How a negotiation tool could become a marketplace

One-sided automation is the starting point

An automated negotiation provider does not necessarily need to recruit every supplier onto a proprietary network before it can deliver value. Pactum currently says its digital workers can run in Coupa, SAP Ariba, through a buyer API or on their own. Its illustrated process takes requests through sourcing and negotiation to a purchase order while leaving the final award decision with the buyer’s team. Its separate explainer describes autonomous systems negotiating with a human counterparty through a conversational interface. These are vendor descriptions rather than independent outcome studies, but they establish an existence example for the product mechanism. ^{1 2}

The commercial consequence is important. Conventional marketplaces face a coordination problem: buyers want suppliers, suppliers want buyers, and each side can be reluctant to join an empty venue. A buyer-side negotiation system can avoid that initial barrier. One organisation adopts; suppliers respond through the channel presented to them; useful transaction history begins to accumulate.

Nothing in that arrangement proves marketplace formation or customer captivity. It shows why neither is required at the beginning. Platform economics can emerge from successful use rather than precede it.

How the tool could acquire marketplace power

The transition would probably be gradual. No single product release needs to announce that the negotiation tool has become a market intermediary. Five developments could move it in that direction.

1. Workflow integration. The system becomes embedded in procurement, enterprise resource planning, contract, payment and supplier-management workflows. Replacing it now requires operational change rather than a simple software substitution.
2. Longitudinal data. It builds a record of requirements, offers, concessions, supplier responses, accepted packages and realised outcomes. Raw events may be exportable while the useful derived models are not.
3. Repeated supplier contact. Suppliers meet the same intermediary across customers or repeated events. Profiles, verified identities, communication routes and performance histories become relationship infrastructure rather than buyer software alone.
4. Surrounding execution services. Qualification, contracting, payment, logistics, audit and dispute handling make the provider more valuable after an agreement is reached. The service begins to influence where transactions finish as well as how terms are discussed.
5. Counterparty discovery and routing. The system starts recommending suppliers or routing opportunities. It is now helping to decide who meets whom, on what evidence and through which commercial path—the central function of a marketplace, even if part of the supplier catalogue remains external.

This is a development path, not a forecast that every provider will follow it. Many systems may remain enterprise tools. Others may integrate into established procurement suites or marketplaces rather than become independent platforms. The point is that one-sided adoption gives the category a credible route towards intermediation without requiring a marketplace on day one.

Marketplace economics change the commercial bargain

Once a provider begins to mediate repeated transactions, the economics change. The OECD notes that where indirect network effects are present, demand for a platform’s services can depend not only on the magnitude of its prices but on how those prices are structured across market sides. A negotiation platform might charge the buyer, the supplier, the transaction, an execution service or some combination. A low apparent subscription price would no longer reveal the whole commercial model. ⁵

Indirect network effects may then appear. More buyer activity can attract more supplier participation or generate richer category evidence. Better supplier coverage can make the system more useful to buyers. More transactions can improve the provider’s operational knowledge, although the extent and transferability of any learning would need to be demonstrated rather than assumed.

The provider may also gain influence over ranking and route selection. It can decide which suppliers are discovered, which evidence counts, which offers are easy to progress and which execution services receive preference. Those choices may benefit the user, the provider or both. They are no longer neutral software settings once they affect access to commercial opportunity.

An open design can allow broad discovery and earn execution through verification, contracting, payment and recourse. A closed design can make outside discovery nominally possible while keeping the useful history, comparison and completion path inside.

Dependence is not the same as lock-in

A good system should become costly to replace in one legitimate sense: it has created value, people know how to use it, and workflows have improved around it. Calling every form of dependence lock-in would make the term useless.

The stronger warning applies when the cost of leaving comes from losing assets the customer reasonably believed it was building for itself, or from barriers unrelated to the provider’s continuing performance. Examples include an unusable export, supplier relationships that cannot continue elsewhere, inaccessible audit history, proprietary policy configuration, punitive exit fees or a degraded route for offers found outside the provider’s network.

