Ada
Ada is an enterprise platform whose AI agents resolve customer service enquiries autonomously across voice, messaging and email. A single reasoning layer drives every channel, with SOC 2 Type II, GDPR, HIPAA and AIUC-1 compliance.
What is Ada?
Ada is the agentic customer experience platform built by Ada Support Inc., a Toronto company that describes itself as AI-native since 2016. It does one thing: it runs AI agents that handle customer service conversations end to end, on voice, on messaging (web chat, SMS, WhatsApp, Messenger, Instagram, in-app and custom channels) and on email. The platform rests on four blocks. The Reasoning Engine is a single intelligence layer that orchestrates several leading large language models behind multi-layer safeguards, so an agent behaves the same way whichever channel a customer picks. The Conversation Hub connects those channels. The Performance Center is where a team builds, launches, analyses and optimises an agent, tracking automated resolution, CSAT, handle time, escalation frequency and cost per resolution, broken down by channel, language and region. The Developer Toolkit exposes APIs, an MCP server and SDKs. Two mechanisms do most of the practical work. Playbooks turn a company's own standard operating procedures into structured, multi-step workflows that fetch live data instead of following a rigid decision tree. Coaching closes the loop: a correction made on a past conversation, whether tone, missing context or a wrong answer, is applied automatically to later interactions on every channel. The vendor publishes ambitious figures. Agents are said to be eight times more productive than human ones, with an 84% automated resolution rate, a 162% rise in CSAT and 357% return on investment, across more than 4.1 billion conversations and over 350 enterprise customers in 85+ countries. Named references include Cebu Pacific, IPSY, Pinterest, monday.com, Sky, Barnes & Noble, Grab, Life360, Loop Earplugs, Betsson, Epos Now and Malaysia Airlines. Compliance is a central argument rather than a footnote. Ada holds SOC 2 Type II certification and claims GDPR, HIPAA, PCI DSS and AIUC-1 compliance, the last being an AI-specific standard the company helped write and was first to earn in its category. Every response is verified against approved sources before it reaches a customer, zero data retention applies with all model providers, and the terms forbid those providers from training on customer data. The product is sold through consultation, not self-service: no free plan, no trial and no published price.
What it does
- Resolves customer enquiries end to end on voice, messaging and email without handing off to a human
- Executes multi-step internal procedures through Playbooks, pulling live data from business systems
- Answers in the customer's language: 60 languages on messaging and email, 42 natively on voice
- Hands over to a human agent with the full conversation context and customer record when needed
- Improves itself through Coaching, where feedback on a past conversation applies automatically to later ones
- Measures automated resolution, CSAT, handle time and cost per resolution by channel, language and region
- Acts inside connected systems such as Zendesk, Salesforce and Twilio instead of only replying
When to use Ada / When not to
A quick filter to help you decide if Ada is the right fit.
When to use Ada
- Customer service and CX leaders in large organisations handling high contact volumes across several channels
- Multilingual, multi-country operations that need one agent to answer in a customer's own language
- Regulated industries named by the vendor: financial services, health and property insurance, gaming, travel, ecommerce and SaaS
- Support teams whose enquiries follow multi-step internal procedures — order status, returns, flight disruption, account updates — rather than simple FAQ lookups
- CX teams that want to own and coach the agent themselves instead of filing tickets with engineering or with the vendor
When not to use Ada
- Small businesses and solo operators: there is no free plan, no trial and no self-service sign-up, only a negotiated order form
- Anyone who needs to size a budget before talking to sales, since no price is published anywhere on the site
- Teams looking for a simple scripted chatbot or a website widget, which is far below what this platform is built and priced for
- Use cases outside customer service — marketing copy, code generation, image or video creation are all out of scope
- Organisations that require on-premises or sovereign deployment, as the platform is SaaS hosted on AWS with an optional Azure database
How to use Ada
A typical end-to-end flow, from setup to results.
