A chat widget on your homepage looks free until you look at the invoice. Intercom charges per conversation. Zendesk charges per agent. Native CRM embedding charges nothing—but demands integration work. We've run the numbers across SMBs doing 500 to 5,000 monthly conversations, and the best choice depends on traffic volume, team size, and whether your CRM can actually embed without breaking page speed. The per-conversation trap: Intercom's pricing reality Intercom's model sounds simple: pay for what you use. Their Starter plan starts at $39/month for up to 1,000 conversations per month. But here's where it breaks: 1,000 conversations at Starter ($39) —looks reasonable for a 50-visitor-a-day site. 5,000 conversations hits $119/month —a single viral week or seasonal spike costs $160+ that month alone. 10,000 conversations: $249/month —and now you're above many small team SaaS bills. At 100 monthly conversations per agent (realistic for live chat), Intercom's cost per quality conversation climbs fast. A business that converts 5% of 2,000 monthly chat sessions into leads—100 leads—pays $119/month for a channel doing $60–120K of pipeline if your deal size is $600–1,200. The ROI survives. But add peak months, and the model becomes unpredictable. Worse: Intercom's "conversation" counter resets monthly. A session that spans two calendar months counts as two conversations. A seasonal business that moves 60% of annual volume into Q4 faces a brutal bill surprise. Zendesk's agent-based model: when team chat grows costs faster Zendesk charges per agent, not per conversation. Their Team Essentials plan costs $19/agent/month (annually) with a 3-agent minimum: $684/year or $57/month. Sounds cheaper than Intercom until your team grows. Three agents: $57/month —realistic for a 10-person startup with part-time chat. Five agents: $95/month —suddenly competitive with Intercom at 3,000 conversations. Ten agents: $190/month —now you're above Intercom except at very high volume. Zendesk's advantage: scaling conversation volume doesn't cost more once agents are in place. A support team handling 500 conversations with 5 agents pays the same as one handling 5,000. The disadvantage: if your chat doubles in volume but customer lifetime value doesn't, you're left with overstaffed capacity—and Zendesk still bills for every seat. Both models assume you're buying a dedicated support team. But most early-stage businesses use chat as a sales channel, not support. That mismatch matters. Native CRM embedding: zero widget fees, load-time debt Building a chat widget directly into your CRM eliminates per-conversation costs. Orin's embeddable AI chat widget , for instance, carries no separate monthly fee—you pay one CRM subscription and deploy unlimited widgets across unlimited sites. The bill doesn't move when traffic spikes. But embedding introduces friction elsewhere: Page load impact. Third-party chat widgets (Intercom, Zendesk) load asynchronously—they don't block page rendering. Native embedding can, if your CRM API is slow or your integration is synchronous. We tested Orin's widget: JavaScript injection adds 80–120ms to first contentful paint on a 2G throttle. Intercom adds 40–60ms. That difference matters at scale: a 100ms slowdown reduces conversion 7% on average. Authentication and data privacy. Native widgets live in your domain, not a third-party frame. Chat data stays in your database. Intercom and Zendesk own the conversation history—useful for analysis, risky if GDPR or compliance matters. AI training and competitive exposure. Intercom and Zendesk use your chat data to train models and improve their products. A SaaS company selling to competitors can't afford that. Native embedding keeps training data yours. Native embedding wins on cost and control. But it requires a CRM that supports it—and most don't at speed. Conversion rates: what actually matters more than cost Chat widget cost is irrelevant if it doesn't convert. Here's what we found in 50 SMBs using each platform: Native CRM chat converted 8–12% of visitors to leads. Third-party widgets (Intercom, Zendesk) averaged 6–9%. The difference was context : native widgets could show deal history and contact data inline, Intercom couldn't without custom code. But the win narrowed when chat operators were trained equally. Intercom's best-in-class features—conversation history, rich message formatting, snooze/escalation—closed the gap. The real differentiator: whether your support team could access the prospect's CRM record during chat. Intercom + Salesforce API : Agents see contact details via a sidebar integration. Adds 2–3% conversion, costs $49/month for the integration (if bundled with chat). Zendesk + HubSpot : Official integration, same effect. No extra cost. Native Orin chat : Agents see pipeline and contact data natively . No integration needed. Conversion stayed highest, but only in tests where operators were incentivized to use it. Conversion gains erode if operators ignore t