Agentic purchasing in a B2B context
When talking about Agentic Commerce and AI agents in the context of purchasing, attention is easily drawn to visions of consumer buying: the agent does the grocery shopping, books a flight or finds the cheapest insurance.
In B2B purchasing, the impact of agency is ultimately an order of magnitude greater, because B2B purchasing is significantly more complex.
Why is B2B purchasing more complex?
An investment decision on a machine worth hundreds of thousands of euros or a service agreement worth tens of thousands of euros does not compare to buying a backpack. Large procurements involve several people, organisations and stages. The process can take months or years and still the end result may fall short, because one stage has been done poorly or in a hurry.
It is precisely this complexity where agents fit exceptionally well. Whereas a B2C agent makes one decision, B2B agent(s) handle a series of successive (and partly overlapping) stages. A significant share of these will happen in the future at the agent-to-agent interface, not between a human and a system.
This is how B2B procurement progresses
Programmatic B2B ordering is not a new idea. EDI standards (EDIFACT, cXML and others) have automated information exchange between companies since the 1970s: orders, invoices, delivery notices.
However, EDI's limitation is clear: it only works for recurring, pre-agreed transactions. Implementation usually requires bilateral integration, which has effectively shut smaller suppliers out. EDI starts when the decision has already been made. Everything before that, requirements definition, supplier selection, negotiations and risk analysis, has remained human work.
AI agents do not directly replace EDI, but complement it. Some of the following stages are already possible with today's tools. But especially agent-to-agent negotiation and machine-readable requests for quotation are only now developing or still future visions. The direction, however, is the same as with EDI in its time: standardised formats, expanding adoption.
Here, in brief, is the idea of a future agentic B2B purchasing process:
Needs and requirements definition:
The agent reviews previous purchases, user feedback and operational data and builds the requirements specification.
Supplier mapping:
The agent gathers information, compares and scores suppliers according to agreed criteria: references, certifications, delivery capability, financial stability.
Machine-readable request for quotation:
The agent submits a request for quotation in a machine-readable format, which the recipient's agent processes automatically. This requires established industry-specific formats, which are not yet widely in use.
Agent-to-agent negotiation:
The buyer's agent negotiates with the seller's agent on price, delivery terms, SLA levels and payment terms.
Due diligence:
The agent "checks what needs checking" in public registers, credit registers, carries out supply-chain risk mapping and compliance checks.
Contract management:
The agent proposes contract amendments and identifies unfavourable clauses.
Internal approval:
The agent calls in the people when the groundwork has been done and it is time to review and make decisions.
Continuous supplier monitoring:
After the contract is signed, SLA monitoring, invoice reconciliation and re-procurement are initiated automatically according to the contract lifecycle.
What does the future look like?
B2B purchasing or selling will not move away from people. Strategic decisions and significant investments require human judgement. But data collection, analysis, documentation, coordination and monitoring are largely automatable.
Preparation for the agentic future of B2B procurement starts now. The good news is that, ultimately, preparation is about data and distributing it sensibly to both programmatic and human customers. For this
How can Crasman help?
At Crasman, we build solutions for our customers that support agentic buying and selling — from both the buyer's and the seller's perspective. We build tools for the agentic future with the market's best Google Cloud tools.
To make getting started as easy as possible, we have put together an Agentic Readiness audit, in which we assess your company's readiness to serve autonomous buyers.
The audit covers your potential online store and the entire digital sales funnel:
Product data quality: Is the technical data sufficiently accurate and structured for agents' decision-making?
Content clarity: Are the products, use cases and differences clearly interpretable, including programmatically?
Findability: How easily can your offering be found through searches, APIs and agents?
Transaction readiness: Do your interfaces enable an agent handling B2B purchasing to carry out tasks programmatically?
Integration capability: How smoothly do your systems communicate with purchasing agents?
Trust & evidence: Is enough information available (prices, availability, delivery, references) to support the purchasing decision?
The price of the audit depends on the scope of the digital service. Starting from €1,250 (+ VAT).
If your procurement process, sales funnel or B2B e-commerce need development, get in touch!
