The pitch sounds almost too straightforward for the convoluted world of enterprise software pricing: What if you only paid for customer service automation that actually worked?
That's the wager Cloud Humans is placing with ClaudIA, its AI support agent. The Y Combinator-backed company now charges exclusively for tickets its AI resolves completely—no human handoff, no charge. Escalate to a real person? That one's free.
It's a model that Intercom and Zendesk have already embraced over the past year, shifting to per-resolution billing as AI agents moved from novelty to operational necessity. But Cloud Humans, which emerged from YC's Winter 2021 batch after a 2020 founding, believes it can undercut the giants with a founder-friendlier approach. Volume-based tiers. No billing when the bot punts. An average 65% reduction in support costs, by the company's math.
The pricing structure went live recently—at least in Brazil, where Cloud Humans displays tiered per-resolution rates in reais on its market calculator. U.S. prospects get directed to WhatsApp, where they can chat with ClaudIA itself for custom quotes. The principle holds either way: Complete AI resolution? You pay. Anything else? You don't.
Whether that proves more attractive than what Intercom or Zendesk offer will hinge on execution. And on how Cloud Humans defines "retained resolution"—a term that can get slippery in practice.
The Mechanics
Here's how it breaks down. Companies pay per fully automated resolution using a volume tier system. More tickets ClaudIA handles, lower the per-resolution rate. Standard SaaS math.
There's a second component if you're using Cloud Humans' own omnichannel platform—Cloud Chat—rather than bolting ClaudIA onto an existing Zendesk or Intercom setup. That carries a separate user license fee. In Brazil, it's R$315 per user monthly.
What the company emphasizes, particularly in its FAQ, is the inverse cost structure. "Atendimentos com envolvimento humano não são cobrados," the Portuguese reads. Tickets with human involvement aren't charged.
That's a clearer line than some competitors draw, where what qualifies as a billable resolution can depend on interpretation. Or optimism.
Cloud Humans offers a 30-day paid pilot. Implementation timelines run 15 to 21 days to reach what the company considers acceptable accuracy. Discounts apply for semi-annual or annual commitments—the usual incentives to lock in longer terms.
Crowded Territory
Cloud Humans isn't alone in betting on outcomes over seats. The customer support AI market has undergone a quiet realignment over the past 18 months, as vendors realized the old helpdesk economics—charge per agent, per ticket, or by volume tier—no longer fit a world where bots handle first contact.
Intercom's Fin now runs $0.99 per AI resolution, with a 50-resolution monthly floor for standalone deployments. Zendesk rolled out its own resolution-based pricing in 2024, advertising rates "as low as $1.50 per resolution," though actual costs vary depending on committed volumes. (Enterprise software pricing rarely means what it says on the tin.)
Then there's Sierra, the well-funded startup from Bret Taylor and Clay Bavor, which also prices on outcomes but doesn't publish rates. Forbes reported last November that Sierra's fees shift based on task complexity—a logical approach, perhaps, but one that introduces ambiguity for budget-conscious buyers. Other entrants include Crescendo, starting at $2.99 per resolution, and Can I Help, which charges $1 per text chat resolution.
The shift reflects something broader. Traditional helpdesk vendors made money whether their software helped or not. Resolution-based billing, in theory, aligns incentives: vendors profit only when their AI actually solves problems. In theory.
What ClaudIA Actually Does
ClaudIA runs on OpenAI's models—GPT-4.1, GPT-4o, and Whisper for voice, according to Cloud Humans' security documentation. The company says it doesn't train on customer data and requires deletion of personal information within 30 days, barring legal obligations otherwise.
The agent operates across WhatsApp, web chat, email, Instagram DMs, and Facebook Messenger. Cloud Humans is an official WhatsApp Business API partner, and its Cloud Chat platform leans heavily into WhatsApp-first workflows: unlimited inboxes, native campaign templates, in-conversation CSAT tracking. Brazil's WhatsApp obsession makes this a logical play.
ClaudIA doesn't just answer questions. It can execute actions in back-office systems through API integrations—what the company calls "MCP" connections, nodding to Model Context Protocol-style architecture. The site mentions "Eddie" automation flows: checking order status, processing payments, generating boletos (Brazil's ubiquitous bank slips), creating tickets, updating records.
Setting these up isn't plug-and-play. Companies provide business descriptions and full technical specs: endpoints, payloads, authentication schemes, error handling protocols. Implementation, in other words, requires engineering lift.
