By the end of this year, the biggest change people feel in daily life will still come through work, according to Kalshi co-founder Luana Lopes Lara. In her telling, no job has yet been fully removed and replaced by AI inside the company. What she sees instead is broad AI augmentation across roles.
Lara, who co-founded prediction market platform Kalshi in 2018, discussed that shift in a podcast interview. Forbes has listed her as the world’s youngest self-made female billionaire. In May, Kalshi raised $1 billion at a $22 billion valuation. The company sits in a distinctive position: it watches where people are willing to put real money on outcomes, and when the topic turns to AI and employment, those prices tend to map changing risk rather than offer reassurance.
AI market questions are moving from capability to consequence
Lara said that one or two years ago, most AI-related market proposals on Kalshi were about capabilities: whether a model could complete a task, which model would win, or when a major math problem would be solved. Now, users are far more likely to ask how AI may affect layoffs, unemployment, and elections.
「A year or two ago, most of the markets people proposed were about AI capabilities; now, more of the requests we get are about AI’s impact,」she said.
For her, that change suggests ordinary users have already moved past asking what AI is and have started trying to price what it may cost them personally. Excitement around model capabilities, drug approvals, and quantum computing has not disappeared. But a market asking whether AI will become the leading cause of layoffs in a given month is now something people track month after month. Tech news, in that sense, is no longer confined to product launches. It is showing up in payrolls and job descriptions.
She also pointed to a long-running AI catastrophe scenario market that she and Kalshi’s co-founder have watched closely. That market has five conditions, and it resolves to yes if three of them occur. At the time of the interview, the implied probability was about 26% to 30%, and trading volume had reached several million dollars. Lara said the figure sits well above what many people instinctively assume, which is why the team keeps returning to it.
What a 73% probability does and does not mean
At the time the interview was recorded, Kalshi’s market put the probability that AI would become the top cause of layoffs in May at 73%, with roughly $30,000 in volume. Lara said AI had in fact been the leading cause of layoffs in March and April.
She added that Kalshi Research has done calibration work showing prices often settle into a fairly stable range even when volume is only around $5,000. More money and less time until resolution usually make the read stronger, she said.
「Even if the volume is only $5,000, you can already see the number converge to a pretty well-calibrated level, so the figure really is worth trusting,」she said.
Still, Lara drew a line around what that 73% means. It does not translate directly into a 73% chance that any given person’s job disappears. The market answers a narrower question: whether AI will be recorded as the top cause of layoffs in a specified month. In prediction markets, value depends on asking a clearly bounded question and letting people who are willing to absorb losses set a price. If the prompt is vague, the number loses force.
Some markets do not resolve for five years, such as questions on when a drug may be approved or when quantum computing may achieve a breakthrough. After Kalshi began paying interest on account positions and U.S. dollars held in accounts, users no longer had to avoid those trades simply because capital would be tied up for a long time. That allows the platform to accumulate more information around distant events instead of having prices chase the day’s headlines alone.
Probability is not prophecy
Lara used the election of an American pope as another example. Before the result, the probability on Kalshi had stayed near 1% for a long period. Afterward, some news coverage said prediction markets had been wrong. Her response was straightforward: 1% is not zero.
「70% is not 100%. Even if a number is 99%, it can still fail to happen once out of every 100 times; 1% is not zero either,」she said.
Her point was that a conclave is almost a closed information system, making outside prediction inherently difficult. When a low-probability outcome happens, that does not retroactively invalidate the earlier probability.
She also said that the same amount of trading volume can mean very different levels of reliability across markets. In politics, one to two thousand dollars may be enough to make a market fairly accurate a week before an election. Six months earlier, it often takes tens of thousands of dollars. Before looking at a percentage on its own, she said, look first at the resolution date, the volume, and the exact market definition. Those three conditions say more about the underlying risk than a standalone number.
Prediction markets, in her view, ask participants to pay for their judgment from the start. Better research gives traders a better chance to make money when they are right. Casual opinions, by contrast, come with a cost if they are wrong. The system does not create an all-knowing crowd, but it does turn 「I think」 into a number with financial consequences, one that can be updated continuously.
Most users come for information, not trading
According to Lara, 70% of Kalshi users do not trade at all. They open the site simply to check updated probabilities on the economy, elections, sports, and cultural events.
「Seventy percent of our users actually trade nothing. They just come in for information and to see the forecasts on different things,」she said.
She compared that behavior to reading a kind of news product written in prices. Instead of reflecting one outlet’s editorial line, the homepage shows judgments expressed in money. Kalshi also compresses thousands of election markets into its K Power index so users can read, in one number, whether U.S. politics currently leans more Republican or Democratic.
Another group uses the platform as a hedge. Lara gave the example of a bar on Manhattan’s Upper East Side that promised free drinks to customers if the Knicks won. The downside could have reached $10,000 to $20,000, so the bar bought the opposite outcome to limit that risk. During hurricane season in Florida, some people have also asked for markets tied to their specific areas, using any gains to offset insurance deductibles. Forecasting can help with decisions; hedging can limit the damage when a judgment turns out to be wrong.
