What if the main user of data is no longer human?

Sanne Roubos
18 November - 4-minute read

You use business data to assess customers, monitor suppliers or identify new sales opportunities. Usually, an employee is involved, weighing up the information. But what changes when an AI agent carries out some of that work autonomously?

An agent is an AI system that performs tasks within predefined boundaries. It can retrieve information, combine sources and prepare the next step. This means access to data alone is not enough. The system must also be able to establish which company the information relates to, what it means and when it can be used. This calls for agent-ready data: data with the context an agent needs to work with it.

agent

Why does an AI agent need more than a data point?

Imagine you are reviewing a supplier’s risk score. You know the company, understand how the score is calculated and check the date if you have any doubts. You use that knowledge to decide whether to take action.

An AI agent does not automatically have that background knowledge. It can only interpret the same score properly if it is clear which business it relates to, how recent the information is and what exactly the score represents. Additional information may also be needed, such as the supplier’s financial position or the corporate group it belongs to.

This is where AI-ready and agent-ready data differ. AI-ready data is available, reliable and suitable for use by AI. Agent-ready data also makes definitions, relationships and rules for use explicit. This becomes more important when an agent does more than analyse information and starts preparing checks or directing a process.

Interesting read: From AI-ready to agent-ready: Is your data ready for Agentic AI?

Does your agent know which company it is dealing with?

You can bring internal data together in one place and still be missing important context. A customer name in your CRM, for example, does not automatically tell you which legal entity is behind it. And two suppliers with different names may belong to the same corporate group.

Suppose an agent needs to assess whether a supplier still fits your risk profile. More records alone will not necessarily help. First, these questions need to be answered:

  • Which legal entity does the information relate to?
  • Does that business belong to a larger corporate group?
  • How up to date is the risk indicator?
  • What is the source of the information?
  • What does the indicator mean?
  • Within what boundaries may the agent use the information?

That context makes an assessment easier to trace and explain. It does not mean the data provider makes your decision for you. Your own processes, policies and risk appetite determine the appropriate next step and when an employee needs to be involved.

Interesting read: AI-ready data: The next step after Databricks and Snowflake

How do you make trust understandable to an AI agent?

As an employee, you may recognise a source or know from experience how to interpret a term. For an agent, that knowledge needs to be made explicitly available in a form that software can process.

Think of the source and date of information, an unambiguous business identity and a clear description of a risk indicator. An agent can then take these attributes into account when processing data. This is what we mean here by machine-readable trust: information about reliability that a system can read too.

This does not guarantee that every outcome will be correct. It does, however, make it easier to trace what an outcome is based on. So the question is not just whether AI can access your data, but also whether it is clear which data is being used and how it should be interpreted.

Interesting read: Altares Dun & Bradstreet launches D&B.AI Ecosystem for trusted AI in business processes

What does this mean for your next step with AI?

Data-as-a-Service is about making data available where you need it. In our view, AI agents add another layer: Context-as-a-Service. You provide not just a data point, but also the identity, relationships, meaning and provenance needed to understand it.

This may evolve further into what you could call Trust-as-a-Service. Not as a seal of approval guaranteeing that every decision is correct, but as a way to make information about provenance, timeliness and context of use available. The more autonomously an agent works, the more important that foundation becomes.

Ready to put your data to work for AI agents? Our experts would be happy to discuss the business data and context your AI agents need.

Interested?
Share on social media

Interested?

Fill in your details or call us directly.
We will contact you within one business day.
Or call us directly
The Netherlands (sales) +31 (0)10 322 03 04 Belgium +32 (0)2 765 00 21

White paper

Data Detox

Get rid of your excess data pounds

Dirty data is costing companies worldwide clamshells of money. In addition, they miss out on profits because they cannot keep up with competitors competitors who are smarter with data and analytics. With this white paper we get you ready for efficient data management.

Pdf of 29 pages, 0.4 MB
Data Detox

A free trial of one of our products? Just like that!

Looking up a company or D-U-N-S number?

Looking up an article or topic?

Suggestions

Je keuze voor

quizz outcome