer never repeats themselves. Short answer: for the most sophisticated native AI-to-human escalation, Intercom and Zendesk lead. For the highest content-trained deflection plus native handoff at a price a small team can afford, SiteGPT is the best value.
iShort answer
Almost every chatbot has a "talk to a human" button now, so that is not a useful way to compare them. What separates them is handoff quality: does the full conversation context transfer to the agent, does escalation trigger at the right moment, how many channels does it cover, and, most importantly, how much does the bot resolve before it ever escalates. This guide compares 7 tools on exactly that.
Key takeaways
| Tool | Best for | Handoff strength |
|---|---|---|
| SiteGPT | Content-trained deflection at value (12+ sources, auto-sync) | Native, full-transcript, on-demand |
| Intercom | Most sophisticated native escalation | Best-in-class (AI-decided, shared inbox) |
| Zendesk | Enterprise omnichannel workspace | Best-in-class (ticket + full context) |
| Tidio | SMB / e-commerce, easy setup | Solid (on-demand + AI-decided) |
| Crisp | Affordable flat-rate omnichannel | Solid (context attached) |
| Chatbase | Custom agent into your helpdesk | Lighter (escalates to 3rd-party desk) |
| Ada | Enterprise resolution across channels | Strong (deep desk integrations) |
The four things that separate a good handoff from a frustrating one. This guide scores every tool on them.
Context transfer
Does the full conversation (ideally an AI summary) reach the agent, so the customer never repeats themselves?
Trigger intelligence
Does it escalate at the right moment, on demand and automatically, not just when a customer hunts for a button?
Channel coverage
Does handoff work across web, email, WhatsApp, social, and voice, with history that follows the conversation?
Deflection before escalation
How much does a content-trained bot resolve first? The best handoff is the one you rarely need.
80%
of common customer service issues will be resolved autonomously by AI agents by 2029, cutting support operating costs by 30% (Gartner). The more your bot deflects, the less your handoff has to carry.
Source: Gartner, March 2025What to look for in a chatbot with human handoff
A handoff button is table stakes. These six criteria are what actually separate a good escalation experience from a frustrating one. Use them to evaluate any tool, not just the ones below.
- Context transfer. When the bot hands off, does the agent receive the full conversation history, ideally with an AI summary of what the customer wanted and what the bot already tried? If the agent has to ask the customer to start over, the handoff failed.
- Trigger intelligence. Escalation should happen at the right moment, not just when a customer hunts for a button. The strongest tools escalate on demand and automatically: when the AI detects frustration, a repeated failed answer, or a topic outside its knowledge.
- Channel coverage. Your customers arrive on the website widget, email, WhatsApp, Instagram, and sometimes voice. Handoff should work across every channel you support, with history that persists if the conversation moves between them.
- Agent notifications and routing. When a conversation escalates, the right agent or team needs to be alerted and assigned (via a shared inbox, Slack, email, or skills-based routing) so nothing sits unanswered.
- Deflection rate before escalation. This is the one most buyers overlook. A bot trained on your own content (a RAG bot) resolves the long tail of questions itself, so you escalate far less. Gartner puts successful AI deflection in the 40-80% range, and the high end is where a content-trained bot pays for itself. The breadth and freshness of what the bot can learn from directly sets this ceiling.
- Pricing model. Handoff and AI often sit behind specific tiers, and the model matters at scale: flat per-workspace, per-agent seat, or per-resolution. Check which plan unlocks the bot and live agent handoff, and model your real volume.
How we tested
Each tool was evaluated against the six criteria above, with a bias toward the two that customers feel most: how completely context transfers to the agent, and how much the bot deflects before it ever escalates.
Every factual claim was checked against primary sources. Pricing and plan limits come from each vendor's official pricing page; ratings come from the actual G2 and Capterra review pages (with review counts); handoff behavior comes from official documentation. Where a figure could not be confirmed first-party, it is labeled as such rather than guessed. SiteGPT's product facts are taken from its official pricing, security, and documentation pages. All figures were accurate as of July 2026; check each vendor's site for the latest.
