Building the Data Shopfront for AI

My last article ended with a question. If AI agents are going to pay to access data, what needs to sit behind the transaction to make it work?

One way of thinking about that framework is as the digital equivalent of a reliable shopfront. When people buy something online, they expect to be able to identify the seller, understand what is being offered and review the price and terms. If something is unclear, they can ask questions or obtain advice. An AI agent needs much the same information, but it must be presented in a form that the agent can interpret reliably.

The agent needs to identify the dataset and the party offering it. It must be able to establish whether that party has authority to grant the licence and what the licence allows. It may also need information about the source and quality of the data before deciding whether it is suitable for the intended use.

Preparing data for commercial use

Companies holding substantial volumes of data are already considering how that information might be commercialised. Before an important dataset can be licensed at scale, it will usually need a considerable amount of preparation. The company must establish what information it holds and how it is organised. It may need to improve the quality of the data and the metadata that describes it. The company will also have to understand the rights and restrictions that apply before deciding how the data may be offered to others.

This is where the idea of a Data Asset Foundation begins to become relevant. A Data Asset Foundation is a separate Isle of Man legal entity established to hold and govern a defined data asset. It does not have shareholders and is governed by a council in accordance with its stated objects and rules. This allows the foundation to be designed around the proper management and commercial use of the data rather than the wider interests of an operating business.

The data does not necessarily have to be moved from the systems in which it is already held. It may continue to be stored and maintained by the operating company, while the data and the relevant rights or permissions are dedicated to the foundation. The foundation can then provide the legal and governance structure through which the resulting asset is registered and licensed.

Any company planning to commercialise a substantial dataset will need to undertake much of the preparatory work in any event. By incorporating a Data Asset Foundation into the project at an early stage, the company can develop the data and the commercial structure around it together.

The operating company may continue to manage the information, but within a framework intended to maintain the quality and integrity of the resulting asset. That framework will include the requirements of the Data Governance Framework, together with any additional standards adopted by the foundation. The company is therefore not simply producing a cleaner database. It is preparing a defined asset with repeated and wider commercial use in mind.

Why should the customer trust it?

Preparing the data does not remove another obvious problem. The company offering the licence is still asking the customer to accept what it says about its own product. It may say that the dataset is accurate, that it was lawfully assembled and that it has sufficient authority to grant the proposed rights. The customer must decide how much reliance to place on those statements.

For a significant transaction, that may require extensive due diligence. The customer may need to examine the source of the information and review the agreements through which it was obtained. It may also have to assess the quality of the data and determine whether it is sufficiently current for the proposed use. That exercise can be expensive and time-consuming. Much of it may then be repeated whenever another customer considers licensing the same asset.

An autonomous AI agent is not well placed to resolve uncertainty by relying on commercial instinct or informal assurances. It needs dependable information that allows it to establish what the asset is, who can license it and whether it is suitable for the transaction being considered.

From assertion to verification

This is where independent assurance and the Data Asset Register can make a difference. Rather than asking customers to rely entirely on the company's description of its own data, the asset would be assessed by an Accredited Assurance Provider against the standards set by the Data Governance Framework.

The Register would then record prescribed information about the asset, including its identity and accreditation status. It would not hold the underlying data or guarantee that every piece of information was accurate. Its purpose would be to provide an authoritative record of the asset and to show that it had passed through an established process before being fully registered.

That should give a customer a better starting point. Instead of having to establish the position entirely from the seller's contracts, policies and technical records, the customer could begin with a registered asset supported by independent assurance. Further enquiries may still be necessary, particularly for an important transaction, but much of the groundwork should already have been done.

The result is not the removal of due diligence. It is a more efficient and reliable basis on which to carry it out. For an AI agent, which must act on information it can identify and interpret, that difference may be particularly important.

A separate commercial structure

A further advantage is that the foundation is a legal entity in its own right. The operating company may continue to collect and maintain the information and use it in developing its products and services. The foundation can hold the registered data asset and provide the structure through which it is governed and licensed.

This creates a degree of separation between the asset and the wider activities of the operating company. It may also provide continuity if the ownership, management or strategy of that company changes. The effect will depend on the rights and permissions that have been dedicated to the foundation and the terms of the surrounding arrangements, but the commercialisation of the asset no longer needs to remain embedded within the company's wider business.

The underlying data does not necessarily need to be moved into systems operated by the foundation. It may remain hosted by the operating company or a specialist provider. The foundation can require it to be managed in accordance with the standards applying to the asset and can enter into licences under its agreed governance arrangements.

This allows the licensing activity to be developed through a dedicated and scalable structure while the operating company continues to maintain the data and carry on its ordinary business.

Who stands behind the shopfront?

Technology remains an essential part of the transaction. It can describe the asset, present the licence terms and confirm which version is available. It can also process payment and provide authorised access. Behind that technology sits the foundation as the legal person through which the registered asset is governed and licensed. The Accredited Assurance Provider assesses whether the asset meets the applicable standards. The Register records its identity and status, while the Data Enforcer has a separate role in supporting continuing compliance with the governance framework.

Together, these arrangements provide the foundations of a shopfront through which people and AI agents could discover and license data with greater confidence.

Companies intending to commercialise substantial datasets will need to undertake much of the preparatory work in any event. They will have to understand their data, improve its condition and establish the rights under which it can be used. A carefully designed Data Asset Foundation allows that work to be carried out with a scalable legal and commercial structure in mind.

AI agents are already beginning to discover content, consider terms and make payments. The next challenge is to create data assets with which they can transact confidently. If AI agents are going to become customers, someone will need to build the shops and establish why the products on their shelves can be trusted.