Let's just say it plainly: most professionals attending our event are using AI for summarising documents, drafting commentary, and basic efficiency tasks. Very few have ventured into predictive analytics or agentic automation. The honest consensus in the room was that most teams are working at the very tip of the iceberg.

That is not a criticism. It is just the reality and it is a useful starting point, because it means there is a lot of runway ahead. The important thing is understanding what you are building toward.

One of the most valuable contributions of the evening came from our guest who leads AI transformation at a global pharma firm. She offered a framework that immediately cut through a lot of confusion in the room because most people, even those actively using AI, don't distinguish between the three categories. And the differences matter enormously.

Organisations that don't understand the distinction end up applying the wrong tools to the wrong problems, then losing confidence in all of them.

Multiple people in the room arrived at this independently, which tells you something. AI will not fix a data quality problem. It will make it worse. Confidently, at scale, and without flagging the error. If your data is messy, inconsistent, or siloed, you will get wrong outputs presented as right ones.

One attendee shared a real world example that landed with the group: a single employee appearing under 15 different name variations across a 100,000 row spreadsheet. That is not a sole case. That is what most finance teams are working with.

The group reached clear agreement on this: governance before tools. You don't deploy AI into your finance function and then figure out the guardrails. You figure out the guardrails first.

But governance is not a document you write once and file. The landscape is moving too fast for that. What needs to be written today will need revising within months, even weeks. It needs to be a live document, owned by someone who understands what they are governing. Not delegated to a committee that meets quarterly and doesn't use the tools.

The challenge for smaller businesses that came up is that they often don't have anyone with the expertise to write meaningful governance in the first place. That is a real gap, and it is one the market hasn't fully addressed yet.

At a global pharmaceutical firm, the AI roadmap flows from the global CEO down. It is not an IT initiative or a finance initiative. It is a business strategy owned at the very top and cascades through every function.

The contrast with smaller businesses in the room was stark. Several described AI adoption as something being championed by one enthusiastic individual, operating without budget or mandate. In those environments, tools get piloted, results get demonstrated, and then nothing moves because the organisation hasn't decided whether it wants to change.

This was the uncomfortable moment in the conversation, and it needed to be said. If your business hasn't formally adopted AI tools, your people are almost certainly using them anyway. Through personal accounts, free tier tools, or applications built on top of models you have never reviewed. That is shadow AI, and it carries real risk: data confidentiality, compliance exposure, and outputs that no one in the business is accountable for.

The instinct to lock everything down isn't wrong, but it doesn't work. People find workarounds. A more effective response is to create a sanctioned environment where AI tools can be used safely and with visibility, so the organisation knows what is happening and employees aren't navigating governance gaps on their own.

This shift is already happening in job descriptions, and it is accelerating. The question is no longer whether a finance candidate has used AI. It is whether they understand how it works, what its limits are, and how to apply it with appropriate judgement.

From a recruitment perspective, we are already seeing this show up in briefs. Hiring managers want finance professionals who are curious about AI, not intimidated by it. The days when "I am learning Excel or Power BI" was a credible answer are behind us. The equivalent statement now is something like: "I understand the difference between generative and predictive AI, and I know when to trust the output and when to interrogate it."

This was the debate that generated the most heat in the room, and it is one that will follow the profession for the next decade. AI is automating many of the tasks that Graduate Accountants have historically learned on: accounts payable, receivables, reconciliations, basic reporting, variance analysis. These weren't glamorous tasks. But they were the tasks through which junior finance professionals learned to read numbers, spot anomalies, and understand what the data was telling them.

If those tasks disappear, and they are disappearing, where does the foundational knowledge come from? There is no clean answer yet. Some in the room believed that AI will free up Grads to work on higher value problems sooner. Others were less optimistic, pointing out that higher value problems require experience and context that Grads won't have if they skip the foundational years. The finance function needs to be intentional about how it manages this transition or it risks developing a generation of professionals who can operate AI tools they don't fully understand.