Artificial Intelligence (AI) has constantly been written about for the past three years. The hype, the digital transformation it promises, the productivity gains, and the threat to jobs, particularly in professional services, and finance, specifically. It is a topical and sometimes divisive subject.
At the same time, many of us have already started using these tools — quietly inputting data into ChatGPT and similar Large Language Models (LLMs) to get answers we did not know, or to do work faster than we used to. And this is often done without fully understanding the data privacy implications, or the quality of the responses we are getting back. Hallucinations are now a daily part of the AI experience, and only those with deep subject knowledge can spot them.
The question worth asking — sitting here in South Africa in 2026 — is whether anything has actually changed. Is this really a turning point, or is it more of the same hype on a longer timeline? My answer is that yes, something has changed.
The CFO question that matters
Whatever side of the AI debate you sit on — hype-and-excitement or doomsday — the question a CFO has to answer is fundamentally a Profit and Loss (P&L) question. How does this make the business more efficient? Either by streamlining back-office tasks, or by improving the customer experience. And critically: on a risk-adjusted basis, accounting for the changes it brings to compliance, governance, and operational risk.
That framing matters because over the past several years the CFO role and the CIO role have moved much closer together. The CFO of today, and certainly of the next decade, needs a real understanding of systems — their capability, their limitations, and most importantly how they get adopted in a way that actually delivers business outcomes.
What changed in 2026
For as long as I have been in finance, the people in our profession have been synonymous with number crunchers. It is a vast oversimplification, but it captures something real about the volume of repetitive processing the function carries.
What has happened in 2026 is that a genuinely capable new number cruncher — in the form of LLM-based AI — has arrived, and the gap between what the technology can do and what is accessible to the average finance team has closed faster than anything I have seen in my career.
Three developments, in particular, have changed the conversation around AI in finance. They are practical, operational, and already affecting how finance teams work.
AI becomes operational
AI capability has moved from novelty to reliability for structured finance work. The current generation of models can draft management commentary from monthly numbers, reconcile transaction descriptions to a chart of accounts, prepare first-draft variance analysis, and handle routine reporting tasks at a quality level where the review time is now materially shorter than the drafting time would have been. This was not true 18 months ago.
AI is already embedded
The platforms our finance teams already use have absorbed AI directly. Microsoft Copilot is built into Excel. Xero has integrated AI into its core platform. Sage, SAP, and Oracle have all done the same. The decision is no longer whether to invest in a new AI tool — it is whether to actually use the AI capabilities that already arrived in the software you are already paying for.
AI aids governance
Enterprise governance has caught up. Data residency, audit logs, access controls, model explainability, and POPIA-compliant deployment are now standard features rather than aspirational ones. There is finally a shared vocabulary that CFOs, auditors, and IT can use to discuss AI risk in a structured way.
The technology is now genuinely capable of meaningfully improving financial processes. And the barrier to access has dropped to a level that finance teams can actually use.
Opportunity with complexity
The most exciting consequence is that financial processes, which used to be the preserve of large companies with serious technology budgets, are now accessible at any level. Real-time management reporting, automated reconciliations, anomaly detection in transaction data, AI-assisted analysis were all available to listed companies and large corporates a few years ago. But they are now within reach of a 30-person owner-managed business instantaneously.
This improves decision-making, exposes inefficiencies in business processes, and can lower the cost to serve. The traditional structure of a finance function — highly trained and qualified staff doing a range of work from low-quality repetitive tasks all the way through to higher-order thinking and problem-solving — was built around the constraint that the lower-value work had to happen somewhere. AI dissolves that constraint.
The lower-value work moves to machines. The higher-order work stays with people, and they can do more of it. This is where I want to be clear about my own view: AI is an opportunity to leverage your finance team, not to replace it. It should bring down your cost-to-process per transaction across the financial processes. It is, ultimately, a productivity tool.
AI is an opportunity to leverage your finance team, not to replace it.
Where adoptions fail
The route to a well-adopted AI capability in finance is, in fact, very complicated. And this is the part of the conversation that is most often ignored.
The gap between how businesses are actually run in South Africa, across any size and any industry, is the most important starting point to understand. AI adoption — even when made dramatically easier by what arrived in 2026 — is fundamentally a business process change.
Businesses are not interchangeable. Even two firms in the same industry, of the same size, can run their operations and serve their customers in such different ways that the same technology change produces different outcomes. Drop a process change into the wrong context and you can damage operations or break the customer experience. I have seen this happen multiple times in my career, with technologies far less powerful than what is on the table today.
What that means in practice is that AI adoption in finance is a team effort, not a finance project. Technology departments are critical for assessing the right infrastructure, integration, and security posture. HR matters for change management — how the finance roles will shift, what training is needed, how confidence with the new tools is built. Risk and compliance need to be involved early so that the controls are designed in rather than retrofitted. Internal audit needs to understand what is being deployed before they audit around it. Operations needs to think about how the upstream processes that feed finance might themselves change.
Get this right, and the competitive advantage is genuine and lasting. Get it wrong, and the cost — both financial and reputational — is significant.
Where I would begin
If I were sitting in a CFO chair today rather than having just launched my own firm, my first move would not be a vendor selection. It would be three quieter conversations.
With the CIO — to map what AI capability already exists in our current technology stack and is sitting unused.
With the head of HR — to discuss how the finance team’s roles will need to evolve and what change management investment will be required.
With the auditor and the audit committee — to start the governance conversation early rather than be confronted with it after the fact.
Then I would pick one workflow — just one — that genuinely costs the team time and produces a generic output, such as month-end management commentary, variance analysis, or accounts payable exception handling. I would commit to AI-assisting that one workflow for a defined period, measure the result honestly against time saved and quality delivered, and use that as the foundation for the next decision.
This is not a strategy. It is a pilot, properly run.
The bottom line
This is the year. The technology is genuinely ready, the platforms have absorbed it, and the governance vocabulary has matured. What is left is the harder work — the change management, the cross-functional partnership, the willingness to engage with the hard part rather than just to buy the software.
The CFOs who do that work in 2026 will look back on this year as the one where their finance function genuinely changed. Those who skip it will spend the year reading more articles and wondering why the promised gains have not appeared.
Thinking about where to start?
Talk to Vedant about what AI-assisted finance could look like for your business.