The 7 best AI chatbots with human handoff
Every tool below follows the same template: what its handoff actually does, pros, cons, pricing, third-party ratings, and who it is best for.
1. SiteGPT: best value for content-trained deflection plus native handoff
SiteGPT trains a chatbot on your own content so it resolves the bulk of questions before anyone needs a human. Its edge is the breadth and freshness of what it can learn from: 12+ content sources (websites, sitemaps, files, YouTube, GitHub, Notion, Google Drive, Dropbox, OneDrive, SharePoint, Box, and help centers), with auto-sync and auto-scan that keep answers current as your docs change. Better, fresher answers mean higher deflection, so escalations are rarer. When a customer does want a human, SiteGPT's native escalate-to-a-human feature hands off the full conversation.
It is honest about its lane: the handoff is on-demand, not AI-decided, so it does not out-engineer Intercom's escalation logic. Where it wins is value: content-trained deflection, a fully customizable and brandable widget, native handoff, and integration with the tools teams already use, at a flat price a small team can afford.
Handoff features
- Context transfer: full conversation transcript plus captured contact details and a link to the conversation are sent on every escalation.
- Triggers: on-demand. The customer clicks an escalation button or simply asks for a human in chat.
- Channels and integrations: website widget as the escalation entry point, with integrations for Slack, Zendesk, Crisp, Messenger, Google Chat, Freshchat, and Zoho SalesIQ (Intercom, WhatsApp, and HubSpot are listed as coming soon), plus a webhook and API.
- Notifications: the team is notified when a customer asks for a person, via email, webhooks, Slack and ticket creation through Zapier, and a Human Support view in the dashboard.
- Deflection: content-trained RAG across 12+ sources, with scheduled auto-sync (monthly on Growth, weekly plus daily auto-scan on Scale) so answers do not go stale.
SiteGPT is also SOC 2 Type II certified, with GDPR compliance and a HIPAA BAA available on Enterprise (security), which matters for regulated teams, though it is not the main reason to pick it.
Pros
- Deepest content integration in this list (12+ sources incl. cloud storage, GitHub, YouTube, help centers), so it deflects the long tail before escalation
- Auto-sync and auto-scan keep answers current automatically, which keeps deflection high over time
- Fully customizable and brandable widget, plus 95+ languages out of the box
- Native handoff transfers the full transcript and contact details, with no repeating for the customer
- Flat pricing with AI included: no per-resolution fees or daily caps
Cons
- Handoff is on-demand (button or ask), not AI-decided on sentiment or repeated failures like Intercom/Zendesk
- Escalation is centered on the website widget, not a full omnichannel live-agent desk
- No native voice channel; WhatsApp is coming soon rather than live
- Third-party review volume is still early-stage
Pricing (billed monthly; annual saves 40%; 7-day free trial on all plans):
| Plan | Price | Messages/mo | Pages | Team |
|---|---|---|---|---|
| Starter | $39/mo | 4,000 | 1,000 | 1 |
| Growth | $79/mo | 10,000 | 10,000 | 4 |
| Scale | $259/mo | 40,000 | 50,000 | 10 |
| Enterprise | Custom | Custom | 500,000 | up to 10,000 |
Human escalation is available on every plan, including Starter.
4/5·Capterra (early reviews)"SiteGPT makes it easy & intuitive to get your chatbot setup & working in no time at all - anyone can do it." Verified user on G2
"SiteGPT has proven to be an invaluable tool for providing swift and accurate responses to visitors' inquiries, with seamless integration with website content that effectively mirrors the website's tone and style." Verified user on Product Hunt
SiteGPT was also Product Hunt #1 Product of the Day. Independent review volume is still small compared with the incumbents.
Best forTeams that want to deflect most tickets with a content-trained bot (trained on everything they already have) and hand off with full context, without paying for an enterprise suite.
