Ada

Ada is an enterprise AI customer service platform built around the ACX Platform, an AI-native system for support automation across chat, voice, email, SMS, social, WhatsApp, Messenger, Instagram, in-app, and custom channels. The platform is organized around the Reasoning Engine, Conversation Hub, Performance Center, and Developer Toolkit, with Playbooks as the structured automation primitive for repeatable support procedures.

The current Ada platform page says its AI agents resolve 84% of inquiries autonomously, support deployments in 85+ countries, and have handled 6.4B+ interactions across 550+ AI agents. Treat those as platform-level marketing and FAQ signals, not a guaranteed result for every deployment. The practical buying question is whether Ada can encode your support SOPs, connect to your CX stack, measure quality, and hand off safely when the AI should stop.

Ada is not a self-serve chatbot widget. It is a sales-led enterprise CX platform for teams with serious conversation volume, regulated workflows, and enough CX operations capacity to own Playbooks, coaching, test cases, integrations, and escalation policy.

What Changed Since The Last Refresh

The prior Ada page was last verified on 2026-06-18 and mainly described MCP, coaching, Web Chat SDK, and handoff changes. The June 25 refresh adds the June 22 Playbooks upgrade and keeps five June operating-surface changes in the buyer checklist.

• Playbooks moved to a structured step model. Ada’s June 22 release says Playbooks now use six step types: SEND, SET, ASK, RUN, IF/ELSE, and GO TO. Buyers get more predictable execution, AI-driven modes where useful, automatic intent-change handling, Knowledge-search pause/resume, escalation behavior, action error branching, and support across messaging, email, and voice.
• MCP is now part of the operating surface. Ada’s MCP overview says connected assistants such as Claude, ChatGPT, or any MCP client can analyze performance, update and create resources, run tests, answer questions, and stage changes safely. This makes Ada more admin-assistive, but it also creates a governance question around who can propose or confirm production changes.
• MCP now enforces dashboard roles. Ada’s June 18 release says Owners and Admins can call every OAuth-authenticated MCP tool, while Agents and Read Only users are limited to read-only tools. That is an important governance control for teams exposing Ada operations to assistants.
• June 16 release notes expand AI-assisted coaching work. Ada says propose_change can now create coaching across six types: reply, action, process, search_knowledge, handoff, and playbook. That matters because optimization is no longer only dashboard work.
• Multilingual Knowledge ingestion became more self-serve. The same June 16 notes added a dashboard toggle for multilingual Knowledge ingestion. Buyers should check language coverage, review workflow, and ownership before enabling it broadly.
• The Web Chat SDK gained deeper programmatic control. June 12 notes added actions such as sendMessage, getConversation, getMessages, getMetaFields, setComposerText, setDelegate, and isOpen, plus headless mode and lifecycle events. That makes custom front ends more realistic, but raises QA and accessibility requirements.
• Zendesk handoff behavior got a sharper setting. Ada’s June 12 notes added a setting to keep the same Sunshine Conversation after Zendesk Messaging handoff, which matters for continuity, auditability, and agent context.

The page also updates Ada’s scale signal from 5.5B+ handled interactions to the current platform FAQ’s 6.4B+ interactions, and replaces the older pricing read with the current platform FAQ: conversation-based pricing is the main public model, while resolution-based pricing is still available for specific enterprise needs.

What It Actually Is

Ada is an enterprise AI agent platform for customer support. It deploys and improves AI agents that can answer questions, execute structured support procedures, escalate to humans, and report on performance. The main design idea is that one Reasoning Engine and one operating surface should govern every service channel instead of splitting chat, voice, email, and social support across separate tools.

Playbooks are the key product primitive. A Playbook is a structured support SOP, such as refunding a customer, changing a shipping address, routing a billing dispute, or checking an order. It gives the AI agent enough structure to act without turning every path into a brittle decision tree.

The June 2026 product read is more operational than the older page suggested. Ada is not only “AI chatbot plus voice.” It now has a stronger optimization layer: structured Playbooks, MCP for admin-assistive changes, role-based MCP access, coaching creation, test-case expansion, Performance Center monitoring, SDK-level Web Chat control, and handoff settings that affect how conversations survive escalation.

When To Pick Ada

• Enterprise CX with real volume. Ada’s own demo path points to companies with at least 300K annual customer-service conversations. Below that, procurement and implementation overhead usually outweigh the benefit.
• Omnichannel consolidation. Teams running separate vendors for chat, voice, email, SMS, and social can standardize on one reasoning and governance layer.
• Regulated support. policy, safety controls, and audit expectations make Ada a stronger fit than lightweight chatbot builders for healthcare, financial services, telecom, travel, and enterprise SaaS.
• Support teams that can govern change. Structured Playbooks, MCP, coaching, multilingual ingestion, SDK customization, and handoff settings are powerful only when CX ops, compliance, and engineering know who owns each change.
• Zendesk, Salesforce, or Twilio-heavy teams. Direct integrations and handoff behavior make Ada more attractive when the support stack is already standardized.

When To Pick Something Else

• SMB or lower-volume support: Intercom, Zendesk AI, HubSpot, or Help Scout-style support automation will usually deploy faster and cost less.
• Builder-first conversational design: Voiceflow gives product and CX teams a more visual builder experience with clearer public packaging.
• Developer-first agent framework: Rasa, LangChain, LangGraph, or a custom build on Claude or ChatGPT gives engineering teams deeper control.
• Voice-only deployment without full CX context: ElevenLabs Conversational AI, Retell AI, Vapi, or CloudTalk may fit better when the job is phone automation rather than a complete service platform.
• Procurement that needs a price page before a call: Ada remains contact-sales. If buying cannot begin without published tiers, this will slow the process.

Pricing

Ada publishes no public pricing tiers. The current platform FAQ says conversation-based pricing is the primary model for most customers, while resolution-based pricing is still offered to enterprises with specific needs. The older pricing explainer frames conversation-based pricing as easier to forecast and compare because resolution definitions can vary across vendors.

Do not quote old third-party ACV ranges as current Ada pricing. They can be used internally as rough procurement preparation only after marking them as unverified. For buyer-facing content, the honest answer is simpler: Ada is enterprise contact-sales, and the contract should be modeled around volume, channels, integration depth, professional services, support operations, and the pricing unit Ada proposes.

Against The Alternatives