Agentic commerce connects the transaction to its cost stack
Ask Maps' ordering and hotel tools and Omilia's new funding show that AI-enabled commerce depends as much on transaction partners, lean models, and sales infrastructure as on the user interface.
Artificial Intelligence··Evening
The map becomes a transaction gateway
Google is adding food ordering, hotel booking, and event-ticket discovery to Ask Maps. When a user asks for a particular dish nearby, the tool finds matching restaurants; an “Order online” option adds items to a cart through supported platforms such as Square, Toast, or Uber Eats. Payment still takes place on the supported platform. For hotels, Ask Maps compares prices and checks availability, but the reservation is completed on the partner's website. Google Maps is therefore shortening the route between discovery and purchase intent while leaving the final transaction with existing commerce partners. The agentic features rolling out in the United States show that a conversational interface is not a store by itself: finding results, comparing choices, preparing a cart, and taking payment remain distributed across several systems. Ask Maps can also remember previous conversations, allowing a user to resume an unfinished plan. Personal Intelligence arrives switched off by default and can personalise responses using Gmail and Google Calendar.[1]
Not every step needs the largest model
Omilia's growth story shows the other end of this distributed transaction chain. The Athens-based company, which has automated voice calls and customer support since 2002, raised 67 million dollars in a Series B round led by Expedition Growth Capital. Omilia says its annual recurring revenue has increased tenfold to 60 million dollars since it took 20 million dollars from Grafton Capital in 2020. Chief executive Dimitris Vassos says a large share of customer queries concern basic information such as account balances, so there is little need to run a large language model for every request. The company presents the resulting unit economics for itself and its customers as a distinguishing feature. In other words, the commercial value of conversational artificial intelligence does not come only from a more capable model. Routing a simple request to a suitable lean process, invoking more advanced capability when needed, and connecting the answer to an existing business flow are also parts of the product. Ask Maps completing an order on a partner platform and Omilia avoiding a large model for every question expose two layers of the same arrangement: the interface captures intent, while the back end allocates the task to the appropriate service and cost level.[2], [1]
Capital moves into sales and connection layers
Omilia plans to use the new money to open an office in the United States, expand its go-to-market team, and hire new leaders for revenue and marketing. The company has about 500 employees and expects to reach 600 by the end of the year. Its customers include Capital One, Discover, RBC, DWP, and PSEG; its technology runs in more than 1,000 Taco Bell locations, and it is talking with two more quick-service chains. That allocation shows that growth funding is not reserved for model development. The teams that connect a product to customer systems, the sales channels that place it, and the capacity to deploy it are scaling too. On the Ask Maps side, similar connection work appears through Square, Toast, Uber Eats, and hotel partners. Google keeps user intent and comparison inside its interface, while payment or reservation concludes in another company's system. Together, the two developments describe an agentic-commerce stack that extends beyond a single model provider. Value is created by carrying a request that begins in natural language into the right catalogue, availability record, cart, payment, or support flow; cost discipline and partner connections help determine whether that transfer can be sustained.[2], [1]