2. Intercom: best-in-class native AI-to-human handoff
Intercom sets the bar for native escalation. Its Fin AI agent and human agents work from the same shared inbox and customer record, so when Fin escalates, the conversation lands in an agent's normal workspace with the full thread and an AI-generated summary. This is genuinely its home turf, and no on-demand button beats it. The trade-off is cost and complexity.
Handoff features
- Context transfer: full history plus an AI summary of what the customer asked, what Fin tried, and why it escalated.
- Triggers: the full set: AI-decided (frustration or repeated loops), on-demand, deterministic rules, and natural-language escalation guidance.
- Channels: Messenger, email, voice, WhatsApp, SMS, Instagram, Facebook, and Slack, with history that persists across channels.
- Notifications: escalations route and assign into the shared Inbox with standard team routing.
- Deflection: Fin is a RAG agent trained on your help center. Intercom cites customer results up to about 65% resolution (verify current figures on their site).
Pros
- The most seamless native AI-to-human handoff: shared inbox, full context, AI summary
- AI-decided escalation triggers, not just an on-demand button
- Broad channel coverage including voice and WhatsApp
- Enterprise-grade compliance (SOC 2 Type II, GDPR, HIPAA via BAA)
Cons
- Costs stack: per-seat pricing plus $0.99 per Fin resolution gets expensive at volume
- More platform to configure and learn than a focused bot
- Premium positioning, so the entry price is only the beginning
Pricing (billed annually):
| Plan | Price | Notes |
|---|---|---|
| Essential | ~$29/seat/mo | Entry tier |
| Advanced | ~$85/seat/mo | More automation |
| Expert | ~$132/seat/mo | Largest teams |
| Fin AI agent | $0.99 / resolution | On top of seats |
Source: Intercom pricing.
4.5/5·G2(3,800) 4.5/5·Capterra(1,133)Best forTeams that want the most sophisticated native AI-to-human escalation and can absorb combined seat and per-resolution pricing.
3. Zendesk: best for enterprise omnichannel workspace
Zendesk is the mature incumbent. Its AI agent opens a ticket that, on escalation, converts into a regular Support ticket in the Agent Workspace, so continuity is preserved and the human agent inherits the transcript, intent, sentiment, and the full customer profile. Like Intercom, its handoff is genuinely strong; the cost is per-seat pricing and setup complexity.
Handoff features
- Context transfer: full transcript auto-attached, plus intent classification, sentiment, entities, ticket tags, and prior cross-channel history.
- Triggers: a default "transfer to agent" step, a "talk to a human" option, escalation blocks in the dialogue builder, or a custom API transfer with metadata.
- Channels: web widget, mobile SDKs, and social messaging, with voice AI agents (EAP) that hand off with transcript, AI summary, and intent.
- Notifications: routing driven by ticket fields (group, skills) into standard omnichannel/skills-based routing.
- Deflection: AI agents grounded in your help center, with RAG-style retrieval (reinforced by Zendesk's 2025 acquisition of Unleash).
Pros
- Mature, full-context AI-to-human handoff inside a proven agent workspace
- Deep routing, skills, and omnichannel coverage including voice
- Enterprise compliance (SOC 2 Type II, GDPR, HIPAA via add-on + BAA)
Cons
- Per-agent pricing adds up quickly for larger teams
- The autonomous AI Agents Advanced tier is priced per resolution and not published
- More setup and administration than a focused chatbot
Pricing (per agent/mo, billed annually):
| Plan | Price | Notes |
|---|---|---|
| Suite Team | $55/agent/mo | Entry tier |
| Suite Professional | $115/agent/mo | Advanced routing |
| Suite Enterprise | Custom | On request |
| Copilot add-on | $50/agent/mo | Optional |
AI Agents Advanced is priced per automated resolution. Source: Zendesk pricing.
4.3/5·G2(7,100) 4.4/5·Capterra(4,083)Best forMid-market and enterprise teams that need a mature omnichannel workspace with deep routing and full-context handoff, and can absorb per-seat pricing.
