Tech
Startups pitch big AI ideas during mini-competition at GeekWire’s ‘Agents of Transformation’
Big Tech is not alone in the AI innovation race. Four startup founders took the stage at GeekWire’s Agents of Transformation event Tuesday in Seattle for a rapid-fire pitch competition.
Ideas from Pay-i, Cascade, Autessa and GemaTEG were pitched to the crowd and a panel of judges, with Pay-i founder David Tepper emerging as winner and most impressive under pressure.
Judges Bryan Hale of Anthos Capital, Yifan Zhang of AI House, and T.A. McCann of Pioneer Square Labs said they were looking for someone who was “both great at presenting but also fantastic at answering the questions.”
Read more about each pitch:
Pay-i (pitch by David Tepper, founder/CEO)
An AI spend management platform that tracks ROI across an organization’s entire AI footprint — not just tokens, but the full cost stack including models, tools, and GPU resources.
David Tepper argued that tokens account for only 72% of the total expense associated with AI, and that the complexity multiplies fast when agents are drawing on multiple models, enterprise discounts, and rented GPU banks simultaneously.
Born from his days tracking Microsoft’s internal Gen AI spend on Excel spreadsheets — a period when he says he once saved his division $300,000 a week by simply asking the right questions — the company targets enterprises spending at least $500,000 on AI annually.
“After all the hype and FOMO wears off, there’s three letters that are going to survive the AI revolution, and that’s ROI,” he said.
Cascade AI (pitch by Ana-Maria Constantin, co-founder/CEO)
An agentic HR and IT support platform that deploys AI agents to handle sensitive employee situations — benefits navigation, mental health resources, leave management — confidentially and around the clock, freeing HR teams for higher-stakes human judgment.
Ana-Maria Constantin opened her pitch with a show of hands, asking the audience whether they’d ever hesitated to go to HR because they weren’t sure whose side HR would be on.
“Imagine if that’s the case for the people in this room, senior leaders working for some of the most successful companies in the world,” she said. “Imagine how regular employees are feeling. That’s the problem we’re working on at Cascade AI.”
Autessa (pitch by Roshnee Sharma, CEO)
A platform that replaces off-the-shelf SaaS with custom-built software staffed by “AI employees” — agents that handle workflows like lead qualification and order processing.
Roshnee Sharma’s pitch opened with a crowd-participation moment: what does SaaS really stand for? “Software as a spend,” she declared.
The company targets mid-market businesses with $20 million to $500 million in revenue, and prices its AI employees at roughly $7 to $10 each.
Judges pushed back on whether results were truly cost-saving or merely cost-neutral; Sharma argued the savings are real because clients avoid having to hire additional headcount: “We didn’t fire people. We got people able to do more of the work that they wanted to do.”
GemaTEG (pitch by Manfred Markevitch, co-founder/CEO)
The outlier of the group: a hardware thermal management company targeting AI data centers, using solid-state cooling technology that requires no water and uses 40% less power than conventional systems.
“AI runs on hardware. It’s not only software,” co-founder and CEO Manfred Markevitch told the crowd, noting that a conventional hybrid-scale data center can consume a million gallons of water per day.
GemaTEG’s granular approach cools at the individual chip level rather than the whole building, and the company claims its systems perform twice as well as conventional ones on a per-watt basis. The company already has installations with the U.S. Department of Energy, and partners in Italy and Switzerland.
Hyperscaler deployment is one to two years out, with chip manufacturer design-in conversations already underway. Judges pressed hard on customer lock-in risk; Markevitch compared the stickiness of their solution to Intel Inside — once designed in, it spans multiple chip generations.
Related coverage:
You must be logged in to post a comment Login