Ada Wants You to Prefer the AI Agent Over the Human

In 2014, two founders in Toronto watched their first startup quietly come apart. The product was a social network for creators called Volley. The problem was not the product. It was the inbox. Customers wrote in faster than two people could ever answer, and every unanswered message was a small crack in the whole thing. Mike Murchison and David Hariri could have hired their way out. Instead they asked a stranger question: what if the answers could handle themselves?

That question became Ada. A decade later it is a company valued at $1.2 billion, powering more than four billion automated customer conversations for brands including Meta, Verizon, Square, and YETI. Its AI agents resolve the kind of everyday requests - where's my order, change my flight, reset my password - that pile up by the hundreds of thousands at any large company. The best-performing customers now automate more than 80% of their support inquiries. But the origin is worth sitting with, because Ada did not start with a model. It started with a job.

01 / The MethodThey worked the support desk on purpose

Before writing much of anything, Murchison and Hariri did something most software founders would never sit still for. They convinced seven companies to let them work as customer service agents. For roughly a year they answered tickets, sat in the queues, and felt the specific grind of the job - the repetition, the scripts, the same twelve questions arriving in a thousand slightly different shapes. Only then did they start quietly running an early version of Ada inside one of those companies, handing their non-technical colleagues a system that could take the boring parts off their plate.

That year is the part of the story that still shows up in the product. Ada was built by people who understood that a support agent's day is mostly pattern, and that patterns are exactly what software is good at absorbing. The company is named after Ada Lovelace, the 19th-century mathematician often called the first computer programmer - a fitting nod for a company that thinks a machine can do work people assumed was too human to hand off.

02 / The ProductResolve, act, improve - not just chat

The word Ada uses for its platform is "agentic," and the distinction it is drawing matters. A chatbot answers a question. An AI agent, in Ada's framing, does the thing - it reads intent, pulls the right answer from a company's own knowledge base, and then takes action inside connected systems like a helpdesk or CRM. It can process the refund, not just explain the refund policy. When it hits a wall, it hands off to a human with the context already attached.

Ada calls its product the ACX Platform, and it runs across voice, messaging, email, and social channels from one place. Teams manage it through a no-code, drag-and-drop interface, which is deliberate: the people who know the support playbook best are usually not engineers. In February 2026, Ada shipped its most technically ambitious piece yet - a patent-pending Unified Reasoning Engine. The idea is one "brain" for the AI agent across every channel, so whether a customer types a message or picks up the phone, they meet the same knowledge, the same policies, and the same brand voice.

03 / The CustomerBuilt for the companies drowning in tickets

Ada is not chasing the corner coffee shop. Its sweet spot is enterprises with at least 300,000 support conversations a year - the point where "just hire more agents" stops scaling and starts hurting. Those are airlines, telecoms, marketplaces, and consumer brands: AirAsia and Cebu Pacific in travel, Verizon and Digicel in telecom, Square and Monday.com in software, Grab and Life360 in consumer apps. More than 350 brands run on it.

For that buyer, Ada leans on a number rather than a vibe. It reports customers seeing a 943% return on investment within four months, an eight-point lift in customer satisfaction, and nearly half of interactions resolved without a human touching them. When support volume is measured in millions, small percentages turn into real payroll and real wait-time math.

04 / The DifferenceThe feedback loop is the moat

Ada operates in a crowded and fast-moving lane. Intercom's Fin, Sierra, Decagon, Salesforce's Agentforce, and Zendesk's own AI are all reaching for the same enterprise budget. What Ada emphasizes is less the model and more the loop around it: agents that get measured, coached, and corrected like members of a team rather than set-and-forget bots. Its investors at Bessemer described this as "fanatical CX loops" - the continuous tuning that pushes resolution rates up over time instead of letting them drift.

The other difference is posture. Ada describes itself as AI-native, meaning the platform was built around the agent from the start rather than bolting AI onto a legacy helpdesk. It builds on Microsoft's Azure OpenAI Service for its models and enterprise-grade security, and its 2025 work extended Playbooks and Coaching to voice, so an AI agent can walk through a multi-step phone call the way a trained rep would.

05 / The PathFrom near-bankruptcy to unicorn

Ada raised over $200 million on the way here, with backing from Accel, Tiger Global, Bessemer Venture Partners, Spark Capital, and FirstMark. It sells the way enterprise software sells - annual and multi-year contracts priced on conversation volume rather than per seat, with custom quotes commonly cited starting around $30,000 a year. There is no free tier and no published price sheet, which is itself a signal of who the product is for.

Whether customers really will come to prefer an AI agent over a person - Ada's stated ambition - is still an open question, and one every company in this space is quietly betting on. What Ada has going for it is the thing it built first: a genuinely close read of the job it is trying to automate. In early 2026 it reported doubling year-over-year growth as demand for agentic customer service surged. The queue, it turns out, was always the product.