How does it work in a real audit scenario?

A good example is expense validation. In the past, we’d select expenses to test, request the supporting documentation and spend time comparing it all manually, then create a workpaper. Today, Clara’s agent identifies which expenses need testing, ingests the source documentation electronically, extracts the necessary data, generates the workpaper and automatically runs the testing with all of the dates, amounts and so on.

"The AI handles the preparation, but the auditor guides it. We’re selecting the right data elements, combining the best methodology steps and reviewing the results."

Thomas Mackenzie

Audit partner in charge, Global Solution Group and audit chief technology officer, KPMG

The human now comes in to review outcomes and apply judgment, rather than spending hours preparing documents. That’s just one example of the kind of capability we’re delivering with this new technology.

What does success with this technology look like here for you, and how are you measuring it?

For us, success is multi-dimensional. First and foremost, it’s about audit quality. Everything we do must improve audit quality and enhance reliability in the capital markets. The more we use technology to perform reconciliations and comparisons, the more we can do it dependably, predictably and consistently. That reduces human error. We’re no longer at the mercy of how someone’s feeling that day.

We can also look at more data. Today, audits rely on sampling, like dropping a bucket in a river and hoping to catch something. If we can do 100% testing, it’s like placing a filter across the river. You catch everything that flows through. That significantly improves our ability to detect exceptions.

Another major benefit is the experience of our professionals. I didn’t know what an auditor really did when I was in university. Once I figured it out, I didn’t expect I’d spend so much time comparing things and reconciling them. We took stories like this into consideration, and this aimed to bring in technology that changes that. It frees auditors to focus on risk, analyze exceptions and have deeper conversations. On top of improving quality, we hope it makes the work more engaging and more attractive for our people, too.

For clients, it’s a much better experience. They don’t want to be told an invoice doesn’t reconcile. They want a smooth, no-surprise audit that delivers insight into how their business is performing. This technology helps us do that.

A good auditor brings skepticism and other human elements to the audit. How does AI do this, and can it enhance, but not replace, the human auditor?

We often say we’re keeping the human in the loop, and that’s central to everything we’re doing. We view technology as something that empowers, enables and transforms our professionals. But we also believe it doesn’t replace them.

Auditors still need to bring judgment, skepticism and experience. So we’re focusing on how we train our professionals to use the technology and how we help clients understand what’s changing. The workforce itself is evolving.

You’ll still need financial accountants, but their development path is changing. They won’t necessarily have a long apprenticeship over years of experience. The technology will accelerate their learning curve. We have to rethink how we train them to be skeptical, analytical and effective with data. That way, they can spot risks, consider implications, and contribute meaningfully, not just follow checklists.

I believe that’s the opportunity we have here. We can use technology to enable deeper thinking by handling the mechanical work.

In the research, development and updates of Clara, how do you balance the rapid demands of the technology combined with the strict and unclear regulatory status of the auditing?

I won’t try to predict where regulation is going because I’d only get myself into trouble, but I can tell you that we need to bring regulators and standard-setters along on the journey. We can’t get too far ahead of them. At the same time, there’s work to be done to help the standards catch up to what technology is now capable of.

In the U.S., we’re not working against regulators, we’re working with them. We share what we’re doing and where we think things are headed, and we gently suggest where standards could become clearer or more permissive. There’s an educational element to it.

"We often say we’re keeping the human in the loop, and that’s central to everything we’re doing. We view technology as something that empowers, enables and transforms our professionals. But we also believe it doesn’t replace them."

Thomas Mackenzie

Audit partner in charge, Global Solution Group and audit chief technology officer, KPMG

We also have a very deliberate deployment model. We pilot new technology, then do a limited deployment, then scale up. It’s not the fastest path, but it lets us control the rollout, show regulators that it enhances audit quality and avoid surprises.

Ultimately, I think market forces will help move things along. Yes, the PCAOB may be in one place now and somewhere else in four years, but in the end, capital markets will price the value of an audit. They’ll demand a certain level of assurance, and that demand will shape how both the profession and the regulators evolve.

Recommended Reading

• ‘Quality audits protect people,’ PCAOB Chair Erica Williams says