- Request a consultation through the Speak to an expert form, the only entry point to the product
- Attend the free consultation: a review of current channels, suggested AI use cases, a tailored demo and comparable customer results
- Negotiate and sign an order form, which fixes scope, billing model and term
- Connect knowledge sources so the agent has something accurate to answer from: help centre, Contentful, GuruKB, Docusaurus, GitHub or Paligo
- Set the agent's persona, tone of voice and content targeting for specific customer segments
- Write Playbooks for the procedures the agent must follow, step by step, using live data
- Connect business systems such as Zendesk, Salesforce or Twilio so the agent can act, authenticate customers and hand off with context
- Deploy the channels: web widget, iOS, Android or React Native SDKs inside your own app, Zendesk and Salesforce blocks, voice and email
- Test before launch, then monitor automated resolution and CSAT in the Performance Center
- Use Coaching to correct tone, context and answers, with improvements propagating across every channel
Pros & Cons
Pros
- One reasoning layer shared by voice, messaging and email, so a fix made on one channel takes effect everywhere
- Compliance is documented and auditable: SOC 2 Type II, GDPR, HIPAA, PCI DSS, AIUC-1, a public trust centre and a dated subprocessor list
- Firm contractual stance on training: model providers are forbidden from training on customer data, with zero data retention enforced across all of them
- Customer results are named and quantified in individual case studies rather than left anonymous
- Wide language coverage, with the 42 voice languages listed explicitly rather than only claimed as a number
- Public and complete developer documentation, including an MCP server and an OpenAPI specification
- A live voice demo agent can be called on +1 833-573-6312 without a contract or a form
Cons
- No published price at all: no grid, no range, no entry point, so budgeting requires a sales conversation
- No free plan and no trial are advertised, and there is no self-service sign-up
- Annual terms renew automatically with a 10% fee increase unless notice is given 30 days in advance
- Aggregated and de-identified information falls outside Customer Data and may be used by Ada without restriction
- No Article 27 representative is designated, although the publisher is established outside the European Union
- General retention periods are not quantified; the only published figure, 90 days after termination, covers data received through Meta services
- The published data processing addendum is dated 20 July 2023, well behind the April 2026 privacy policy and the March 2026 subprocessor list
Pricing & Plans
No free plan and no free trial are advertised, and no price figure is published anywhere on the site. Ada describes a usage-based model in which the customer is billed for each conversation between its AI agent and an end user, with a resolution-based alternative offered to enterprises that have specific requirements. Fees are set in a negotiated order form, invoices are payable within 30 days by wire transfer or ACH, and renewal terms carry a 10% fee increase unless 30 days' notice is given. A public ROI calculator is offered in place of a price list. The only cost-free item announced is the initial consultation with an ACX expert.
- the customer pays for every conversation the AI agent has with an end user. No public price.
- billed on resolutions rather than conversations. No public price.
- reasoning
- agent behaviour controls
- every channel
- measurement and coaching
- human handoff and integrations. No feature tiers are published.
Data, GDPR & hosting
A consolidated view of how Ada handles your data.
GDPR overview
GDPR implementation is concrete and documented. The privacy policy, last updated on 21 April 2026, sets out legal bases (contract, legitimate interest, legal obligation, consent) and the rights to access, correction, deletion, restriction and withdrawal of consent, with deletion requests answered typically within 30 days. A data protection officer is designated at dpo@ada.support, and privacy requests go to privacy@ada.support. A data processing addendum and a dated subprocessor list are both published. International transfers rely on the European Commission's adequacy decision for Canada, on standard contractual clauses and on the DPA. One gap is worth noting: no representative under Article 27 is named anywhere on the site, although the company is established outside the European Union. End users of Ada's customers must address their requests to that customer, not to Ada.
Who owns the data?
The customer terms of use state that, as between the two parties, the customer owns all right, title and interest in its Customer Data. Ada receives a worldwide, royalty-free, non-exclusive licence to access that data in order to provide, maintain and improve the service, to meet its contractual and legal obligations, and to create aggregated and de-identified information. That aggregated and statistical information is expressly excluded from Customer Data and from the customer's confidential information, and Ada may use, transfer, modify and display it without restriction. Ada separately owns the services and the Ada models. Generic inputs containing no confidential information may be reused for Ada's own purposes or for other customers.
