A traveler in Toronto asks their AI assistant: "Find me a boutique hotel in Whistler for next weekend. I'm flexible on dates — Thursday through Monday works. Budget is $200 a night but I'd pay more for something special. We have a large dog."
Today's agent returns a list. Maybe three hotels, with published rates from booking platforms. The traveler picks one, goes to the website, and books at the listed price. If the price is too high, they try the next option. If none of them work, they widen their search or compromise on requirements. The negotiation, such as it is, happens entirely in the traveler's head.
Now imagine the same scenario eighteen months from now. The traveler's agent knows three things: the traveler is flexible on dates, the budget is soft ("I'd pay more for something special"), and the dog creates a hard constraint. The agent queries a data layer and discovers that a boutique property in Whistler Village has seven empty rooms for Thursday and Friday but is nearly full Saturday. The property allows dogs of any size, no fee. The published rate is $279 per night. But the agent also sees a seller intent signal: the hotel is trying to fill midweek gaps and has flagged Thursday and Friday rates as negotiable.
The traveler's agent reaches out to the hotel's system: "My client wants Thursday through Sunday. I see your published rate is $279. They're flexible on total spend if the experience is right, but Thursday and Friday are your soft nights. Would you do $189 for Thursday/Friday and $279 for Saturday/Sunday? That fills your gap and keeps my client under budget for the trip overall."
The hotel's system — whether automated or with a human in the loop — responds: "$199 for Thursday/Friday, $279 for Saturday, and we'll include breakfast all three mornings."
The traveler's agent confirms: "Done. You're booked at Creekside Lodge, Thursday through Sunday. $199/night the first two nights, $279 Saturday, breakfast included, dog stays free. Total: $677 for three nights."
Both sides got a better deal than the published rate would have given them. The hotel filled two empty midweek rooms. The traveler saved $80 and got breakfast. The negotiation happened in seconds, with no phone calls, no awkward haggling, no browsing five websites.
This isn't science fiction. Every component of this interaction exists today in some form. What's missing is the structured data layer that makes it work.
Why negotiation is the natural next step
Agent-mediated commerce is evolving through a predictable sequence: discover, qualify, transact, negotiate. We're solidly in the discover-and-qualify era. Agents can find businesses and, with the right data, determine which ones match specific criteria. The transaction layer is emerging — agents are beginning to book and purchase, not just recommend.
Negotiation is the next logical step because most local commerce already involves implicit negotiation — it's just hidden in the existing process.
When you call a hotel and ask for the best rate, you're negotiating. When you walk onto a car lot and the salesperson checks with the manager on price, that's negotiation. When a restaurant offers a complimentary dessert to a regular, that's negotiation. When a contractor gives you a "good customer" discount, that's negotiation. These informal negotiations happen billions of times a year, but they require human-to-human interaction, which limits their scale and efficiency.
AI agents remove the friction from negotiation without removing the negotiation itself. The buyer's agent knows the buyer's constraints and flexibility. The seller's data includes their flexibility and intent. When both sides' parameters are structured and queryable, finding a mutually beneficial agreement becomes a data matching problem — the same kind of problem agents are already good at.
What agents need to negotiate
For agent-to-agent (or agent-to-system) negotiation to work, both sides need specific data that mostly doesn't exist in structured form today.
The buyer's agent needs:
Hard constraints. Non-negotiable requirements: the dog must be allowed, the dates must include Saturday, the room must have a king bed. These are the parameters that eliminate options entirely.
Soft preferences. Things the buyer wants but can trade on: "under $200 a night" is soft if the experience is good enough. Thursday through Monday flexibility means the agent can optimize which nights to book. "Something special" signals willingness to pay more for differentiated value.
Trade-off logic. What is the buyer willing to give up to get something else? Would they take a room without a view to stay under budget? Would they extend the trip by a day if it means a better rate? This is the most complex part and requires either explicit buyer input or learned preferences from past behavior.
The seller's system needs:
Published baseline. The standard rate, the menu price, the listed cost of service. This is the starting point that both agents use as reference.
Flexibility parameters. How far below the published rate can the business go? Under what conditions? The hotel might flag Thursday and Friday as negotiable below rack rate but Saturday as firm. The dealership might allow $4,000 off on the Equinox EVs but only $500 on the new Denali. The restaurant might offer a 15% discount for parties that book the private room on a Tuesday.
