Nine years in steel cylinders running nuclear reactors. Then McKinsey. Then Palantir. Then the corner office at Citi Bike, sold to Lyft. Now he is teaching AI agents to do the work most chatbots only pretend to.
Kris Sandor runs Readyly out of New York with a team of about thirty-two people and an unfashionable conviction: that the chatbot era is closing. Not because language models cannot chat, but because most organizations do not want to chat. They want a permit issued, a refund processed, a wildfire alert translated into the right language by 9 a.m. Readyly builds the agents that do those things across voice, chat, email, and web. They fill forms. They integrate with the systems behind the systems. They translate. They route. They escalate to a person who actually has time to listen.
If that sounds like a slide, it is also a backlog. Readyly has been named to the U.S. Chamber of Commerce list of America's Top 100 Small Businesses, and to Context VC's Top 150 veteran-founded startups. Both lists are mouthfuls. The work behind them is not.
“These are businesses that instead of spending money on customer support, they're spending that on marketing, on developing a new product, which from a macro economic perspective is much better for our economy.”
Kris Sandor — SABM podcastThere is a useful test for a founder: ask what they would do if the company disappeared tomorrow. Most invent another company. Sandor, by the shape of his career, would probably just start operating something else. The pattern is consistent. He picks an unloved system — a bike-sharing network, a tier-one support queue, a municipal call center — and turns it into infrastructure people stop noticing because it works.
A rough chart of a deliberate career. Bars are vibes, not years.
Readyly's agents handle inbound voice across languages without funneling callers through eight numbered menus to reach the wrong department.
The agents read, write, and act — submitting forms and pulling from integrated systems instead of replying with links.
Permits, resident inquiries, emergency alerts, multilingual notifications. Built for municipalities that have never bought AI before.
Sandor's stated ambition: automate a large share of routine support so headcount moves to product, marketing, and care that needs a human.
Anthropic Claude, OpenAI, Gemini, Code Llama, Mistral. Vectors via Pinecone and Weaviate. LangGraph for orchestration. React and Python in the building.
High-security posture, secure integrations, government-grade requirements baked in — the leftover habit from a career adjacent to defense.
He speaks Hungarian. The surname is a clue. He likes mentoring; the SABM interview makes that obvious without him saying it directly. He spent his twenties below the ocean's surface running a reactor, which is one way to learn that systems work or they kill you. He talks about customer support not as a cost line but as a redirect of capital toward growth — an unusually macroeconomic frame for a founder selling SaaS.
There is a quiet pattern. The Naval Academy is service. McKinsey is craft. Palantir is mission. Citi Bike is infrastructure. Readyly is all four at once, applied to phone trees and permit windows. The constant is that he picks problems other people have given up on solving and approaches them like they are mechanical, not magical.
Most AI companies sell to other software companies because software companies move fast and pay quickly. Sandor went the other way. Readyly's loudest keywords read like a municipal procurement plan: multilingual emergency alerts, permit automation, resident inquiry handling, public safety communication, ai for civic engagement. There are reasons to find this strange. There are better reasons to find it shrewd. Governments do not churn. They buy slowly and renew forever. They have the highest stakes for getting an interaction right, and the lowest budgets for hiring more humans to do it.
A submariner understands constraints. A McKinsey consultant understands incentives. A Citi Bike CEO understands what it costs to keep something physical running through a New York winter. Stack those skills and a govtech bet starts to look obvious rather than contrarian.
From below the ocean, to above city hall — the operator's playbook stays the same: respect the system, automate the boring parts, escalate the rest to humans who have time to care.
YesPress Dossier — Editorial Note