Earned dependence means the customer can leave but prefers the provider’s continuing value. Engineered captivity means leaving destroys accumulated capability or market access beyond what is necessary to change systems. Real arrangements will sit between those poles, which is why the contract and architecture matter before switching becomes urgent.

The Negotiation System Portability Test

Use this test when procuring or materially expanding an automated negotiation system. Ask the provider to demonstrate the exit state after a representative period of operation. The instrument is Andrew’s Intel editorial synthesis, not a published industry standard.

On a narrow screen, each row is shown as a separate card.

A provider does not fail this test merely because migration would take work. The question is whether the customer can preserve the substance of its commercial memory, authority and relationships, then continue operating through another route.

Accio is an adjacent signal, not a lock-in case

Alibaba’s Accio Work makes one possible convergence visible because it comes from an established marketplace owner. Alibaba’s current product page says Accio Work can browse the web and operate multiple e-commerce backends; for sellers already on Alibaba.com, it describes the service as deeply integrated with that platform. Reuters reported the 23 March 2026 launch and quoted Alibaba International vice-president Kuo Zhang as saying that actions involving financial transactions, payment execution or access to private files require explicit, granular user permission. ^{3 4}

That evidence does not show that Accio has trapped users in a negotiation marketplace. It does not establish neutral supplier ranking, equivalent progression for outside suppliers, realised network effects or switching difficulty. Its relevance is narrower: a marketplace owner is combining boundary-crossing agents with privileged access to its own commercial infrastructure. That is one possible route from business tool to market intermediary, not proof that the destination has been reached.

The EU’s Digital Markets Act provides context rather than a direct rule for this category. For designated gatekeepers and the listed core platform service concerned, its duties include interoperability in specified situations, business-user access to certain data generated through use of the service, and freedom to promote offers and conclude contracts with acquired end users outside the service; it also prohibits self-preferencing in ranking. Which obligations apply depends on the service and designation. The regime shows why portability, outside contracting and ranking become material questions once a digital service acquires infrastructural power. It does not establish that today’s negotiation providers are gatekeepers. ⁶

The decision comes before the marketplace

A buyer evaluating an automated negotiation system today does not need to pretend that a closed marketplace already exists. It should recognise that the early contract can determine who owns the ingredients from which future platform power may be built.

The decision owner is therefore not only procurement technology. Commercial leadership, procurement, legal, data governance and operations should agree which records must remain portable, which supplier relationships belong to the buyer, how derived intelligence may be used, whether external options can be progressed and what happens to authority at termination.

This is not a case for weakening the product before it has proved useful. Deep integration and accumulated knowledge may be exactly why the system creates value. The safeguard is to make continued use a recurring commercial choice rather than the only way to retain the capability already built.

What the evidence does not establish

The reviewed evidence does not show a mature proprietary marketplace for automated commercial negotiation, widespread customer lock-in or a settled winning business model. Pactum describes buyer-side digital workers operating across procurement systems and suppliers. Alibaba describes a broader business-agent platform connected to its marketplace. Both are commercially interested sources.

The pathway from tool to marketplace is therefore Andrew’s Intel analysis. Platform economics explains why repeated participation, data and transaction services could create market power. Current products demonstrate some of the necessary components. They do not prove that those components will combine, that users will become captive or that an open model will outperform a closed one.

The sentence for the next procurement review

Show me what we can take with us after this system has learned our suppliers, terms and decision rules.

If usable history, relationships, authority and outside routes survive, the provider can earn our continued business. If the useful capability exists only while we remain, the switching cost is being designed now.

A useful negotiation tool does not need to begin as a marketplace to accumulate the ingredients of one.

Evidence note

Vendor material is used only to establish described product mechanisms. Pactum supports an existence example for one-sided automation. Alibaba and Reuters provide an adjacent signal of agents connected to established marketplace infrastructure, with Reuters reporting Alibaba’s permission statement rather than independently validating it. Official OECD and European Commission material supplies platform-economics and regulatory context. The five-stage development path, dependence distinction and portability test are Andrew’s Intel editorial synthesis.