Businesses facing the retirement of experienced treasury staff are looking at AI agents to help keep payment and account workflows running. The immediate pitch is not handing agents the keys to company money, but getting them to connect the systems and steps that people currently hold together.
Watch Desk analysis
What happened
Katherine Weislogel, head of Commercial Payments Client Success at Fifth Third, told PYMNTS that one client had employees who had spent an average of 30 years in their roles and was unsure how to replace their knowledge. She described a commercial payment process that can involve roughly five steps, across different people, systems and decisions.
Weislogel’s argument is that agents could coordinate those hand-offs, while companies document the rules and exceptions experienced staff know by heart. She also described account onboarding as a process where information can be requested repeatedly by different departments. In the smoother experience she outlined, a customer would submit information once and hear back when the account and services were live. She said the difference could be days versus hours. That is an example of the potential, not evidence that AI has already delivered those results.
Why it matters
This puts a less glamorous, more immediate problem at the centre of the AI-agent debate: business continuity. Automating a task in isolation may leave the same bottlenecks between departments. Coordinating the whole workflow could matter more, but it requires connecting systems and capturing the practical knowledge that keeps awkward processes working.
The distinction between administrative work and financial authority matters too. Weislogel said firms are assessing permissions incrementally: an agent might analyse cash positions or prepare forecasts, while initiating a payment is a different decision involving fraud, liquidity and authorisation risks. She does not expect businesses to hand over money movement to agents in the next 18 to 24 months.
Our read
The most persuasive case here is not “let the agent run treasury”. It is “stop making every process depend on a few people remembering how disconnected systems fit together”. That is a real operational challenge, and a more useful starting point than autonomy for its own sake.
But preserving knowledge takes more than installing an agent. Companies still need to identify the rules, exceptions and decisions involved, then show that automation improves the service without quietly moving financial authority somewhere nobody intended. PYMNTS reports Weislogel’s account and examples; they are not proof of a deployed system or measured results.
What to watch
- Whether businesses document workflows and exceptions before deploying agents.
- Whether early use stays with coordination and administration, or expands into payment decisions.
- Whether companies publish concrete results, such as faster onboarding or fewer hand-off delays.
Discussion spark: Should companies use AI agents first to preserve operational knowledge as experienced staff retire, or prioritise keeping those workflows under human control?
Sources and evidence
- Source update (19 October 2026, 08:03 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.