Coupa Connects - Jakarta: A Trusted Data Foundation for Agentic Procurement

On 29 September, Cogniviti Labs joined Coupa Connects Jakarta as an event sponsor. The programme explored the agentic enterprise and the role of AI in procurement maturity, opening up a practical discussion about how intelligent spend management can move towards autonomous action while retaining business control.

At our booth, conversations with procurement and finance leaders returned to three concerns: giving agents appropriate boundaries, improving spend visibility, and responding to pressure on suppliers and budgets. Across those discussions, one prerequisite kept coming into focus: trustworthy data.

Before an agent acts, it needs a reliable record

Consider an agent asked to route a purchase request. Before it can make a useful recommendation, it needs to understand the supplier, the item, its category and the rules that apply. A duplicate supplier record or an incomplete description can make that decision harder to explain and harder to trust.

The practical starting point is to examine the information behind a chosen workflow. Which records does it depend on? Where do teams still resolve ambiguity by hand? Who owns the corrections? Those questions help turn a broad AI ambition into a manageable first step.

What we demonstrated with CogniPurge MDM

At Coupa Connects Jakarta, we showed CogniPurge MDM and its capabilities for AI-driven classification against UNSPSC, deduplication and enrichment. These capabilities address the data issues that surfaced in our booth conversations, including repeated supplier records, inconsistent item descriptions and missing categories.

For teams assessing their own readiness, these three areas offer a useful structure:

• Classification: Can records be grouped consistently enough to support meaningful spend analysis?
• Deduplication: Can teams identify records that refer to the same underlying supplier or item?
• Enrichment: Is the information needed for the selected workflow complete and usable?

Improving these areas creates a clearer foundation for automation. It does not remove the need to define permissions, review exceptions or decide when a person should intervene.

Start with a workflow you can explain

A practical pilot should have a clear purpose and a bounded scope. Identify the action an agent may take, the evidence it needs and the situations it must refer to a person. Then check the data supporting that workflow before expanding its responsibilities.

The review should cover more than whether an action completed. Can the team understand why it happened? Can an exception be traced to its source? Can a record be corrected without leaving the same problem elsewhere? These questions make the pilot useful as a learning exercise as well as a technology demonstration.

Continue the conversation

Thank you to the Coupa team and everyone who spoke with us in Jakarta. The questions raised at the event are relevant well beyond the booth: how can procurement teams put AI to work while maintaining visibility and control?

If your team is exploring that next step, we would welcome a conversation about the data and implementation work behind it. Explore CogniPurge MDM or contact Cogniviti Labs.

Read the original event post on LinkedIn.