4. Tidio: best for SMB and e-commerce
Tidio pairs an easy-to-deploy live chat with its Lyro AI agent, which is a strong fit for Shopify and WooCommerce stores. Lyro deflects well and escalates sensibly (on request, when it cannot answer, or by rule), making it one of the better SMB options for both deflection and handoff.
Handoff features
- Context transfer: conversation context is preserved (roughly the last few hours) so the human agent picks up with continuity.
- Triggers: on-demand, AI-decided (unanswered questions or detected frustration), and rule-based via audiences.
- Channels: website widget, Facebook Messenger, Instagram, WhatsApp, and email.
- Notifications: handed-off chats land in an Unassigned folder for any available agent (auto-assignment on higher tiers).
- Deflection: Lyro is a RAG agent that works only from content you provide, with an advertised up-to-67% resolution rate and a 50% resolution guarantee on higher tiers.
Pros
- Fast to deploy; strong fit for e-commerce (Shopify, WooCommerce)
- Lyro deflects well (up to 67% claimed) and escalates on demand or automatically
- Social and WhatsApp coverage plus a free plan to start
- Strong third-party ratings
Cons
- Conversation-based pricing can climb as traffic grows
- Lyro AI is a metered add-on that reviewers find confusing to price
- No HIPAA support stated
- Less deep routing than Zendesk or Intercom
Pricing (paid tiers billed annually):
| Plan | Price | Notes |
|---|---|---|
| Free | $0 | 50 conversations |
| Starter | ~$24/mo | Live chat + basics |
| Growth | ~$49/mo | Advanced analytics |
| Plus / Premium | $749/mo+ | Custom |
| Lyro AI add-on | from $32.50/mo | 50 AI conversations |
Source: Tidio pricing.
4.6/5·G2(1,880) 4.7/5·Capterra(590)Best forSmall and mid-size e-commerce teams that want easy setup, social and WhatsApp coverage, and solid AI deflection with handoff.
5. Crisp: best for affordable flat-rate omnichannel
Crisp is a multi-channel shared inbox first, with a content-trained AI agent (Hugo) layered on top. Its flat per-workspace pricing is attractive for startups, and Hugo hands off to a human with context attached. The main caveat is compliance: Crisp's own docs state it has not completed a SOC 2 audit.
Handoff features
- Context transfer: Hugo hands the conversation to a human with the context attached; AI-handled threads live in an automated inbox agents can pick up.
- Triggers: routing rules, escalation settings where Hugo decides it is out of depth, and manual agent takeover, all designed in a workflow builder.
- Channels: website chat, email, WhatsApp, Facebook Messenger, and Instagram in one shared inbox.
- Notifications: escalations surface in the shared inbox.
- Deflection: Hugo is a RAG agent (built on a Claude model) trained on your website, knowledge base, Q&A, and uploaded files.
Pros
- Affordable, predictable flat per-workspace pricing (extra seats just $10/mo)
- Genuine omnichannel shared inbox across web, email, and social
- Content-trained AI (Hugo) hands off with context attached
Cons
- Crisp's own docs state it has NOT completed a SOC 2 audit (self-attested only)
- AI is a paid credit add-on at every tier, and the free plan has no AI
- HIPAA is not substantiated on Crisp's own security docs
- Fewer enterprise routing and admin controls
Pricing (flat per workspace; +$10/mo per extra seat):
| Plan | Price | Notes |
|---|---|---|
| Free | $0 | 2 seats, no AI |
| Mini | $45/mo | AI unlocks here |
| Essentials | $95/mo | More AI credits |
| Plus | $295/mo | Highest credits |
Source: Crisp pricing.
4.5/5·G2(179) 4.6/5·Capterra(148)Best forStartups and SMBs that want an affordable, flat-rate omnichannel inbox with a content-trained AI, where predictable pricing matters more than audited compliance.
6. Chatbase: best for a custom agent into your helpdesk
Chatbase is a custom AI-agent builder first, and a helpdesk second. Its deflection is strong (a RAG agent trained on your content with real-time data sync), and it escalates via an "Escalate to Human" action into a connected desk (Zendesk, Salesforce, Intercom, Freshdesk, HubSpot, Help Scout, Zoho) or its own newer native Help Desk. Be realistic that its live-agent layer is younger and lighter than the dedicated platforms.
