Earlier signal detection and more predictive risk analytics
AI puts audit and risk leaders in the driver’s seat of a sports car.
Every leader knows bad news has to travel fast. AI is putting audit and risk leaders in the driver’s seat of a sports car.
For most of the history of the function, we could not actually deliver on the "fast" part. We could analyze harder, sample smarter, write up findings more crisply. But the cadence of audit was set long before us. You saw the pattern after the quarter closed, often long after. The business partner received the information with the decision window already shut.
The sports car changes that.
Full-population analysis has replaced sampling in a growing share of our work. Pattern recognition surfaces anomalies across transaction, communications, and behavioral data as those anomalies form. The signal reaches a business partner in the moment they can still act on it.
A concrete example. One of our risk intelligence projects compared anonymized travel and expense data across specific geographies against the known risk drivers of those markets. The patterns that surfaced were things every company can picture: employees issued corporate cards they never used, personal expenses run through corporate ones, approvals executed by assistants instead of by managers, people approving their own expenses. Depending on what the pattern showed, the response ranged from additional training, to a change in corporate policy, to a targeted audit or a full investigation. The old sampling model would have caught any one of these months later, if it caught them at all. We were able to guide management to make changes that have systemically reduced risk.
Getting to this actually working took three things most organizations underestimate.
The data infrastructure has to carry the weight. Signal detection at scale requires unified access to financial, operational, and HR data without a six-week ETL project for every new question. Most organizations are not there yet. Building it is a multi-year capital commitment, and calling it an AI initiative understates what it is.
The governance model has to match the pace. If a signal surfaces in real time and the governance process still takes six weeks to decide what to do about it, the capability stays theoretical. We had to rebuild some of our own governance processes to move at the speed of the signal. That was the less glamorous half of the transition, and the half that made the rest work.
The team has to have the judgment to weigh the signal. Pattern recognition produces false positives. A mature team is one that knows which signals to escalate, which to watch, and which to set aside. That judgment sits in the people on the team, and it takes years to build.
Audit has always been under pressure to deliver news fast. The new reality is drinking from a growing fire hose of data, structuring and assessing it with good judgment, and building the level of trust with stakeholders across the business that lets us communicate with clarity and speed. If you are driving a faster car and arrive at your in-laws hours earlier than expected, you’d better hope you have the kind of relationship where they’ll welcome you. As we spend more time on real-time and proactive work, I realize the importance of the trust that the team and I have built with our stakeholders. They don’t expect us to show up early or raise issues. When you change, you can’t underestimate the need to also reset expectations with those around you.
Audit and risk are two of the very few functions that can see both left to right and top to bottom across the company. That breadth of visibility is both tremendous responsibility and now incredible opportunity. Few functions are better able to take advantage of the increased speed of collective intelligence.
Audit and risk leaders are in the sports car first. Any function that reviews data, evaluates transactions, or monitors compliance has access to the same car. Finance, security, compliance, operations. Most leaders are not yet driving at the speed the signal allows.
Get new essays in your inbox
One essay at a time. No promotion. Unsubscribe anytime.