For companies already committed to Zendesk, Intercom, Freshdesk, or HubSpot, ClaudIA offers native integrations. Or they can go all-in on Cloud Chat as their primary support infrastructure.
The Results (According to Cloud Humans)
The company claims an average 65% reduction in tickets requiring human agents, with some implementations hitting 80% to 85% automation. CSAT scores improve by 35 percentage points compared to traditional chatbots, Cloud Humans says.
One case study the company highlighted on LinkedIn—an implementation it refers to as "Lia"—saw ticket retention jump from 4% to 42% within 15 days. Eventually: 70%. First response times landed under 30 seconds in every case. CSAT clocked in at 88%.
The company's SaaS vertical page lists other examples. CRM Bonus: 85% reduction in requests needing human escalation. Yampi: more than 70% ticket load reduction with 95% CSAT. Contraktor: 75% of support interactions automated. The page notes these figures were updated in September 2025.
Standard caveats apply. Companies showcase their best results, not their median ones. And definitions matter: What counts as "retained"? Does the customer need to confirm satisfaction? Does an LLM verify the resolution? Or does silence equal success?
Cloud Humans' public materials don't spell out the exact criteria. Zendesk's model, for comparison, uses LLM verification to determine whether a resolution occurred—a technical answer to what's often a judgment call.
Under the Hood
ClaudIA's architecture relies on an internal system Cloud Humans calls IDS: essentially question-title and answer pairs that the system searches semantically. Companies can configure context, identity, and tone. Recently, Cloud Humans gave clients direct access to edit base prompt modules—a nod to the reality that generic AI rarely works out of the box.
The help center documentation, last updated in August 2025, describes guardrails and "micro-agents" that ask clarifying questions rather than guessing. There's tooling for simulating conversations, auditing responses, calibrating the agent over time.
The company emphasizes ClaudIA's ability to interpret not just text but documents, photos, and audio. It includes frustration detection—a feature that sounds minor until you've watched a chatbot repeatedly miss signals that a customer is about to rage-quit. When ClaudIA determines a customer is satisfied, it closes the conversation.
Whether these features deliver as promised in messy real-world deployments is always the question. Demos impress. Production is where things get complicated.
Why This Matters Now
The launch arrives as AI agents transition from experimental curiosity to operational reality in customer support. A Google Cloud survey in Brazil found that half of companies cited customer experience and support as early use cases for agentic AI. Capgemini's 2026 World Cloud Report on financial services projected AI agents could unlock $450 billion in value by 2028, with customer support high on the deployment list.
For SaaS companies, the logic is compelling. Traditional support economics don't scale gracefully: as ticket volume climbs, you hire more agents, margin compresses, and suddenly growth doesn't feel like growth. Outcome-based pricing for AI that genuinely resolves issues—rather than just deflecting them or creating more work—offers different math.
But the devil, as always, lives in the definitions. What is a resolution? When does the clock start and stop? How much credit does the AI get if a human has to finish the job?
Cloud Humans pitches primarily to the SaaS vertical for now, with dedicated pages and case studies targeting that audience. The company hasn't disclosed funding beyond its YC pedigree. U.S. pricing remains quote-based, not published—a choice that either reflects customization flexibility or reluctance to compete on sticker price with better-known rivals.
The product has been in market since at least mid-2023, based on its Product Hunt launch. But the explicit pay-per-resolution model with the "no charge for human escalations" clause appears recent. Documentation suggests the pricing calculator went live within the past few weeks.
What Remains Unclear
Cloud Humans faces the classic challenger dilemma. Intercom and Zendesk have distribution, brand recognition, and existing customer relationships. Can a YC startup undercut them enough on price to overcome those advantages? Especially when the incumbents can adjust their own pricing faster than a smaller player can build market share?
There's also the question of accuracy at scale. A 70% retention rate sounds impressive in a case study. Whether that holds across hundreds of deployments, in varied industries, with wildly different support complexity—that's harder to predict from marketing materials.
Still, the bet makes sense. As AI gets better, pricing tied to outcomes should favor companies whose technology actually works. And for founders tired of seat-based SaaS fees that never quite align with value delivered, paying only for what resolves is at least honest.
Whether Cloud Humans can execute on that promise, and whether its pricing proves materially better than Intercom's or Zendesk's in real deployments, will determine if this is differentiation or just another pitch deck talking point.
For now, the model exists. Companies can test it. And if ClaudIA escalates too often? Well, at least you won't get billed.
