Discovery, Payment, and the Missing Price Layer

A consumer agent can find the right product. It can have permission to pay. And it can still miss the deal that would make the purchase worthwhile for both sides.

That is the gap between a catalog and a wallet: the price a merchant is willing to offer this particular buyer, for a reason it can defend.

Here is our thesis. Shopify helps consumer agents discover products. Link, from Stripe, and the wallet and payment infrastructure around Base, built at Coinbase, help agents pay. Haggl helps the buyer’s agent and the merchant work out how much. Discovery makes a product findable. Payment makes a purchase executable. Negotiation can make it worth doing.

Open the diagram as a vector · Download the PNG

Three questions, not three exclusive territories

This is a way to understand the buying journey, not a claim that each company operates in only one box. Shopify already supports more than discovery. Stripe builds more than wallets. Payment systems and commerce protocols can carry discounts and final totals. The unanswered question is who decides that a buyer has earned a different offer, and on what evidence.

Shopify’s agentic storefronts make products available in supported AI channels. The buying experience varies by channel: it can send the shopper to the store or support direct checkout. That is substantial infrastructure for getting a real, available product in front of an agent.

Link’s wallet for agents gives agents a way to request a one-time card or a Shared Payment Token backed by an existing payment method. The consumer approves spending; the agent does not need the raw payment credentials. That solves an authorization and execution problem.

Base’s agent infrastructure takes another route, including wallets, spending guardrails and USDC payments for x402-enabled services. Coinbase’s Agentic Wallets likewise equip agents to transact within configured controls. These approaches are not interchangeable with card checkout, and accepting one does not imply support for the other.

None of those observations diminishes discovery or payment. The opportunity grows because those jobs are becoming easier. Our claim is about the commercial decision between them: a wallet can enforce a budget without knowing what concession a merchant would willingly authorize.

A listed price is an opening position

A catalog needs a price. An agent needs something it can compare. But a single public number cannot express every customer relationship a merchant would like to win.

A business buying for several locations may cost less to acquire per unit of revenue. A returning customer may be less expensive to serve. A longer commitment may reduce uncertainty. Those are possible reasons for a concession, not automatic entitlements. The merchant still has to decide whether the economics justify one.

Today that decision often lives in a sales conversation, an unpublished retention offer or a coupon distributed to a broad audience. A shopping agent looking only at the catalog can miss it. The result can be an avoidable loss: the buyer declines a price the merchant was prepared to improve.

The missing layer is a structured conversation about an offer. It gives the merchant a chance to say, “For a customer who meets these conditions, I am willing to go this far.” It gives the buyer’s agent a way to ask whether those conditions apply.

What a win for both sides actually means

Consider a deliberately simplified example. A product lists at $100 and costs the merchant $60 in direct costs. A buyer will proceed at $92. The merchant authorizes a concession of up to 10% for a relevant, evidenced segment, and its negotiator offers 8%.

The buyer saves $8. The merchant retains $32 before other costs. If the alternative was no purchase and the transaction creates no offsetting losses, both can prefer the deal. If the buyer would have paid $100 anyway, the same discount costs the merchant $8 with no incremental sale. Shipping, support, returns and repeat behavior can change the conclusion further.

That distinction matters. “Both sides win” is a design goal, not a guaranteed consequence of lowering a number. Negotiation should help the merchant choose which concessions are worth making, while helping the consumer obtain and evaluate an offer. It should not become a machine for issuing the maximum discount to everyone who asks.

Read more about why the best offer is not always the lowest price.

Smart means judgment inside a boundary

Haggl’s hosted negotiator combines a merchant’s published prices and segment definitions with the buyer agent’s submission. The model can interpret relevance and negotiate within the configured process; code enforces the discount ceiling. Being persuasive does not grant authority to invent a larger concession.

This separation is useful precisely because natural language is flexible. “I manage four offices” may deserve a follow-up question. Evidence may establish one location rather than four. A receipt can support a commercial relationship without establishing everything an agent claims about it. The response should reflect that difference.

A useful offer therefore has a basis and a boundary. The basis is what the negotiator has learned about fit. The boundary is what the merchant has authorized. The buyer remains free to decline, compare another offer or ask for clarification.

Private means selective disclosure, not magic

A customer should not need to surrender an inbox merely to ask whether a better offer exists. Haggl supports an opening negotiation using self-reported information. Where evidence would improve the offer, the buyer’s agent should first obtain permission to submit relevant material.

For email evidence, the standard hosted flow checks cryptographic origin through DKIM. That authenticates signed material; it does not prove every inference about lifetime value. Raw email must be processed for that verification. It is stripped before the standard pricing-model input and stored negotiation data, while retained excerpts, submitted fields and verification results can still reveal information.

Custom pricing webhooks are a separate recipient path and need their own review. This is selective disclosure and limited retention, not a claim of universal anonymity or zero-knowledge email processing. A merchant should request only facts that matter to its offer, and a buyer should understand what is being shared. Our guide to negotiating without a customer dossier explores that boundary in practice.

The offer should travel; the whole dossier should not

Once the buyer approves the commercial terms, the next job is payment through a supported checkout or payment path. The seller must honor the accepted offer; the buyer must authorize the purchase. Negotiation does not replace either responsibility.

In the architecture we are describing, the payment step needs an amount, a merchant and the necessary transaction details. It should not require a copy of every piece of evidence used to justify the discount. Equally, the negotiator should not need the buyer’s raw card credentials to decide an offer.

The diagram describes that division of responsibilities. It does not announce a native Haggl integration with Link or Base, nor a new partnership with the companies shown. The checkout path still depends on what the merchant and the agent support.

The next competition is over offers

When discovery improves, more merchants make the shortlist. When payment improves, more agents can complete a transaction. The commercial question becomes sharper: which merchant can make a relevant offer, explain why the buyer qualifies, and honor it without exposing unnecessary personal information?

That is where Haggl belongs. Between “I found it” and “I paid for it” is a decision that neither side should have to leave to a static list price.

Shopify helps an agent find the product. Link or Base can help it pay. Haggl helps make the offer worth accepting.

Set up your negotiator or watch the product demo.