Businesses Are Debating the Wrong AI Questions

Bill Gates is calling for safeguards on AI because it could cause “a billion deaths” if left unchecked, a statement that comes on the heels of a lead researcher at Anthropic saying he believes there’s a more than 10 percent chance that AI “could kill all humans” within the next decade. While governments, philosophers and the handful of top AI makers debate guardrails, though, it’s crucial for businesses using AI to recognize that if they are focusing on what’s being splashed across the headlines, they could be ignoring the biggest risks that businesses actually face in the AI era.

In addition to fears that the bots will overpower the humans, much of the enterprise AI discourse centers around issues like the perils of rogue agents, sky-high token costs, relatively autonomous frontier AI models and non-sovereign AI.

Threats like these are taking up a lot of the oxygen in the room when it comes to AI concerns, but focusing on these trends instead of on the real-life business risks can lead to poor outcomes that could have been avoided. I’ve seen dozens of businesses significantly reduce their AI risk by focusing their attention much more closely on business outcomes than on technological ones. Here’s a look at how they can do that – and why they should.

Focus on the Business Case, Not the AI Model

Enterprises need to understand that they don’t actually need to use the latest AI models – or pay astronomical token costs to turn every employee into a citizen developer – in order to significantly improve their return on investment.

What they need to do is take a close look at specific areas where their financial or time expenditure could be lowered, and build a plan to tackle that specific problem with all the tools at hand – including not only agentic AI but also generative AI, solvers, machine learning and even rule-based programs, so they can select the right one for the right task.

For example, AI-based forecasting models can reach over 90 percent accuracy when it comes to predicting staffing needs, McKinsey reports. I recently experienced this with a security company that dispatches tens of thousands of security guards to thousands of sites for global giants like Nike and Amazon. Like many other labor-intensive, shift-based businesses, it was facing skyrocketing costs in the form of overtime pay and expensive subcontracting when demand spiked.

No frontier models were needed to increase this company’s ROI and there was no need to lock in on a specific AI provider or technology. Instead, the security company’s starting point was fixing one of the biggest business problems it was facing, and the tools were selected on the basis of how they solved that specific challenge. Smart solvers can deploy teams in an optimized way, cutting overtime costs and avoiding expensive subcontracting thanks to better hiring anticipation. In this case, generative AI combined with other tools reduced turnover, improved planning and helped build schedules that worked better for clients as well as employees – for instance, suggesting how to reduce the number of shifts that required a long commute and how to evenly distribute work-free weekends. Implementing recommendations like these reduced overtime costs 30 percent and brought down subcontracting expenses, while reducing the time it took to build staffing plans and improving the quality and longevity of those plans. When the security company implemented the AI recommendations, it ended up increasing EBITDA by 200 basis points, a major win in an industry with notoriously thin margins. Just because the business problem is big doesn’t mean the AI solution has to be risky.

Indeed, there are many concrete business challenges – such as improving inventory allocation, optimizing collection reconciliation, speeding up bank reconciliation, increasing forecasting accuracy, automating invoice matching and browsing catalogs to answer consumer questions – that are being solved with minimally invasive AI tools, dramatically increasing the bottom line without running the risk of killing humanity or even requiring the use of frontier models or unchecked agents.

Watch Out: Moving Away From ROI Could Incur an AI Hype Tax

In addition to mitigating the chances of out-of-control AI, sticking to concrete business goals can also protect enterprises from incurring the AI hype tax my company found in an analysis of the share price and annual reports of 275 global publicly traded companies.

The 2026 AI Barometer found that the stock prices of the largest global listed enterprises showed a strong correlation between going big on AI and enjoying a superior stock price evolution, but that correlation held only as long as those enterprises disclosed concrete actions and precise targets. The biggest lesson is that general statements about AI transformations aren’t enough for investors, based on the share price analysis.

So what did work? The report found a large stock outperformance gap in favor of companies that were specific about the concrete actions they’d be taking to implement AI and the improvements in efficiency or revenue that would result. Those businesses outperformed companies that mentioned AI in a vague way in their financial reports by 14 percentage points the following year.

I’ve seen firsthand that by sticking to concrete revenue and productivity goals solved by the most efficient and low-risk AI tools – often operating as mini applications embedded into the IT landscape – rather than the newest and glitziest AI models, enterprises can largely avoid the kinds of doomsday outcomes that grab headlines and instead further the business outcomes that increase the bottom line.