John Kessler sent Cameron Fink a LinkedIn message. He was looking for advice about an accelerator. What followed was a conversation scheduled for half an hour that lasted two and a half hours. Fink then called his friend Ned Koh. The stranger on the screen, he told him, was the smartest person he had ever met.
That is a pleasingly human beginning for Aaru, a company devoted to simulating people. A message arrived. Somebody answered. The meeting ran long. No predictive machinery was required to recognize that the afternoon had changed direction. Fink’s account has the small, untidy particulars that disappear when a founding story is reduced to a fundraising announcement.
He is now Aaru’s cofounder and chief executive, based in New York. The business builds simulations intended to anticipate how populations will respond to products, prices, advertising and other decisions. Its customers want an answer to a question that has kept whole industries employed: what will people actually do?
Fink’s story is about the pursuit of that answer, and the less predictable business of finding the people to pursue it with.
“Thank God, this time I did answer my LinkedIn DMs.”
Cameron Fink, recalling the introduction to John Kessler
The friendship before the forecast
Fink and Koh met on the first day of their freshman year of high school, in their first class. Their early collaborations included programming their way around school Wi-Fi so they could avoid handing in homework on time. They also started a quantitative insurance business. Entrepreneurship was already an activity they shared, rather than an identity waiting for a business card.
The distinction matters. A longstanding collaborator has seen your ideas before they acquire a pitch deck. He knows which enthusiasms survive the weekend. By the time Kessler appeared, Fink and Koh had already practiced the awkward work of making things together. The company would inherit a friendship as well as a technical ambition.
Fink’s childhood crossed the United States, Australia and Canada. His early work included research in AI, machine learning and quantitative political science. He founded four companies while still at high school, later attended Dartmouth, and left before completing a degree. The direction was toward building, with formal education becoming one route he chose to leave.
Those details help explain the range of the problem he eventually took on. Aaru sits between disciplines: the engineering of models, the interpretation of behavior and the practical question of whether someone will buy what a business is selling. An elegant answer in one field can still be useless in another.
A company made of hypothetical customers
Aaru was founded in March 2024. Fink became CEO, Koh president and Kessler chief technology officer. Kessler brought experience in simulation research at MIT’s City Science Lab. Their shared question was whether human behavior could be modeled at the scale and complexity of the world outside the computer.
The practical starting point is familiar enough. A business has a proposed product or campaign and a particular audience in mind. It needs to know how that audience might respond. Traditional research recruits people, asks questions, collects answers and interprets the results. Each stage costs time; each also creates opportunities for the sample or the question to bend the answer.
In Fink’s description of Aaru’s approach, observed behavior carries particular weight. Purchase records, advertising responses and election outcomes show decisions people made. His argument is that the difference between a stated preference and an actual choice is a central obstacle to understanding customers. Most of us have supplied evidence for that proposition, usually somewhere between a carefully prepared shopping list and the checkout.
The commercial promise is the ability to test a decision before committing to it. Change a price. Try another message. Look at a different audience. For a chief executive selling simulation, the challenge is to make those experiments useful enough that an organization changes how it works.
- 01Define an audienceWho might respond?
- 02Model responsesWhat choices might they make?
- 03Compare optionsWhich decision looks stronger?
- 04Check outcomesHow well did the model perform?
Conceptual illustration of the decision cycle, not a product screenshot.
The forecast that met an election
A prediction business acquires a particularly unforgiving kind of deadline. Eventually, something happens. In the 2024 US presidential election, Aaru’s model predicted that Kamala Harris would win the Electoral College. Donald Trump won instead. Most of the predictions in the test went wrong.
Fink defended the result by pointing to the narrow margins and the model’s margin of error. He argued that a close probability should be understood as a close probability. He also maintained that AI polling offered advantages in speed and cost. Those were his defenses of a public result that had failed to name the winner.
