Tungsten Automation’s Andrew Ng argues that AI’s near-term role in corporate payments is less about letting agents send money and more about connecting fragmented systems, checking transactions and preparing decisions for human approval. His account puts a practical question on the table: how much financial work can be automated before the final click becomes too consequential to delegate?
Watch Desk analysis
What happened
Ng, Tungsten Automation’s head of payments and embedded finance, told PYMNTS that companies’ separate payment systems often leave finance teams with duplicated controls, approval processes and reconciliation work. He argues that structured e-invoicing data, combined with payment histories, counterparty information and company policies, could give AI tools better context for assessing transactions.
The tasks he sees as suitable for AI include gathering information, validating instructions, screening counterparties, recommending payment routes, flagging anomalies and preparing transactions for approval. He argues that people should retain oversight of consequential decisions such as releasing funds, where mistakes can be difficult to reverse.
Why it matters
This is a more grounded picture of agentic finance than handing a bot the company cheque book and hoping for the best. AI could help finance teams make sense of data spread across payment rails, while people keep authority over the money leaving the building.
Ng also makes a broader case for investing in a shared control layer that connects data, policies, approvals and payment options, rather than treating every payment rail as a separate island. That is his argument, not evidence that the approach is already working at scale.
Our read
The useful distinction is between automating preparation and delegating authority. Transaction checks and recommendations may be good candidates for AI; releasing funds is a different proposition. E-invoicing could supply more structured data, but cleaner inputs do not by themselves guarantee sound decisions. Finance teams should judge any system by what it can explain, what it flags for review and who remains accountable when the answer is wrong.
What to watch
- Whether companies put AI into payment preparation and screening workflows.
- How finance teams test recommendations against fraud, compliance and reconciliation outcomes.
- Whether structured e-invoicing data becomes a practical foundation for shared payment controls.
- Where organisations draw the line between an AI recommendation and permission to release funds.
Discussion spark: Should AI be allowed to release routine business payments under strict limits, or should a person approve every transfer?
Sources and evidence
- CFOs Pay a Price for Every Disconnected Payment System – PYMNTS.com (29 September 2026, 08:04 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.