Kyriba Adds Shareable TAI Skills and Document Uploads for Treasury Teams

Kyriba has expanded TAI, its Trusted Agentic AI for treasury, with shareable Skills and document uploads. Users can now share reusable AI workflows across teams and upload documents of up to 3 MB to support payment reviews, exception research, and financial analysis, the company announced.

Both capabilities broaden how treasury teams use the AI. More people can run the same workflow, and the agent can draw on additional supporting information. Together, they move TAI further from answering individual questions toward supporting repeatable treasury work across a team.

Where Treasury Teams Lose Time on Manual Processes

Treasury decisions depend on pulling together bank balances, forecasts, payables, and policy limits, often under time pressure.

Kyriba’s 2026 CFO Risk Radar, a survey of 1,354 CFOs and senior finance leaders, suggests many teams are still building the automation and visibility that work requires. In the survey, 76% said they are not running continuous or automated modeling, and 79% reported some degree of financial impact from inadequate risk visibility over the past year.

The daily work behind that visibility is detailed and repetitive.

Félix Grévy, Kyriba’s SVP of platform, data, and AI, describes one treasury team whose global cash staff check bank statements every morning and investigate differences between balances in their ERP and in Kyriba. The same team is working through more than 300 transaction-mapping rules, with FX budget codes producing the most exceptions.

Kyriba points to HelloFresh as an example of what AI can change. The company’s three-person treasury team uses Kyriba’s AI to identify hard-to-reconcile payments in seconds and answer audit questions that previously took hours of research.

Kyriba’s two new features are aimed at the same kind of work.

What Kyriba’s New Capabilities Bring to TAI

Kyriba builds its reusable workflows as what it calls TAI Skills. A Skill is a set of plain-English instructions that tells TAI how to carry out a specific treasury task the same way each time, with a record of each step for the user to check.

Sharing can let one analyst build and test a process, then hand it to the rest of the team.

One Kyriba example addresses the budget-code problem described above. It scans recent cash flows for missing codes, suggests coding based on past patterns, and sends unclear cases to a person for review.

Document uploads add a second source of input. Users can attach files of up to 3 MB so the agent can work from them alongside the data already in Kyriba.

The planned connections to other applications fit the direction Kyriba has described for TAI. At KyribaLive 2026 in April, it showed a capability called Canvas that brings data, recommendations, and approval-ready workflows from multiple systems into one guided process, with the treasurer approving each action.

Why Governance and Data Quality Will Determine AI Adoption

How far teams take these features depends on two conditions: whether people can account for what the agent did, and whether the data behind its work is sound.

On the first, TAI is designed to operate within user permissions, segregation-of-duties controls and established approval processes. Skills provide a record of the instructions and steps for users to review, while recommendations remain subject to human judgement and applicable approval thresholds.

Bob Stark, Kyriba’s Global Head of Market Strategy, adds a readiness test he calls Know Your Agent: before acting on a recommendation, a team should be able to say which agent produced it and who is accountable, what data it used, what it concluded and why, and which authorized person reviewed it.

“The accountability has to come before the autonomy,” Stark wrote.

On the second, much of the data for SAP customers starts in the ERP. More than 1,000 Kyriba clients run SAP, and Kyriba’s SAP-certified integration for SAP ECC 6.0 and SAP S/4HANA feeds payments and open payables and receivables into the platform.

In SAP environments, the two conditions go hand in hand: shared workflows make the work consistent and traceable, and current SAP data makes its results reliable.

What This Means for SAPinsiders

Turn repeated treasury tasks into shared workflows. Processes like daily balance checks and cash-flow coding already follow known steps, which makes them natural candidates for reuse. Teams get the same results each time and a step-by-step record.

Establish accountability and data quality before expanding AI. Decide who owns each shared workflow, who reviews its recommendations, and whether the underlying ERP data is reliable. These controls give treasury teams a stronger foundation for expanding AI use.

Build toward connected, governed execution. As treasury teams connect more systems and data sources, they should focus on workflows that make the path from information to recommendation to approval easier to understand. The aim is not autonomy for its own sake, but faster, more consistent, and more accountable execution.