What we're seeing, Autumn 26. Updated quarterly.
Trend 2: A new front door
Alongside the website and the app, a player's personal agent is becoming a front door, and iGaming is no exception.
Our hypothesis
Alongside the website and the app, a player's personal agent is becoming a front door for all businesses, and this will be no different in iGaming. A person gives it a task and it acts for them. We expect it to start with research and comparison before it reaches transactions, as it has in shopping. In gambling, we think it will begin with discovery for entertainment: the best bets for today, or specific best prices. Offers exist for an individual behind login, so they'll come later, and will need operators to open a new front door for agents. The pattern is early, but the evidence base is moving at a rapid pace and could be seismic in how it shifts value within the industry.
What this means for you
Operators commonly design the website and app for a human. An agent needs different things: information it can read quickly, a way to identify itself, and clear rules on what it may do. The questions to settle first are who owns this inside the business, which agents you welcome, and what they may do. This is also going to be a major test of brand loyalty: how do you get a customer to not just ask their agent for the best bets for today, versus a specific closed question, "Show me the best bets on X operator today"? This change will quickly bring into focus whether customers buy from your brand or just because of your prices. Humans have loyalty. Will agents?
Prediction markets and traditional sports betting are becoming increasingly interesting side by side. If agents compare prices on a player's behalf, some businesses may decide to compete on price alone, and to step away from the regulatory requirements that owning a player's account brings. This raises a real question for sportsbook operators: could an operator retire its consumer-facing business, move onto prediction markets, and make its money from its trading skills like a hedge fund, rather than spending millions on brand marketing in a crowded space?
On casino it is less clear how this plays out, but think of it from the player's side. Would you rather ask your agent to tell you when a seat at your favourite live dealer table is ready and funded, and drop straight into it? Or open the app, go to the lobby, click the thumbnail, open the live dealer lobby and search for the table? The lobby is designed for browsing. An agent skips it.
What we are seeing that we think is important
- -Personal agents are gaining users fast. On Meta's Q3 call (21 Oct 2026), Zuckerberg said more than 10M people in the US and Canada have set up Muse accounts, and shopping is its most common task, at about $200 a purchase.
- -Agent traffic is rising and costing the businesses that serve it. Dan Rowe reports that peers across industries see agent traffic up, order volume flat and serving costs rising. His proposed fix is verifiable agent identity through Web Bot Auth.
- -Agents acting for bettors are being described. iGB (24 Sep 2026) sets out how a bettor could give an agent a budget, with the agent placing bets within limits where a platform allows. Automated betting already exists in racing syndicates and on exchange APIs. Polystrat, launched in Feb 2026, lets users fund an agent that trades on Polymarket. FanDuel's AceAI helps customers research and build bets, but only within the app when it is opened. It's an early pattern, but not agent to agent: a good example of human to agent.
- -The plumbing is being built. Marie Haynes describes Google's Universal Cart, Universal Commerce Protocol and WebMCP, which let agents compare inventory and check out inside Google's surfaces. She says WebMCP is still early.
- -Suppliers are opening up to agents. Fast Track introduced agent-to-agent support on 30 Sep.
- -The standards are still forming. Forkast reported that three agent specs promised for October were still unshipped, and that vendor extensions are fragmenting the stack.
- -Prediction markets are taking sports volume. Blask (23 Jul 2026) reports Kalshi at $31B notional volume in June 2026, about 85% of it from sports contracts, and Kalshi and Polymarket together holding 48.85% of its US demand index across 392 brands. FanDuel Predicts and DraftKings Predictions are both live, with roughly $600M of 2026 marketing committed to their own event contracts. InvestmentNews (15 Apr 2026) cites Bernstein forecasting $240B of prediction market volume in 2026 and $1 trillion a year by 2030.
What we're finding out next
- -Which player tasks agents will do first in gambling: research, price comparison, claiming offers, or placing bets.
- -How an agent identifies itself to an operator, and which agents an operator allows.
- -What an agent-readable offer, odds feed or terms page looks like.
- -Who owns this inside an operator, and what it costs to serve agent traffic.
- -Whether a sportsbook operator can run as a trading business on prediction markets, and what that does to the cost of brand marketing.
- -How casino play changes when an agent holds a seat, funds it and drops the player straight in.