Handoff features
- Context transfer: the escalate action creates a ticket/conversation in the connected platform and passes the relevant history and user info.
- Triggers: trigger-based via an AI Action that fires when the user asks for a human or the AI cannot resolve after attempts.
- Channels: website chat, WhatsApp, Slack, email, and Messenger; voice/telephony on Standard and above.
- Notifications: native Help Desk (Standard+) offers conversation views, assignment, and takeover; specific notification mechanics are thinly documented.
- Deflection: RAG agent trained on website crawl, uploaded files, and Q&A, with real-time sync to connected systems.
Pros
- Strong content-trained deflection with real-time data sync
- Escalates into 8 major helpdesks, or its own native Help Desk (Standard+)
- SOC 2 Type II and GDPR compliant; low entry price with a free tier
Cons
- Native live-agent handoff is newer and lighter than dedicated helpdesk platforms
- The Help Desk and voice features require the Standard plan ($120/mo) or above
- Message-credit limits by tier; HIPAA only on Enterprise
Pricing (billed annually):
| Plan | Price | Notes |
|---|---|---|
| Free | $0 | 50 credits |
| Hobby | $32/mo | 500 credits |
| Standard | $120/mo | 4,000 credits; adds Help Desk, voice, API |
| Pro | $400/mo | 15,000 credits |
Source: Chatbase pricing.
4.3/5·Capterra(73)G2 shows roughly 4.7/5 across a small review base (~14; verify on the G2 page).
Best forTeams that want to build a custom, content-trained AI agent that deflects well and escalates into the helpdesk they already run.
7. Ada: best for enterprise resolution across channels
Ada is an enterprise AI-agent platform built around measurable automated resolution across chat, email, and voice. Its handoff is strong through deep native integrations: it transfers in real time (or creates a case) into Zendesk, Salesforce, or Intercom with transcript context. The catch is that pricing is custom-quote only, which puts it out of reach for most SMBs.
Handoff features
- Context transfer: transfers the transcript and captured customer data; for Intercom, the full transcript is visible to the agent.
- Triggers: real-time live escalation, asynchronous ticket follow-up, off-hours handling, and error-recovery fallback.
- Channels: web/messaging, email, and voice (SMS and social are not explicitly named on Ada's own site).
- Notifications: context-based routing to different handoffs based on conversation variables, with availability rules.
- Deflection: RAG agent grounded in your content and connected systems; Ada frames the metric as "automated resolution" and cites per-customer results rather than one headline number.
Pros
- Enterprise omnichannel automated resolution across chat, email, and voice
- Deep native handoff integrations (Zendesk, Salesforce, Intercom) with transcript context
- Strong compliance posture (SOC 2 Type II, GDPR, HIPAA)
Cons
- Custom, quote-based pricing only, with no public numbers and geared to enterprise budgets
- Overkill for SMB and mid-market teams
- Longer deployment and sales cycle than self-serve tools
Pricing:
| Plan | Price | Notes |
|---|---|---|
| All plans | Custom quote | No public pricing; book a consultation |
Source: Ada pricing.
4.6/5·G2(170) 4.7/5·Capterra(15)Best forEnterprises that want an omnichannel AI agent focused on automated resolution across chat, email, and voice, with deep helpdesk handoff, and can engage a custom sales process.