For the profile of a young founder, this episode is more revealing than another valuation milestone. It places his central ambition against an observable outcome. Probability allows for uncertainty; customers still need to decide what to do. The gap between those two truths is where much of the work of selling prediction takes place.
A simulated population can produce a tidy table. Reality has no obligation to cooperate with its formatting. For Fink, the business therefore depends on explaining what a result supports, how much uncertainty it carries and what a decision-maker should make of it.
Getting invited into the room
The next phase brought Aaru into established businesses with their own research operations. In March 2025, Accenture announced an investment through Accenture Ventures and a collaboration with Accenture Song. The plan included integrating Aaru’s private-sector model, Lumen, into work involving product development, marketing, customer strategy and customer service.
That relationship gave Fink’s company a route into projects where simulation had to fit an existing commercial process. It is one thing to demonstrate a model. It is another to have a strategist use its output while deciding how to spend a client’s money. A partner brings access, but also colleagues with questions, deadlines and methods of their own.
Interpublic followed with a partnership in August 2025. Its plans covered campaign development, creative testing and audience work, with simulation to be incorporated into its Interact platform. Chief solutions officer Jayna Kothary was also joining Aaru’s strategic advisory board. The agreement connected Aaru’s forecasts with an advertising network’s data and creative operations.
Fink described the intent as helping creative people do their work. That is an interesting sales task for a founder whose product appears capable of putting a numerical judgment on an idea before the public encounters it. The value depends on whether the number helps someone take a better risk. A beautifully quantified reason to do something dull would be a disappointing use of all that computation.
When the customer already knows the answer
EY offered a concrete comparison. In August 2025, its wealth and asset management team tested whether Aaru could reproduce findings from its global research. The original work covered 3,600 affluent investors across more than 30 markets. The simulation took a day; the conventional project took six months.
Across 53 matched single-choice questions, EY reported a median Spearman correlation of 0.90. That measures agreement in rankings, rather than a universal percentage of correct predictions. Some answers diverged substantially. The distinction is useful: a headline number introduces a result, while the underlying differences determine what someone can responsibly do with it.
One day.
Median rank correlation: 0.90 across matched single-choice questions. The original study covered 3,600 investors in more than 30 markets.
A SPECIFIC TEST, NOT A GUARANTEE OF FUTURE RESULTSFor Fink, this kind of exercise gives an unfamiliar product a familiar examination. A prospective customer possesses a body of research and can inspect the comparison. The conversation moves from whether simulation sounds plausible to where its answers agree, where they differ and whether those differences matter.
In September 2026, EY announced a formal alliance with Aaru. The two organizations described work on decisions involving messages, pricing, products and business strategy. The relationship had progressed from a test of the machinery to a means of putting it to work.
The future has a payroll
Redpoint announced in April 2026 that it was leading Aaru’s $80 million Series A. For Fink, the financing brought the resources to expand a company whose product demands both technical development and customer trust. Money can pay for experiments and employees. It cannot settle whether a forecast deserves belief.
Aaru’s September 2026 evaluation put a larger set of results into view: 2,993 questions across 186 studies. The company reported a mean observed total variation distance of 7.62 points and a mean absolute error of 3.53 percentage points. These describe different kinds of error; neither means that the system is correct on every future decision.
The progression is striking enough without pretending the uncertainty has disappeared. A friendship became a founding team. A model became something enterprises tested. Partnerships and financing supplied room to grow. At each step, the claim became more consequential because more people might act on it.
Fink now has to manage the institution around the idea. A prediction company needs engineers, researchers and people who can explain an unfamiliar result to a skeptical buyer. It also needs the discipline to separate an attractive story about its technology from the work the technology has actually demonstrated.
The most appealing detail in his origin story remains that overlong first conversation. A founder trying to anticipate human decisions found a collaborator through an unsolicited message and an afternoon that escaped its calendar slot. The future arrived as a person asking for advice. Fink had the good sense to keep talking.