Google Cloud exec on software’s great reset and the end of certainty: we’re shifting from predictability to probability

We are currently witnessing the biggest collision in the history of software.

You’ve likely heard the terms “deterministic” and “probabilistic” in discussions about AI, but what does that really mean? And what does it mean for your business?

On one side, we have the deterministic model. This is how we have built software and businesses for 50 years. Every piece of software you bought – from your CRM system to a basic spreadsheet – has been precise, rule-bound, and intolerant of error. Input A plus Input B always equals Output C. If it doesn’t, you have a bug that needs fixing.

On the other side, we have Generative AI that breaks this rule. It is probabilistic, creative and context-dependent. The same inputs can produce different outputs. It’s a reasoning engine, not a calculator.

This allows you to ask questions that deterministic systems can’t answer. What will tariffs do to my revenue this year? How would conflict in the Taiwan Strait affect my commodities pricing? Questions like these have no certain answer, but foundation AI models can analyze vast amounts of data and model multiple outcomes to inform your decisions. 

The friction you are feeling right now in your operating model, from compliance to quality control, is because we built our business systems to hunt down and eliminate uncertainty. However, you can’t force a probabilistic engine into a deterministic operating model. To harness the full power of Generative AI, leaders must stop treating AI like a faster spreadsheet. 

The winners in this new era will be the companies that stop trying to suppress uncertainty and start operationalizing it. Here are three shifts required to re-wire your business and fully capitalize on the AI future.

Measure Autonomy, Not Just Efficiency

In the deterministic world, software value is measured by access (seats) and efficiency (how fast a human can work). We treated software as a tool to amplify the worker.

Generative AI flips this model. We’re moving from software-as-a-service to “service-as-software,” where the value lies in the outcome, not the tool. If an AI agent drafts a legal brief or resolves a customer ticket, the metric is no longer how much time did a human save by using the software, but whether the human needed to be involved at all.

This requires different metrics. We need to stop measuring effort and start measuring autonomy. Was the AI agent consistently factual? Did it reduce time to decision? What’s the task completion rate? And the metric that matters most for expanding your margins: Did the AI agent resolve the issue without human intervention? The goal isn’t a faster workforce; it’s a workforce that can scale infinitely because the bottleneck (the human) has been removed from the loop.

Manage Uncertainty, Don’t Eliminate It

Most companies try to jam probabilistic AI into deterministic, rule-bound operating models. It doesn’t work. When traditional leaders see an AI model hallucinate, they panic. They want to shut it down until it is “100% accurate”.

But 100% accuracy is a deterministic fantasy. The right approach wraps the probabilistic engine in guardrails that manage the uncertainty. At Google, we talk about “grounding” and confidence scores. Leadership teams need to stop asking “Is this answer right?” and start asking “How confident am I of this output?” At Google, we teach our employees that AI agents aren’t designed to produce answers, but to generate reasoning. 

To avoid this, we need to build systems where AI operates autonomously when confidence is high, and fails gracefully to a human expert for review when confidence drops. 

Much like Google’s AlphaFold provides confidence levels for its predictions about protein structures, your business AI needs to give leaders a score they can respond to. These interventions become the feedback loop that trains the model and drives continuous improvement.

Turn Data Into Feedback, Not Just Facts

This technology doesn’t replace humans, but it shifts their primary function from execution to expertise.”.

In deterministic systems, data was a ledger of truth for reporting and historical analysis. With Generative AI, data becomes instant feedback and action. Your historical data trains your future workforce – your fleet of autonomous AI agents. Messy data creates an incompetent digital workforce.

This demands an evolution of the human role. In a deterministic world, we hired armies of junior employees to do rote execution. In a probabilistic world, the AI does the grinding. It generates the first draft, the initial code, the baseline analysis, instantly.

We now see an evolution unfolding. Initially, humans do the work while AI assists. Then, AI does the work while humans supervise, stepping in when needed. Eventually, AI operates independently while humans audit periodically. If a human needs to approve every decision made by an AI agent, you’ve merely created an expensive spell-checker. 

This creates a massive talent shift. We don’t need people who can just execute; we need people who can audit. AI can generate average work instantly. You need humans who are expert enough to recognize “great” versus “good” in seconds. We need editors-in-chief. We need talent with the expertise to look at an AI output and instantly distinguish between “plausible” and “brilliant.” The apprenticeship of toil is gone; we need to build an apprenticeship of judgment.

The Sailboat and the Train

The greatest competitive advantage will belong to leaders who can tolerate ambiguity in exchange for exponential speed.

Think of it this way: For decades, we have been building faster trains. A train runs on rails (rules). It is efficient, predictable, and goes exactly where you planned. Today, we are building sailboats. They rely on the wind (probabilistic data) and can go places rails cannot reach. But without a rudder (guardrails) and a compass (ground truth), you will capsize.

The leaders who demand 100% certainty will remain stuck in the past, perfecting the efficiency of a dying model. The future belongs to those who learn to embrace and navigate probability.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

This story was originally featured on Fortune.com