He once solved a customer's problem with gigabytes of machine-learning infrastructure. His co-founder solved it better with 100 lines of code. That sting became Cotera.
Ibby Syed. The grin of a man who fired his best-paying customers - and was right to.
Walk into a Cotera sales call and Ibby Syed will not show you a slide deck. He will build you a working AI agent, live, in about three minutes - one that watches your call transcripts in Snowflake and quietly drafts the case study you forgot to write. The demo is the pitch. The product talks for itself.
Today Cotera calls itself the agentic AI platform for enterprise work: software that lets a team build, run and trust AI agents that take real action across operations, support and go-to-market. The agents live on top of a company's own data warehouse - Snowflake, Redshift, BigQuery - so the data never leaves the building. Bilt, OpenSea, Skims and Coterie are among the names on the wall. The number that matters is roughly a million dollars in annual recurring revenue, earned by 15 enterprise customers, run by a team of about ten.
None of that is the interesting part. The interesting part is how badly it nearly went the other way.
Before the agents, before Y Combinator, there was a spreadsheet problem at a fast-growing exercise-bike company. Ibby Syed left Harvard in 2019 with an applied math degree and a short stint at Drift doing time-series and NLP work, then walked into Peloton as its very first data science hire.
That is a particular kind of job. There is no team, no pipeline, no playbook - only a company convinced its data is worth something and no one yet to prove it. Syed built the plumbing: the early ELT processes, the data transformations, the first production machine-learning libraries. He rose from Data Scientist to Lead to Manager, running data science for retention, growth and customer experience - the unglamorous math of predicting who will upsell, who will churn, and which support tickets are about to become a problem.
It taught him the lesson that would later define Cotera: the value of data is not the dashboard. It is the decision someone makes because of it. A churn chart that tells you 25% of customers left is a fact. It is not a feature. Once you have seen the chart, you have no reason to open it again.
Syed would spend the next several years learning that lesson the hard and expensive way - by building a company on top of exactly the thing he should have distrusted.
"People aren't actually logging into the product. The value they get is from talking to me and my team."
— Ibby Syed, on the moment he realized Cotera was a consulting firm in disguise
Cotera began in 2021 when Tom Firth got into Y Combinator and Syed jumped in beside him. The first product was a customer-analytics platform: connect it to a warehouse, compare cohorts, build personas, see why people churn. Through the W22 batch and a seed round, the two founders ground out customers the old-fashioned way - two-sentence LinkedIn messages, the occasional Starbucks gift card, 25 meetings a week between them, an 8-to-10% response rate that any sales leader would envy.
It worked. The first customer paid $200 a month for an MVP Tom built on a Jupyter notebook in days. Eighteen months later they were at $150,000 to $170,000 in ARR. By every startup vanity metric, this was traction.
It was also quicksand. Nobody logged in. Customers got their value from phone calls with Syed and his team, not from the software. He had accidentally built a consulting business with employees and revenue - which is precisely the kind of business that is psychologically impossible to walk away from.
In 2023 a customer asked for something small: pull the topics out of our support tickets. Syed - the trained data scientist, the man who built Peloton's ML stack - answered the only way he knew how. He stood up a heavy, gigabytes-deep data science solution. It worked, roughly.
Then Tom Firth wrote about 100 lines of code against the new OpenAI API. It did the same job. Better. Faster. Cheaper. No infrastructure.
"That was when it clicked," Syed has said. The ground had moved. The expensive expertise he'd spent a career accumulating had just been undercut by a few prompts - and if it could happen to him, it would happen to everyone. The question stopped being how do we do this for customers and became how do we let customers do this themselves.
So Cotera became an AI agent builder. Fully prompt-based - no drag-and-drop flowcharts. You write what you want the way you'd brief a new hire, hand the agent some tools and access, set how often it runs, and let it loose.
To force the company onto product instead of services, Syed deliberately let go of the consulting-heavy customers - the ones generating the comfortable revenue. Cutting income to find focus is the least intuitive move a founder can make. He made it anyway.
A major retailer wanted a bot to scan Poshmark for stolen merchandise. Instead of building it, Syed ran a training session so the retailer's own ops team could build it in Cotera. They got excited - and started building more agents on their own.
His outbound now sends prospects real leads - generated by Reddit-monitoring agents that surface people complaining about the exact problem Cotera solves. Value first, pitch never. In a saturated inbox, the working thing wins.
"We are still the AI guys. We just now teach instead of do things for people."
— The pivot, in one sentence
Graduates Harvard, B.A. Applied Mathematics.
Data Scientist at Drift - time-series and NLP modeling.
Peloton's first data science hire; climbs to Manager of Data Science across retention, growth and CX.
Co-founds Cotera with Tom Firth.
Y Combinator W22; seed round; churn-analytics product.
The OpenAI moment. The pivot to AI agents begins.
Rebuilds Cotera as a prompt-based agentic platform on enterprise warehouses.
~$1M ARR, 15 enterprise customers, team of ~10.
Roughly the same founders, the same customers, the same NYC office. Different answer to one question: do we do the work, or do we hand people the tools?
Syed names focus as the trait that separates founders who ship from everyone else. "Focus is very, very difficult." Easy to say; the firing of paying customers is what it actually costs.
His blunt forecast: hyper-specific AI startups will rocket to millions in ARR and then collapse, because what's trivial to build is trivial to copy. Horizontal tools with real moats survive. He names Zapier and n8n - not other AI startups - as Cotera's true rivals.
"I don't want to be a billionaire who's alone." He stays in New York, guards time for his wife and friends, and treats that boundary as a feature of the plan rather than a tax on it.
There's a line he borrows from the show White Collar to explain the emotional physics of a startup: a happy ending depends on where you stop the story. He pictures the founder's life as a sine wave, each trough hopefully a little higher than the last. It's an oddly honest frame from someone selling optimism - which may be exactly why customers buy it.
He is also impatient with effort that masquerades as work. Syed's view is that most people move through their days thoughtlessly, defaulting to volume when what the moment needs is diligence. The whole premise of Cotera is the opposite instinct: hand a person an agent that does the thoughtless part, and you free them to think. The pitch and the philosophy are the same sentence said twice.