Artificial intelligence
The agentic buying path
Publiée le December 3, 2025
Publiée le December 3, 2025
Agent assistants are becoming increasingly commonplace in everyday life. An “agent” can take care of booking a table, ordering a dish, purchasing an item or subscribing to a service using digital interfaces. To use these tools with confidence, it’s important to know the steps involved, understand the regulations and know what rights protect consumers. The following sections present in-depth articles (around 2,000 words each) on six types of agentic buying journey: booking (reserving a table at a restaurant), service (on-demand transport), order (ordering a meal), commerce (buying goods online), subscription (taking out car insurance) and appointment (making an appointment with a garage mechanic). Each article describes concrete steps, best practices and consumer rights.
The Agentic Booking path automates the search for and reservation of a restaurant. The user expresses a simple intention: “Book me a table for two in an Italian restaurant tonight around 8pm”.
The agent then performs all the steps traditionally carried out by hand: selecting establishments, analyzing availability, checking reviews, taking dietary requirements into account, choosing a budget, comparing booking options.
This type of agent also manages unforeseen events: changes to the number of guests, late arrivals, cancellations, payment of a deposit if necessary.
It must also integrate rules: cancellation conditions, no-shows policy, PMR accessibility, variations linked to platforms (TheFork, Google Reserve, OpenTable).
This agentic path is particularly promising in the restaurant sector, where real-time availability, contextual preferences (proximity, weather, ambiance) and personalization create a real advantage.
Here, the agent takes charge of the entire transport booking process (Uber, cab, VTC, shuttle). All the user has to do is say, “Can you book me an Uber to Gare de Lyon in 10 minutes?
The agent automatically analyzes geolocation, journey times, estimated fares, options (XL, sedan, eco), traffic conditions, alternatives in the event of delays, and even the carbon footprint.
The agent can also manage unforeseen events: change of address, driver delays, traffic jams, driver cancellations or platform malfunctions.
The value of this course lies in :
reliability (intelligent planning),
pricing transparency,
time optimization (anticipatory care),
safety (route analysis, license plate verification, automatic sharing).
Eventually, these agents will be able to orchestrate multi-modal trips (train + Uber + scooter) to reduce total cost and travel time.
The Agentic Order route corresponds to the automated ordering of a meal via Deliveroo, Uber Eats, Just Eat or local platforms.
The user gives an intention: “Order me a poké salmon from the highest-rated restaurant and deliver it at 12:45.”
The agent then manages :
food preferences (gluten-free, vegetarian, allergies),
order history,
estimated delivery times,
assessment of restaurateurs,
costs (delivery, service, small basket),
loyalty programs and discounts.
One of the major challenges is managing the unexpected, such as out-of-stock situations, restaurant closures, extended lead times, ingredient substitutions.
The agent must act proactively: proposing an alternative, renegotiating delivery or adjusting the order.
This pathway has been massively adopted because it responds to a simple intention: to eat quickly, well and without friction.
The Agentic Commerce path represents the future of e-commerce: the user expresses a need, and the agent finds, compares, buys and tracks the product concerned.
For example: “Can you buy me a pair of Nike Air Max 270 size 42 at the best price and delivered before Thursday?”
The agent takes charge of :
multi-platform search (Amazon, Zalando, Veepee, official websites),
price comparison, delivery times, free returns,
checking for counterfeits,
notice consultation,
the legal warranty of conformity,
managing returns in the event of a problem.
In some cases, the agent takes into account biometrics (foot scan) or product compatibility.
It also anticipates purchasing preferences (new, reconditioned, second-hand), environmental impact and seasonal promotions.
This is the start of a revolution in online purchasing, with consumers delegating the entire purchasing process, including after-sales service.
Here, the agent doesn’t buy a one-off product, but subscribes to a recurring service on behalf of the user: car insurance, energy, internet, SVOD subscription, food box.
Example: “Get me car insurance at the best price for a 2022 electric vehicle.”
Agent:
compares offers (price, guarantees, exclusions),
analyzes deductibles and ceilings,
takes into account driving history and bonus-malus,
checks legal compliance (compulsory civil liability),
detects ecological promotions or bonuses,
monitor the contract (renewal, termination, change of vehicle).
This pathway has enormous strategic value: it enables us to considerably reduce the household’s fixed costs by optimizing each subscription.
Subscription agents will become the new “automated family budget managers”.
Here, the agent manages the scheduling of a car maintenance appointment.
Example: “Book me a slot at a garage to change my brake pads before Saturday.”
Agent:
compares available garages,
verifies certifications,
consult prices (packages, quotes),
reads the notices,
estimates how long the vehicle will be out of service,
reserves a time slot compatible with the user’s schedule,
prepares documents (estimates, repair orders),
manages unforeseen or additional work.
This type of agent must manage a crucial element: legal compliance, particularly if the garage carries out unauthorized work, if the user cancels, or if the invoice diverges from the initial estimate.
This saves a lot of time, and makes often complex processes more reliable.
These six routes demonstrate the power of the agentic model:
reduces friction,
eliminates repetitive tasks,
reassures the consumer,
personalizes the experience,
optimizes total cost.
In the years to come, every purchase intention could be carried out automatically by a reliable, contextual, multimodal and interoperable agent.