Prof Yudhvir Seetharam | The case for friction in an age of automation

Businesses have always focused on removing friction from their processes, with faster approvals, quicker responses and more automated decisions being key markers of efficiency.

More recently, AI and data-driven systems have accelerated these gains, enabling organisations to process information and act at a speed that would have been impossible even a few years ago.

Low-risk, repeatable decisions are well suited to automation. When an outcome is predictable, the process is well understood and errors can be corrected quickly, technology can improve both speed and consistency.

There is a point, however, at which removing friction creates a different kind of risk. Not every decision should happen at maximum speed. Decisions that could materially affect customers, revenue, reputation or long-term strategy often require organisations to pause, assess context, and apply judgement.

This is where friction by design becomes valuable. The principle is simple: the higher the risk, the greater the level of scrutiny should be. The objective is not to slow a business down unnecessarily or create additional bureaucracy, but to ensure oversight remains proportionate to the potential impact of a decision.

For a small business, that responsibility often rests with the owner. AI can help gather information, compare options, identify trends, and produce recommendations, but ultimate responsibility for major decisions still lies with the person running the business.

As organisations grow, responsibility becomes more distributed. Management teams, specialist functions, governance structures, and boards exist to bring different perspectives to important decisions and challenge assumptions before action is taken.

AI should be viewed in a similar way. It can be a valuable voice in the room, but not the final authority. Its outputs may be informed by vast amounts of data, yet decisions still need to account for ethical considerations, reputational implications and factors that may not be fully reflected in the data itself.

This becomes particularly important when businesses rely on external technology providers. Organisations may not build every AI tool they use, but they remain responsible for how those tools affect customers. The technology provider therefore becomes part of the business's own risk profile. Customers are unlikely to distinguish between the supplier and the organisation that selected the technology. As with any critical supplier relationship, businesses need confidence in who they rely on and whether they are comfortable attaching their reputation to that relationship.

Friction by design is not only relevant at the point of decision-making. It is equally important when reviewing outcomes, particularly the exceptions and edge cases that broader performance measures can overlook.

Businesses typically monitor performance through error rates, complaint volumes, and other aggregate measures. The challenge is that these can mask issues at the margins. A process that works effectively for most customers may still create poor outcomes for a smaller group.

Individual complaints therefore deserve close attention, as they can expose flaws in decision logic, unintended consequences or experiences that broader performance metrics fail to capture. They may also reveal wider issues affecting people who have simply chosen not to complain.

As technology evolves, organisations must continually reassess which decisions warrant additional scrutiny and how automated outcomes are reviewed. Frameworks that were appropriate six months ago may no longer reflect customer expectations, emerging technologies, or the scale at which those technologies now operate.

The goal is not to choose between human judgement and automation. As technology becomes more capable, the real differentiator will not be how much friction organisations remove, but whether they know where to retain it. In an age of automation, deliberate pauses may become one of the most important tools of risk management.

*Prof Yudhvir Seetharam is the Chief Analytics Officer, FNB Commercial.