HCLTech Wants AI Agents to Run Corporate Finance

For years, enterprise finance has been a story of automation. ERP systems digitised processes. Workflow tools moved work faster. Robotic process automation (RPA) took over repetitive tasks.

But the human decision-maker remained at the centre.

HCLTech now believes the next phase will be different. AI agents could move corporate finance from automating individual tasks to running entire processes, making routine decisions and responding to changing business conditions in real time.

The catch is trust, and how much decision-making should companies hand over to machines? And who remains accountable when an agent gets it wrong?

Upjit Ghuman, Executive Vice President and Global Head of BPO at HCLTech, believes the answer lies in what he calls “governed autonomy.”

“Ask a finance leader what two decades of technology actually delivered, and the honest answer is usually: faster paperwork. ERP, workflow tools and RPA digitised the steps, but a person still had to decide at every junction. So the work kept waiting on people. Autonomous finance changes how the work itself gets done,” Ghuman tells AIM.

In May, HCLTech launched its Autonomous Finance Platform (AFP), built on Google Cloud’s Gemini Enterprise AI stack. The platform is designed to move enterprise finance from process automation to what HCLTech calls “outcome orchestration.”

The idea is simple. Instead of an AI system completing one step and handing the work back to an employee, agents can coordinate an entire finance process, make routine decisions and improve outcomes over time.

That could change how enterprises manage everything from invoice processing and collections to forecasting and cash management.

But it also raises a bigger question for an industry built around rules, controls and human approvals: where should autonomy stop?

From Automation to ‘Outcome Orchestration’

AFP combines Google’s Gemini Enterprise stack and secure cloud infrastructure with HCLTech’s finance expertise and operations experience.

It uses agentic AI across key finance functions, including Invoice-to-Pay, Order-to-Cash, Record-to-Report and Financial Planning and Analysis (FP&A).

The ambition is to move beyond automating individual tasks. “The shift is from process automation to outcome orchestration,” Ghuman says.

An agent could, for instance, identify an overdue payment, assess the customer’s history and initiate the appropriate collection action. It could also coordinate with other agents or systems to complete the process.

But HCLTech is not proposing a finance function that runs without human oversight. “We call this governed autonomy, not uncontrolled autonomy,” Ghuman says.

Under this model, agents handle routine decisions and execution within predefined boundaries. Finance leaders retain control over policies, risk thresholds and delegation levels.

High-priority decisions remain with people. If an agent encounters a questionable Invoice-to-Pay decision, the transaction can be routed to a human for review before execution. That distinction could become critical as agentic AI moves deeper into enterprise systems.

Traditional automation generally follows predefined rules. Agents can interpret context, make decisions and take action across multiple steps. That makes them more powerful, but also introduces a new layer of operational risk.

The question is no longer just whether an automated process works.

It is whether the organisation can explain why an AI agent made a particular decision, whether that decision was within policy and who approved the rules governing it.

Ghuman says HCLTech is accountable for the platform itself, including security, explainability, audit trails and guardrails. Clients, meanwhile, define the operating policies and risk thresholds within which agents can act.

The division of responsibility is deliberate. The agent may execute the work. But the organisation remains in control of the rules, he says. 

The Opportunity: A Universal Finance Problem

The business case for autonomous finance starts with a problem that most large enterprises already understand. 

Money gets stuck in the wrong places. Revenue leaks through cracks in complex processes. Payments arrive late. Forecasts become outdated before they reach decision-makers.

“The distance between how fast the business moves and how fast finance can respond, that gap is the opportunity,” Ghuman says.

That gap has become more important as businesses operate across increasingly complex supply chains, geographies and digital channels.

Finance teams may have access to more data than ever. But having data is not the same as acting on it quickly.

For CFOs, the promise of agentic AI is therefore bigger than reducing headcount or processing invoices faster. It is about turning finance into a more responsive operating function.

That is why early adoption is likely to be strongest in sectors with high transaction volumes, complex finance systems and significant working-capital pressures.

Banking, insurance, manufacturing, retail, consumer goods, telecom, healthcare, and energy and utilities are among the industries that could see early use cases.

Large shared-services and global capability centre (GCC) environments could also become important testing grounds. These organisations already have standardised finance processes, large transaction volumes and centralised controls — the conditions needed to deploy agents at scale.

The Numbers — And the One That Matters Most

Ghuman pushes back on the idea that autonomous finance is primarily a cost-cutting exercise.

He sees the opportunity across three areas: revenue, cost and cash.

The potential impact, however, depends on where an organisation starts.

On cash, HCLTech estimates that autonomous finance could improve cash flow by up to 25%. The company also points to a potential 36% reduction in the cash-conversion cycle and a 23% reduction in days sales outstanding (DSO).

The productivity opportunity is also significant. Ghuman cites potential cost-efficiency gains of 2–3%, productivity improvements of up to 50% and a roughly 35% reduction in exceptions.

On the revenue and control side, he points to potential improvements in discount capture of about 40% and straight-through processing of roughly 35%.

Duplicate payments could fall by close to 90%, he says, along with reductions in revenue leakage and improvements in working capital.

These figures are potential outcomes, not guarantees. They depend on the quality of an organisation’s processes, data, controls and operating model. AI cannot fix a fundamentally broken finance process simply by sitting on top of it. Enterprises still need clean data, clear policies and well-defined workflows.

In fact, agentic AI could expose weak processes faster than traditional automation did. That makes the transition as much about redesigning finance operations as deploying new technology.

The Real Test Is Trust

For HCLTech, the real test of autonomous finance is not whether AI agents can execute processes. It is whether enterprises will trust them to make decisions with real financial consequences.

The question is shifting from “Can a machine do this task?” to “Can it be trusted to decide what happens next?”

That shift could reshape the finance workforce and the BPO industry. As agents take over repetitive, rules-based work, humans are likely to take on more exception handling, policy setting, oversight and strategic decision-making.

For enterprises, the challenge will be defining where agents can act independently and where human intervention remains necessary. HCLTech’s approach is to give agents autonomy within defined policies, thresholds and governance frameworks.