Reuse rights
Reuse rights are asymmetric, and worth reading closely. The customer keeps ownership of its own data and may extract it: after termination Ada retains Customer Data for a reasonable period and lets the customer request and export it. Ada's own reuse is bounded on one side and open on the other. On the bounded side, the terms forbid the third-party model providers from storing Customer Data beyond what is needed to deliver the service and from training or fine-tuning any model on it, and the trust pages state that customer data is never used to train models, with zero data retention enforced with every LLM provider. On the open side, Ada may build aggregated and de-identified information from customer data and then use it for any purpose without restriction, including to improve its own models, and it may reuse inputs it considers generic for other customers. On the marketing site itself Ada uses analytics and advertising cookies, offers the usual industry opt-out mechanisms, and states plainly that it does not honour browser Do Not Track signals.
Data retention & training
Hosting summary
Ada hosts its platform on Amazon Web Services, in regions including the United States, Canada, Ireland and Germany, with Microsoft Azure available as an option for database hosting. The published subprocessor list gives a location for each third-party service, generally the United States, Canada or Europe. Data is encrypted in transit with TLS 1.2 or above and at rest with AES-256, behind role-based access control, multi-factor authentication and audit logging. Cross-border transfers rely on the European Commission's adequacy decision for Canada, on standard contractual clauses and on the data processing addendum. Note that the website itself sits behind Cloudflare's anycast network: the IP address the domain resolves to, geolocated in the United States, says nothing about where customer data is actually processed.
Things to keep in mind
Risks and trade-offs to weigh before adopting Ada.
- Pricing is invisible until you meet a salesperson, which makes independent comparison with rival platforms difficult
- The 10% automatic uplift at each annual renewal compounds quietly if nobody diaries the 30-day notice window
- Aggregated and de-identified information escapes the definition of Customer Data and can be used without restriction, including to improve Ada's own models
- Inputs the vendor considers generic may be reused for other customers, a boundary that is easier to write than to audit
- Hosting spans the United States, Canada, Ireland and Germany, so data residency is something to negotiate, not something granted by default
- The published performance figures — eight times more productive, 84% automated resolution, 357% ROI — are the vendor's own claims and should be tested against your own volumes
- Automating the majority of contacts erodes the human team's exposure to real customer problems, which is where product insight and escalation judgement are usually learned
Setup & Integrations
Technical difficulty
Moderate, and never self-service. Everything starts with a sales consultation and a signed order form. Day-to-day administration is designed to be no-code and run by the CX team: knowledge sources, persona, Playbooks and coaching need no engineer. Genuine technical work remains for the integration layer — connecting Zendesk, Salesforce or Twilio, authenticating customers, embedding the web or mobile SDKs, using the APIs and MCP server. Ada offers its ACX Practice methodology and dedicated ACX Experts alongside, and publishes full developer documentation.
Deployment
Integrations
Supported languages
Behind Ada
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Which channels does Ada cover?
How many languages does Ada support?
How much does Ada cost?
Is there a free plan or a free trial?
Is customer data used to train AI models?
Who owns the data?
Where is the data hosted?
Which certifications does Ada hold?
Does Ada offer an API?
Can I try the product before signing anything?
Should you pick Ada?
Ada is a mature, enterprise-grade product rather than a tool you try over a weekend. Its strongest argument is architectural: one reasoning layer serves voice, messaging and email, so what a team fixes on one channel takes effect on all of them, and the Coaching loop turns that into steady improvement instead of a rebuild. Its second strongest argument is governance. The customer keeps ownership of its data, model providers are contractually barred from training on it, zero data retention applies across every LLM provider, and the compliance posture — SOC 2 Type II, GDPR, HIPAA, PCI DSS, AIUC-1, a public trust centre, a dated subprocessor list — is unusually well documented for this category. The reservations are real too. Nothing on the site lets a buyer estimate cost: no grid, no range, no entry price, only a usage-based model described in words. There is no free plan, no trial and no self-service route in, so evaluating the product means engaging a sales cycle. The contract renews annually with a built-in 10% increase unless notice is given. Aggregated and de-identified information sits outside the customer's ownership and may be used freely, including to improve Ada's models, and inputs judged generic may be reused for other customers. No Article 27 representative is designated despite the publisher being outside the European Union, general retention periods are not quantified, and the published data processing addendum is nearly three years older than the privacy policy. The honest summary is that Ada suits customer experience organisations in large, multilingual companies with the volume to justify the investment and the team to run the agent themselves. Smaller structures, and anyone who needs a price before a meeting, should look elsewhere.
- Choosing a selection results in a full page refresh.
- Opens in a new window.