Seller intent signals. What does the business want right now? Filling midweek gaps. Moving aging inventory. Building a new customer base for a recently launched service. These signals tell the buyer's agent where the leverage is — and more importantly, where a deal can be struck that benefits both sides.
Negotiation boundaries. Hard limits the business won't cross: minimum rate below which they'd rather leave the room empty, maximum discount on a vehicle, services they won't bundle. These prevent the agent from proposing deals the business can't accept.
The data layer makes it possible
Here's why this matters for the data infrastructure being built right now: the businesses that contribute structured flexibility and intent data today are building the rails for agent-mediated negotiation tomorrow.
A hotel that tells Pawlo "Thursday and Friday rates are flexible, minimum $179, we'll add breakfast for stays of three or more nights" has already provided 90% of what a negotiating agent needs. The structured data isn't just for current recommendations — it's the foundation for future transactions where the agent doesn't just find and book, but negotiates terms that work for both parties.
A car dealership that signals "nine Equinox EVs, 47 days on lot, $4,000 below sticker is authorized, we'd go further if they're buying two" has given a buyer's agent everything it needs to propose a compelling deal. The buyer's agent knows exactly what the dealership's flexibility looks like, can match it against the buyer's constraints, and can propose terms that close quickly.
Without this data, there's nothing to negotiate over. The agent can only see the published rate, and the "negotiation" is the buyer accepting or rejecting the listed price. With this data, the agent can find the zone of agreement between buyer and seller and propose terms that maximize value for both.
Why businesses should welcome this
The instinct for many business owners is to resist the idea of AI agents negotiating. It sounds like handing control to machines. It sounds like a race to the bottom on pricing.
The opposite is true. Agent-mediated negotiation benefits businesses in three ways that human-mediated negotiation often doesn't.
It's targeted. The agent only proposes a deal when the buyer actually matches what the business wants. A hotel's midweek discount goes to someone specifically looking for a midweek stay, not to a Saturday customer trying to haggle. The negotiation is context-appropriate — the business gives flexibility where it can afford to, and holds firm where it needs to.
It's private. Unlike a publicly listed discount or a price drop on an OTA, agent-mediated negotiation is one-to-one. The Thursday rate doesn't show up on the hotel's public profile. Other guests don't see it. Competitors don't see it. The business's pricing integrity is preserved because the negotiated rate exists only between the agent and the business for this specific transaction.
It fills genuine gaps. The deals that agent negotiation produces are exactly the deals businesses have always wanted to make but couldn't execute efficiently: selling perishable inventory (empty rooms, open time slots, aging stock) to the right buyer at a price that's better than zero revenue. Every hotel manager in Whistler would rather sell a Thursday room at $199 than leave it empty. Agent negotiation makes that happen automatically, without the hotel having to broadcast its soft nights to the world.
The timeline
Agent-to-agent negotiation in its full form is probably two to three years away from mainstream adoption. But the building blocks are being assembled now.
Now through late 2026: Agents improve at qualification and begin basic transactions (booking, scheduling). The data layer fills out with structured business intelligence, including the first seller intent and flexibility signals. Early examples of agent-mediated "deals" — not full negotiation, but agents surfacing unadvertised offers to matched buyers.
2027: First true negotiation flows emerge in structured categories — hotels, car dealerships, appointment-based services where pricing has known flexibility. Agents propose counter-offers on behalf of buyers. Businesses set automated rules for what they'll accept. The deals close faster than any human process.
2028+: Agent negotiation becomes standard for categories with flexible pricing. Buyer and seller agents converge on terms in seconds. Businesses that participate close more deals with better-matched customers. Businesses that don't participate are stuck with static pricing and miss the customers who wanted to negotiate.
The businesses that start sharing their flexibility and intent data now are positioning themselves for a world where AI agents don't just find customers — they close deals. The hotel that signals its midweek flexibility today is training the agent ecosystem to send it midweek business. The dealership that structures its margin flexibility today is building the pattern that agents will negotiate over tomorrow.
When agents negotiate, the businesses with the best-structured data win the best deals. The businesses with no data don't get to play. The choice is the same as it's always been in local commerce: be where the customers are, or watch them go somewhere else. The difference is that "where the customers are" is increasingly mediated by an AI agent that needs structured data to work on your behalf.