As you read this, an accounts payable (AP) clerk is busy keying invoice data into a system that could read it automatically. Another is chasing an approval by email that a workflow engine could route in seconds. A third may well be reconciling a supplier statement that artificial intelligence (AI) could cross-reference against the ledger overnight.
None of these AP staff chose a career in finance to do that work. And increasingly, they will not have to. That’s precisely where fear arises: If AI can do all of those tasks, what is left? The anxiety is understandable.
However, as someone who builds the AI systems that power these workflows, my perspective is that AI is not coming for AP jobs. It is coming for the parts of the AP job that nobody actually wants to do.
Accounting is deterministic, AI is probabilistic
Accounting, at its core, is a deterministic discipline. A VAT calculation either conforms to the rules or it does not. There is no “probably correct” in a set of audited accounts.
AI works differently. Every prediction an AI model makes has a probability attached to it. Even the most capable models available today are fundamentally probabilistic systems. They infer and predict. They surface the most likely answer based on patterns learned from historical data.
In most domains, that is fine and often remarkable. In accounting, it creates a specific and important gap because the output of an AP process can not live in a world of probabilities. It lives in a world of certainties. A model might correctly extract and match invoice data at an impressive rate, genuinely transforming the economics of AP operations. But in a business processing hundreds of thousands of invoices a month, even a small error rate produces a meaningful volume of items where humans need to step in, understand the context, apply the rules, and take accountability for the answer.
High automation rates and human oversight are not contradictory ideas. They work together.
This is the second disruption in finance
Decades ago, businesses ran their accounting operations with literal floors of office workers performing calculations by hand. When automated software arrived, it did not just speed things up, it eliminated that layer of work entirely. Despite this, finance did not collapse as a profession. It grew. Those who adapted moved into new roles involving analysis, forecasting, and decision-making that the software could not handle. What looked like the end of a job role at the time turned out to be the start of a more valuable one.
AI in Accounts Payable is the same story, one chapter later. The manual invoice keying, the approval chasing, the statement reconciliation – this is today’s floor of clerks and AI is clearing it.
While independent research and real-world practitioner experience show that AI substantially reduces the volume of manual processing, it does not eliminate the need for human involvement. It simply concentrates it.
When execution becomes cheap, judgment becomes valuable
There is a useful principle that keeps showing up across different professions touched by automation: As the cost of execution falls, the value of judgment rises.
When invoice routing and matching become largely autonomous, AP professionals can stop spending their days on process and start spending them on decisions. They have more time to ask important questions.
Which vendor relationships need attention? Where is cash sitting unnecessarily? Which exception requires escalation? Which anomaly is a data error, and which one is a fraud signal worth investigating?
With AI handling the analytical burden, AP teams can spend their time on quality control, stakeholder relationships, and compliance. This is work that directly drives spend optimisation and better financial control.
Where human value grows
● Cash flow and working capital: When AP teams are not buried in processes, they can engage more meaningfully with treasury. AP moves from a downstream processor to an active contributor to liquidity strategy.
● Supplier risk: AI is good at flagging anomalies. It is not good at interpreting them (yet). Business context, supplier history, and relationship knowledge still sit with people, meaning trust remains human-led.
● Complex exceptions: Real-world AP contains ambiguity: partial deliveries, pricing disputes, contract interpretation, non-PO service invoices. These are precisely the situations where experienced AP professionals earn their place.
● Controls and accountability: Someone needs to define the policies AI agents operate under and audit their decisions. When an AI system approves a fraudulent invoice, the algorithm is not accountable. Finance teams require human ownership of the process.
Practical steps for AP professionals to thrive in the AI era
● Build basic AI literacy: Understand how the model extracts data, matches invoices, and flags exceptions. That knowledge makes you incredibly difficult to replace.
● Own the exceptions: Somewhere between 2% and 7% of invoices or transactions will consistently need human judgment, regardless of how good the automation is. Become the expert who resolves disputes, interprets ambiguous contracts, and applies context that a model cannot infer.
● Get involved early: AP professionals who contribute to how systems are configured report the highest job satisfaction and see the fastest improvements in their teams.
● Strengthen relationships: Use the time you save to deepen ties with suppliers, treasury, and procurement. Relationships cannot be replaced by AI.
● Move up, not out: Once AI has automated processes, the AP professionals who matter most will be the ones who understand cash flow, supplier strategy, and compliance well enough to make decisions on information that AI can only surface. Start building that knowledge now.
Will some entry-level, manual roles disappear? Yes, and it is important to acknowledge that directly. However, the skills that made someone effective at AP are not obsolete. They will become the foundation for overseeing a more powerful system.
Done well, AI transforms AP from a transaction processing function into a source of financial intelligence. The tools handle scale while people handle consequence.
Don’t ask whether the floor of clerks is being cleared out, ask what your teams are doing with the space this creates.
ENDS/