Side-by-side comparison
| Tool | Context transfer | Trigger intelligence | Channels | Deflection | Entry price (AI + handoff) | SOC 2 / HIPAA |
|---|---|---|---|---|---|---|
| SiteGPT | Full transcript + contact | On-demand (button or ask) | Web widget (+ Slack, Zendesk, Crisp, Messenger, more) | Content-trained RAG, 12+ sources + auto-sync | $39/mo flat | SOC 2 II / HIPAA (BAA, Ent) |
| Intercom | Full history + AI summary | AI-decided, on-demand, rules, guidance | Web, email, voice, WhatsApp, SMS, social | Fin RAG (~65% cited) | ~$29/seat + $0.99/resolution | SOC 2 II / HIPAA (BAA, Ent) |
| Zendesk | Full transcript + intent + sentiment | Transfer step, talk-to-human, escalation blocks, API | Web, mobile, social, voice (EAP) | AI agents RAG | $55/seat (+ AI add-on) | SOC 2 II / HIPAA (add-on + BAA) |
| Tidio | Context preserved (~3h) | On-demand, AI-decided, rules | Web, Messenger, Instagram, WhatsApp, email | Lyro RAG (up to 67%) | ~$33/mo (Lyro add-on) | SOC 2 / No HIPAA |
| Crisp | Context attached | Rules, escalation settings, manual | Web, email, WhatsApp, Messenger, Instagram | Hugo RAG | $45/mo flat | Self-attested (no audit) / No |
| Chatbase | History + user to ticket | AI Action (trigger-based) | Web, WhatsApp, Slack, email, Messenger | Content RAG + sync | $120/mo (Standard) | SOC 2 II / HIPAA-eligible (Ent) |
| Ada | Transcript + data | Live, async, off-hours, fallback | Web, email, voice | Content + systems RAG | Custom quote | SOC 2 II / HIPAA |
Prices are not apples-to-apples: SiteGPT, Crisp, and Chatbase are flat per-workspace; Zendesk is per agent/seat; Intercom (from ~$29/seat) adds $0.99 per Fin resolution on top of seats; Tidio's Lyro is a metered add-on; Ada is custom-quote. Figures are as of July 2026, so always confirm on the vendor's pricing page.
$141/mo less than a 4-seat Zendesk Suite Team
SiteGPT Growth is $79/mo flat and includes up to 4 team members, 10,000 messages, and native handoff. Four Zendesk Suite Team seats are 4 × $55 = $220/mo, before Zendesk's per-resolution AI Agents Advanced add-on. Roughly $1,692/year saved on seats alone. As of July 2026.
How to choose
Match the tool to what you actually optimize for.
What matters most for your handoff?
- If you want the most sophisticated AI-decided escalation and have budget→Intercom or Zendesk
- If you want to deflect the most tickets before escalation from content you already have→SiteGPT (12+ sources + auto-sync)
- If you're an e-commerce/SMB team that lives on social and WhatsApp→Tidio
- If you want the lowest predictable flat price and don't require audited SOC 2→Crisp
- If you're building a custom agent to plug into an existing helpdesk→Chatbase
- If you're an enterprise automating chat, email, and voice→Ada
How important is compliance (SOC 2 / HIPAA)?
- If you need HIPAA at a small-team price→SiteGPT (BAA on Enterprise)
- If you need HIPAA at enterprise scale→Intercom, Zendesk, or Ada
- If SOC 2 is a hard requirement→Avoid Crisp (self-attested, no audit)
Use-case cheat sheet
| Scenario | Best pick | Why |
|---|---|---|
| SaaS company with docs, help center, and a GitHub wiki to learn from | SiteGPT | 12+ content sources + auto-sync drive the highest deflection |
| Healthcare team needing HIPAA without an enterprise suite | SiteGPT | HIPAA BAA on Enterprise at a fraction of suite pricing |
| Large support org wanting AI-decided escalation | Intercom | Fin escalates on frustration/failure into a shared inbox |
| Enterprise with complex routing and voice | Zendesk | Mature agent workspace, skills routing, voice EAP |
| Shopify store handling DMs and WhatsApp | Tidio | Lyro + social/WhatsApp coverage, easy setup |
| Startup wanting the lowest predictable flat price | Crisp | Flat per-workspace pricing, omnichannel inbox |
| Team that already runs Zendesk/Salesforce and wants a custom bot | Chatbase | Custom RAG agent that escalates into your desk |
| Enterprise automating chat, email, and voice at scale | Ada | Omnichannel automated resolution + deep handoff integrations |
Frequently asked questions
Basics
What is human handoff in an AI chatbot? Human handoff (or live agent escalation) is when an AI chatbot transfers a conversation to a human agent. Good handoff transfers the full conversation history so the customer never repeats themselves, and it triggers at the right moment, on demand or when the AI detects it cannot resolve the issue.
