Most of the AI-in-finance story is told through models. The plumbing underneath is changing too. Markets are being redesigned to run almost around the clock, and the contracts traded on them are becoming more granular. The first software to use those systems at machine speed will be AI agents. The dossier behind this piece is thin on the models themselves. It is much stronger on the rails agents will run on, so that is where this piece spends its time.
The contracts: from index bets to company KPIs
On September 30, 2026, at Robinhood's third annual HOOD Summit in Houston, Cboe Global Markets announced plans to launch a new category of binary contracts tied to company-specific key performance indicators. The launch was expected in October 2026 and is subject to regulatory approval. Robinhood would be the first retail broker to offer them.1 Cboe's Steve Quirk framed the product this way: “Earnings contracts give retail investors another tool to inform their strategies and offer an even more precise way to trade on anticipated company KPIs.”1
For a technology reader, the point is that these contracts resolve on discrete, machine-readable facts. That makes them natural inputs and outputs for automated systems. We would be inferring that link. The dossier does not say agents will trade them.
Robinhood's latest quarter shows how fast this line is growing. Its Q3 results, reported October 27, showed event-contract revenue rising more than sixfold to $173 million. Crypto revenue fell 55% year on year to $120 million. Total revenue rose 12% to a record $1.42 billion. Net income fell 14% to $476 million, and diluted earnings per share fell 16% to $0.51.2 Prediction contracts are now large enough to show up in an income statement, and they are growing while the older crypto line shrinks.
The clock: 23 hours, and a regulator asking whether 24 is possible
Nasdaq says the industry's move to a 23-hour trading day is expected on Sunday, December 6, 2026. That depends on SIP readiness and SEC rule changes. Subject to SEC approval, Nasdaq will add a 9 p.m. to 4 a.m. ET session.3 (The SIP is the shared data feed that publishes market prices.) Separately, the CFTC announced an industry roundtable on December 15 on whether derivatives markets can move to 24/7 trading.4
Both dates are scheduled events, not completed changes, and both depend on approvals. Still, a market that never closes is one where human oversight is the bottleneck. That is the setting in which software agents become attractive.
The fight over who writes the rules
The rules are not settled. On October 27 the CFTC sued Utah to block state efforts to shut down CFTC-registered contract markets under state law. It said it has started legal proceedings against nine other states.5 Event contracts are growing inside a jurisdictional dispute that has reached the courts.
The AI layer: who is this agent?
The clearest AI-specific development in our material is Baselayer's. The San Francisco startup helps financial institutions verify businesses and assess fraud risk. It has raised $35 million to extend its identity technology to AI agents. M13 led the Series A, with Picus Capital participating.6
Founder Jonathan Awad states the problem plainly: “Agents spin up and they spin down. How can you trust this random one-task agent?”6 His warning about the alternative is blunt. Unless agents can establish that they act on behalf of legitimate people or businesses, “agents will just get blocked everywhere.”6
Awad also describes the commercial friction. Relationships with financial institutions typically take 12 to 18 months to establish, and large merchant partnerships can take up to 24 months.6 Identity infrastructure for agents therefore cannot be shipped like a software update. It depends on slow-moving institutions agreeing to trust it. Awad also said of Baselayer's own sales motion: “We've essentially streamlined 10 years' worth of selling into two years.”6 That is a company's description of its own progress, not an independent measurement.
The investor's caution: AI is not a calculator
Norwest partner Sean Jacobsohn offers a useful counterweight for anyone excited about agents in finance: “You don't want AI doing calculations because it is not good at math. There are certain workflows it can handle where it doesn't produce precise numbers. But when you need precision, accuracy and calculations, you can't rely on AI for that.”7
This matters for the products above. Contracts that settle on a KPI, and systems that move money, depend on exact numbers. Jacobsohn also says that incumbents such as Workday, ADP, SAP, UKG and Dayforce are hard to disrupt at the core. In his view the opening is in their secondary products, where “the category isn't their core business.”7 He sees a potential opportunity against NetSuite and Sage, with disruption happening “downmarket” first.7 The AI-native finance stack, on his account, advances at the edges first.
The one hard dataset: Robinhood's SEC-checked numbers
Robinhood is the one company in this story whose numbers we can check against SEC filings. Its revenue grew from $1.865 billion in fiscal 2023 to $2.951 billion in 2024 and $4.473 billion in 2025.8 That is about 2.4 times the 2023 level in two years. Net income swung from a loss of $541 million in 2023 to profits of $1.411 billion in 2024 and $1.883 billion in 2025.8 Earnings per share went from -$0.61 to $1.56 to $2.05.8
Research and development spending barely moved: $349 million, $323 million and $355 million over the same three years.8 Revenue more than doubled while the engineering budget stayed roughly flat. We would not read that as proof of anything about AI. The filings do not say where the money went. But it shows a platform whose growth is not coming from a rising technology bill.
Cash held steady rather than surging. It was $4.835 billion at the end of 2023, $4.332 billion in 2024 and $4.261 billion in 2025. Operating cash flow was $1.181 billion in 2023, -$157 million in 2024 and $1.638 billion in 2025.8 Quarterly cash ran from $5.829 billion in Q2 2023 to $5.362 billion in Q2 2026.8 One seam is worth noting. Two records for Robinhood's cash at September 30, 2025 disagree: $4.331 billion in one and $4.3 billion in the other, a difference of roughly $31 million. We use the SEC-sourced figure, $4.331 billion, and treat the rounder number as a rounding of it.8
How much to trust the sources
Our measured reliability for the sources here is modest. For the NewsEOD wire carrying the Cboe release, 53% of 6,464 checked claims held up. For Crunchbase News, which carries the Baselayer and Norwest material, 49% of 41 checked claims held up.1,6,7 The Crunchbase sample is small, so treat that figure as indicative. The quotes above come directly from the published interviews. The surrounding claims, such as plans and funding, rest on the companies' own accounts. The Robinhood financials are the only figures here checked against a source of truth.
What to watch
- Whether Cboe's KPI contracts received regulatory approval and actually launched. The October 2026 timing was conditional.1
- The December 6 Nasdaq session and the SEC rule changes it depends on.3
- The CFTC's December 15 roundtable on 24/7 derivatives.4
- The CFTC's litigation with Utah and the nine other states.5
- Whether agent-identity firms like Baselayer turn funding into the institutional relationships Awad says take 12 to 24 months.6
Our dossier contains no evidence that AI agents are trading these contracts today. The case for agent-driven markets is so far a design argument. It is not an observed fact.