She said the platform’s work is more involved than simply listing a yes-or-no question. If users say they want to know whether AI is affecting an election, the team first has to break 「affecting」 into conditions that can actually be resolved. A country’s political lean is also difficult to express through a single seat count, which is why K Power combines forecasts from the House, the Senate, and races across the map. A useful number often begins with a long process of definition.
Work is still where the biggest shift will land first
Lara did not say any job at Kalshi has already disappeared because of AI. What she sees is amplification across the board: engineers coordinating around 20 cloud agents at once, new hires using agents to catch up on company context, and managers getting faster visibility into projects that are stuck.
「I haven’t yet seen a role fully switch to ‘we don’t need this person, we have AI.’ Every role is being enhanced by AI,」she said.
She still expects work to be where ordinary people feel the most visible change by the end of the year. For now, she is more constructive on physically involved roles, such as fitness coaches, and craft-oriented jobs. She did not draw a conclusion on how far robotics may go later. Engineering roles, long treated as relatively safe, look less secure to her now. The title may remain, but the way tasks get done, the number of people a team needs, and the standard for hiring have already begun to change.
She used travel planning as an everyday example. In the past, arranging a weekend trip meant searching for places to stay and building an itinerary manually. Now she asks ChatGPT to plan two days for her, but she still has to open the websites herself and book each hotel. That break point captures the current stage, she said: AI has taken over thinking and organizing in many cases, while the final steps of acting across systems still often remain with a person.
How a 170-person company runs with an AI-first approach
Kalshi has about 170 employees. Lara described market operations as a factory that constantly tracks error counts, delays in market launches, delays in result adjudication, and coverage rates. AI systems collect those metrics at the base layer, then summarize emails, documents, and team updates for managers.
「AI gets me context faster and lets me know what is happening faster, so I can manage more threads and more people more effectively,」she said.
Sunday used to be her busiest day because she had to review what got done the previous week, what needed to happen next, and how the full set of metrics looked. Now many of those organizing tasks are handled by agents, which she said has made room for Sunday brunch. Management has not disappeared. It has shifted from moving information around to judging information. Access control remains difficult because legal materials and monitoring records can only be opened to a limited group.
The company wants every new hire to start with a personal agent connected to email, documents, and work records. The hardest piece right now is permissions. The same system has to help new employees come up to speed quickly while also making sure legal cases and market surveillance materials remain visible only to approved staff. The richer the context becomes, the less vague access design can afford to be.
More agents still mean a need for full-time owners
One of the more counterintuitive details in the interview was that Kalshi is using more agents while still hiring engineers. Lara said she and her co-founder are already fully occupied with day-to-day operations. If each of them tried to build internal AI systems with only 5% of their time, the result would not be stable. The company therefore created dedicated teams and placed engineers into market operations, design, and legal workflows to work through processes one by one.
「If I only spend 5% of my time on it, it won’t be done well. If AI is going to become an important part of the company, really strong people need to own it full time,」she said.
Actual workflows are far more complicated than demos, she said. Product feedback is scattered across Slack, Discord, posts on X, and internal discussions. To sort that automatically, the company also has to track which issues have already shipped. Kalshi tested multiple QA products and found none that met its needs. The design team, meanwhile, wants to shorten the cycle where engineers push an experiment live first and designers come back later to polish it. Stories about the solo founder are compelling, but stable AI systems still need clear owners.
Kalshi has also just launched perpetual futures, marking its first move beyond prediction markets. Lara said that without current AI capabilities, the new product would either have dragged on the core business or required significantly more hiring. Her focus is not mainly on employing fewer people. It is on using AI to build more products and make the company larger once efficiency rises.
Hiring is shifting from task execution to relearning
Kalshi allows candidates to use AI in engineering interviews. Lara said the question that used to dominate two years ago — whether someone could complete a technical task — no longer separates candidates as clearly. The team now spends more time on system reviews and retrospective discussion of past projects. Design candidates are asked how much AI they have used, and legal interviews now include questions such as how a candidate would use AI to handle a case of that kind.
「The way you worked in the past is not going to be the way we work at Kalshi. We need people who are willing to re-figure it out: things they used to be very good at may no longer need to be done by hand,」she said.
What she values more are low ego, willingness to learn, and reliable execution. In her framing, hard work is not about counting hours. It is about doing well what has been entrusted to you. As tools keep changing, the willingness to let go of old advantages has become part of hiring judgment. That also explains why engineers are being pulled into design and legal processes. The company is not trying to change one isolated task. It is trying to change how each team handles tasks.
Lara also described the culture as direct. If colleagues dislike a piece of work, they say so immediately. She noted that Kalshi spent three to four years seeking regulatory approval, then sued its own regulator after two years of communication, and ultimately secured room to offer election markets. She reduced that experience to a simple rule: do everything that can be done, so that if you fail, you do not later discover there was still one more step you never took.
For individuals, the interview did not offer a one-line answer. A 73% probability is worth taking seriously, but it should still only be treated as 73%. The more useful move is to rewrite anxiety into questions that can be checked: which parts of your work have already been taken over by agents, which judgments still rely on experience, and whether your team’s hiring standard has changed recently. Probability will not decide the next step for anyone, but it can warn people not to use a sense of safety from two years ago to plan for today.