Which AI chatbot has the best human handoff? For the most sophisticated native AI-to-human escalation (AI-decided triggers, full context plus an AI summary, and a mature shared agent inbox), Intercom and Zendesk lead. For the highest content-trained deflection plus native handoff at SMB pricing, SiteGPT is the best value.
Can the AI chatbot escalate to a human automatically? It depends on the tool. Intercom, Zendesk, and Tidio can escalate automatically when the AI detects frustration or a failed answer. SiteGPT and Crisp escalate on demand and by rule. Automatic escalation catches customers who would otherwise abandon the chat.
Handoff and context
Do all AI chatbots transfer conversation context to the human agent? No. Every tool here transfers at least the transcript, but depth varies. Intercom and Zendesk pass the full history plus an AI summary, intent, and sentiment. SiteGPT sends the full transcript and captured contact details. Builder-first tools may only pass a ticket with limited context unless configured.
Does SiteGPT integrate with Zendesk, Slack, or Crisp for escalation? Yes. SiteGPT integrates with Slack, Zendesk, and Crisp, plus Messenger, Google Chat, Freshchat, and Zoho SalesIQ, so escalations and notifications reach the tools a team already uses. Intercom, WhatsApp, and HubSpot are listed as coming soon.
Deflection
What is a good ticket deflection rate for an AI chatbot? Gartner cites successful AI support deflection in the 40-80% range depending on the use case. A content-trained (RAG) bot that answers from your own docs and help center drives the high end, and the more it deflects, the less often a handoff is needed at all.
How many content sources can a SiteGPT chatbot train on? SiteGPT trains on 12+ content sources, including websites, sitemaps, files (PDF, DOCX, CSV, and more), YouTube, GitHub, and cloud storage such as Google Drive, Notion, Dropbox, OneDrive, SharePoint, and Box, plus help centers. Auto-sync keeps the trained content current, which keeps deflection high over time.
Pricing and compliance
How much do AI chatbots with human handoff cost? Models differ. SiteGPT, Crisp, and Chatbase charge a flat monthly rate (SiteGPT from $39/mo, Crisp from $45/mo, Chatbase's handoff-capable Standard plan at $120/mo). Zendesk ($55/agent/mo) and Intercom (from ~$29/seat/mo) charge per seat, and Intercom adds $0.99 per Fin resolution. Ada is custom-quote only.
Which SiteGPT plan do I need for human handoff? Human escalation is available on every SiteGPT plan, including Starter at $39/mo. Higher plans add more team seats to handle escalations (Starter 1, Growth 4, Scale 10) and add integrations, API access, and auto-sync.
Which of these are SOC 2 and HIPAA compliant? SiteGPT, Intercom, Zendesk, Chatbase, and Ada hold SOC 2 Type II; Tidio has completed a SOC 2 examination; Crisp states it has not been through a SOC 2 audit. For HIPAA, SiteGPT (BAA on Enterprise), Intercom (BAA), Zendesk (add-on + BAA), and Ada support it; Chatbase is HIPAA-eligible on Enterprise; Tidio and Crisp do not state HIPAA support.
For a deeper look at compliance-specific options, see the guides to the best HIPAA-compliant AI chatbots and GDPR-compliant chatbot platforms, or browse the best conversational AI chatbots overall.
Sources
- SiteGPT: pricing, features, security
- Vendor pricing pages: Intercom, Zendesk, Tidio, Crisp, Chatbase, Ada
- Ratings: G2 and Capterra review pages, linked per tool above (review counts as shown)
- Market data: Gartner, "Agentic AI Will Autonomously Resolve 80% of Common Customer Service Issues by 2029" (March 2025)
Last updated: July 2026. All pricing and features verified as of July 2026.