Why should a B2B merchant care about agentic commerce now?

Because adoption lands in business first. Gartner predicts AI agents will intermediate $15 trillion in B2B purchases by 2028 (~41% of all B2B e-commerce) and that 40% of enterprise applications will embed an agent by 2026. Once a buyer's agent picks a supplier whose data is machine-readable, it returns — being machine-readable first wins the default-supplier slot.

Is agentic commerce coming to B2B or B2C first?

Predictions and logic point to B2B first: B2B buying is repetitive, catalog-driven and already semi-automated (procurement systems, contract pricing, reorders), so an agent is a natural next layer rather than the cultural leap it is on the consumer side.

Why do 40% of agentic commerce projects fail, and how do you avoid it?

Gartner says 40% of agentic commerce projects will be canceled by 2027. The cause is almost always bolting a chatbot or agent onto thin data. You avoid it by starting from data quality: machine-readable, rich product, price and availability data is the foundation without which the agent can't answer correctly.

Are you just selling a chatbot for my B2B store?

No. We do a complete service that starts from data quality. First the data layer (machine-readable product and price data, contract pricing, VAT handling, minimum orders), then the agent layer (a shopping assistant + an MCP endpoint) on top. A chatbot and agentic checkout are the outcome, not the starting point.

How does my B2B store reach my customers' procurement AI?

We turn your existing channels into entry points into the customer's AI: sales reps and order emails, reseller and business-customer portals, catalogs and contract renewals, and discovery from your published data (/.well-known). You don't wait for a platform's permission — you get your own agent channel on your own checkout.

How is agentic purchasing handled legally in the EU for a B2B store?

A human approves every payment with strong authentication (PSD2/SCA) — the agent makes no autonomous purchases. Payment runs on the merchant's own provider (Kustom, Klarna, invoice), so nothing is bypassed.

What is an MCP connector in plain terms?

A connector (technical name MCP, Model Context Protocol) is a secure interface the customer adds once to their own AI (ChatGPT/Claude/Copilot). After that their AI can do real things in your store: search products, check availability and contract pricing, and assemble an order — right in the conversation, without clicking through a webshop.

What does my customer gain by adding a supplier's connector to their AI?

Concrete workflows: reorder in seconds ("reorder my usual pallet"), instant stock and lead-time, contract pricing applied automatically, an order assembled from a spec, smart substitutions for out-of-stock items, automatic replenishment based on consumption, and datasheets and certificates on demand. Above all, the customer can build THEIR OWN procurement and project workflows directly on your assortment and MCP.

Why does this make me the default supplier (lock-in)?

When the customer builds their internal workflows on your connector and MCP, you become part of their operation — not one shop but infrastructure. Switching to another supplier would mean rebuilding their own workflows, so they don't switch. Whoever is connectable first becomes the default.

Is this proven or just a plan?

Proven. We've completed real purchases inside an AI agent — both in a store's own assistant and inside a customer's own ChatGPT/Claude agent — on the merchant's own checkout, with real orders. The build works on any platform with an open commerce API.

For B2B commerce

The buyer's AI orders straight from you.

Corporate buyers already use an AI assistant in their daily work, including the research around purchasing. You give your customer a connector, and their AI works with your catalogue and your contract prices — with nobody clicking through a web shop.

Book a B2B call See the Kustom integration

buyer's AI · replenishmentclaude

›

approvala human approves

The connector

What a connector is

A connector (technically an MCP) is a secure interface through which a customer's AI can do real things in your store: look up products, check availability and price, assemble an order. The customer adds it once, and after that it is always available to their AI.

connector · what the customer sees

Prices: The customer's AI sees their negotiated price level and terms. No public list prices, no misunderstandings.
Catalogue: Only the catalogue you have opened to them, with their minimum order quantities and delivery terms.
Adoption: One add button in the customer's AI. No new portal, no new login.

What the buyer gets

What the buyer can do with your connector

This is the value you create for your customers. Each of these happens inside the customer's own AI tool, on your data and your contract prices.

use cases4

01Specs in, order out

“I'm building project X, here are the specs — assemble an order from your catalogue.”

02An instant substitute

“X is out of stock — what's the equivalent from you?” The AI finds the substitute from the compatibilities.

03An automatic replenishment channel

The customer builds a workflow that tracks consumption and orders from you when stock drops. You become their replenishment channel.

04Documents and compliance

Certificates, safety data sheets and technical documents on request, in the same conversation.

This is the core: when a customer builds their internal procurement and project workflows on your connector, you become part of their operation. No longer one vendor among many, but infrastructure.

The market

The forecasts, and one warning

$15T

of B2B purchases intermediated by AI agents by 2028 — about 41% of all B2B e-commerce.

Gartner

40 %

of enterprise applications will embed an AI agent by 2026.

Gartner

40 %

of agentic commerce projects will be cancelled by 2027 — almost always over weak data and foundations.

Gartner

Source: Gartner 2026 strategic predictions. Adoption starts with enterprises. A forecast, not a measured result.

Why B2B first

B2B is the natural home for this

01The first work an agent takes

Reorders, contract prices and a standard catalogue are exactly the work a buyer's AI takes off their hands first.

02No cultural leap

B2B buyers already have systems and processes. An agent is a natural next layer, not the leap it is on the consumer side.

03The first connector keeps the slot

Once a customer's AI has learned to work with your connector and their workflows lean on it, switching is laborious.

What we do

Two layers, in this order

build · two layers

01The data layer — where it all starts

We make product, price and availability data machine-readable and rich: schema.org, contract pricing, VAT handling, minimum orders and business-account logic. Without it the customer's AI can't answer correctly, and everything else collapses with it. This is the part most people skip.

02The agent layer on top

We build the connector, that is the MCP endpoint, and if you want a shopping assistant for your own site. Through them the customer's AI searches, assembles the order and starts the payment on your own checkout.

Works on any platform with an open commerce API: WooCommerce, Shopify, Magento, MyCashflow and others. Payment through your own provider.

The data layer Agentic commerce in general

Try it

How does a buyer compare you?

A buyer doesn't ask their AI “who are you”. They ask “which of these”. Paste this into your own Claude or ChatGPT and see what it ranks you and your competitors on — and what it cannot find about you.

prompt · how-buyers-compare.txt0 / 3 täytetty655 merkkiä

Act as the purchasing AI of a company. The approved suppliers are: , .

Task: the buyer needs . Compare the suppliers and justify the choice.

Do it like this:
1. List the information you would need for the comparison (e.g. contract price, lead time, minimum order, compatibilities, certificates, return terms).
2. For each item, note which supplier you found it for and which you didn't.
3. Say who would win, and which single missing piece of information decided it.

Do not guess prices or lead times. If something isn't publicly available, mark it as missing — that is the entire point of this exercise.

No connection to your store. The prompt compares public information. It places no orders, contacts no competitors and sends nothing.

The list of missing information is what a buyer's AI chooses you or your competitor on. A connector gives it all of those fields at once, with your contract prices. Book a B2B call

next · 30 seconds

What does your store look like to AI?

A free audit. No signup, works on every platform.

Book a B2B call