How Ada Grew Toward $70M Revenue by Answering Thousands of Tickets by Hand First
Ada became the customer-support automation leader by refusing to automate first. Mike Murchison's early talk lays out the philosophy — software isn't valuable, it's a multiplier — and the dataset shows what it compounded into: ~$70M revenue and a unicorn round.
On this page
Years before Ada became the reference platform for AI customer support, its co-founder Mike Murchison stood in front of a room of founders and told them their software wasn’t valuable. Not as a provocation — as a method. “Software actually isn’t valuable,” he argued. “Software is a multiplier of value that is created in the physical world.” Users, praise, even revenue — all “very dangerous proxies for value creation in the early days.”
Then the line that explains the company: “With Ada, we actually answered thousands of customer support tickets a month manually before we wrote one line of code. That was very, very, very painful.”
Software as medicine
The method has three steps, in Murchison’s own framing.
- Do the job by hand until it hurts. Genuinely hurts, at volume, for months.
- Write software as medicine. “You’ll know what software to build because you were feeling the pain so intensely… it’ll be so obvious.”
- Keep the analog past as your go-to-market. “The way you sell to prospects is by communicating stories around how your software takes away pain that you yourself experienced. That is your competitive differentiator.”
It’s the same discipline other archive founders arrived at independently — Workboard coaching alignment by hand before tracking it in software, UserTesting building its business on literally watching users struggle — but nobody stated it more cleanly, or earlier. By the time of the talk, Ada had gone from answering thousands of tickets a month by hand to automating hundreds of thousands a day. Asked whether AI would automate everything within a decade, his answer was characteristically unhyped: “Things have been being automated for hundreds of years… viewed from a historical lens, I don’t think this era will be that distinctly different.” That was years before the LLM wave made his company’s category the hottest in enterprise software.
What the grind compounded into
The GetLatka dataset carries Ada’s trajectory from philosophy to scale:
| Year | Revenue | Event |
|---|---|---|
| 2020 | ~$18M | $44M Series B |
| 2021 | $35M | Series C: $130M at $1.07B — unicorn |
| 2022 | ~$46M (est.) | — |
| 2023 | ~$58M (est.) | — |
| 2024 | ~$70.6M (est.) | — |
The Toronto company rode the exact curve Murchison predicted calmly: chatbot skepticism, then the generative-AI turn that transformed “automated support” from cost-cutting purchase into board-level mandate — with Ada holding the incumbent’s advantage of a decade of real resolution data.
The margin question the whole category now faces
There’s a sharp cross-current in our own archive worth reading against this post. Amanda Kahlow of 1Mind — building AI agents for sales — told us in 2026 that she deliberately avoids Ada’s home turf. Her full argument is here.
Support automation is “a very low margin business… you can’t have a million conversations and have it cost a lot, because there’s not a big upside to that,” whereas an agent that closes million-dollar pipeline justifies its inference bill trivially.
Hundreds of thousands of resolutions a day is exactly the volume where infrastructure discipline compounds into a moat — and where the customer’s alternative (human support at scale) is so expensive that even thin software margins price well.
Both theses are getting tested with nine-figure revenue bases now; the dataset will keep score.
Current data lives on Ada’s GetLatka profile; the original talk is here. The takeaway for builders hasn’t aged a day: before you automate anything, do it by hand until it hurts — then sell the scar tissue.
SourcesMike Murchison’s talk (linked above); the GetLatka dataset through late 2025; Amanda Kahlow’s 2026 interview on 1Mind.
