Damian Rezaee

By Damian Rezaee

A calculator is a useful comparison for a lot of software. Put in the same numbers, get the same answer every time. No judgment required, no oversight needed, no risk of the tool quietly making something up. It’s tempting to apply that same mental model to AI in business: treat it as a fast, reliable input-output machine, hand it a task, and trust the result.

That analogy is wrong, and treating AI like a calculator is where a lot of companies are getting into trouble.

I think about this constantly from the marketing side, because brand is where the calculator mindset does the quietest damage. Nobody notices it in a single output. They notice it eighteen months later when the brand starts sounding like everyone else.

The core problem: AI is not deterministic as much as it is probabilistic!

A calculator is deterministic. Give it the same input, and it will give you the same output every time. The logic behind the answer is also clear and traceable. Generative AI does not work in the same way. Ask it the same question twice and the wording, structure, or even the conclusion may change, even when your question stays the same. Industry analysis of enterprise AI in 2026 makes this distinction clearly: deterministic systems are designed to be repeatable and auditable, while generative AI works with probabilities, interprets uncertainty, and produces an answer rather than simply following a fixed set of instructions.

This difference matters because every AI-generated response comes with some chance of being wrong, drifting from the original task, or producing something the business did not expect. When AI is used again and again across many tasks, those small risks can build up over time and create larger problems. A calculator that occasionally gave you the wrong answer with complete confidence would be almost impossible to trust. Generative AI can do exactly that. This is why treating AI output as ground truth, in the same way you would trust a calculator or a spreadsheet formula, is a serious mistake.

So judgment is the king, speed is the army!

Harvard Business Review’s coverage of AI in 2026 points to a similar issue: the main challenge for many organizations is no longer how fast AI can produce work. The real challenge is whether employees know what to trust, what to question, and what needs to be improved before they use what AI gives them. Many companies have trained people to use AI tools, but not necessarily to judge the quality of AI output. That creates a problem exactly where the calculator comparison falls apart. A calculator usually does not need someone to check its answer. AI does.

This also affects how companies develop future talent. Good judgment has traditionally been built by doing the difficult, repetitive, and sometimes boring work early in a career. That is often how people learn to recognize patterns, notice mistakes, and understand why something works or does not work. But AI is now taking over many of those tasks. HBR research warns that junior employees may lose some of the experience that used to help them build judgment. Over time, this could create managers who are expected to judge work they never really had the chance to do themselves, and that could weaken the future leadership pipeline.

AI is supposed to save work, but the evidence shows it can create more!

Part of the reason the calculator comparison sounds appealing is that we assume AI simply takes work away from people. A calculator saves us from doing long division by hand, so it is easy to imagine AI doing the same thing on a much larger scale.

But an eight-month Harvard Business Review study of 200 employees at a U.S. technology company, published in February 2026, found something very different. In practice, using AI led to longer workdays and more burnout. Shocking? As AI made some tasks faster, employees were often given more work, while also having to spend more mental energy checking, correcting, and judging what AI produced. The burden was especially heavy for entry-level and associate employees.

So yes, the mechanical work goes down while the mental work can go up. Someone still has to check the output, fix mistakes, decide whether it makes sense, and take responsibility for the final result, even though they did not actually produce it themselves. And taking responsibility for someone else’s work is stressful enough. Taking responsibility for AI’s work can be even more stressful.

A calculator does NOT come with that hidden cost.

Businesses are starting to treat AI more like infrastructure than just another tool

Harvard Business School faculty looking at AI trends in 2026 describe a clear shift: AI is moving from being an optional tool people choose to use into something that sits inside everyday workflows, business decisions, and customer experiences. That changes what it means to use AI properly. If AI is just a tool you use for one task, you can check the result before moving on. But once AI becomes part of the system itself, businesses need clear rules, oversight, and boundaries around where AI can make judgments and where more fixed, controlled systems are still needed. At that point, AI starts to look less like a spreadsheet app and more like infrastructure that needs to be managed carefully.

The problem is that many organizations have not fully adjusted to that shift yet. A 2026 Harvard Business Review Analytic Services study of 385 business decision-makers found that although many companies already have AI in active use, most are still mainly using it to make existing tasks faster or more efficient. They are not yet using it to seriously rethink the processes that drive business results. In other words, many companies are still adding AI on top of old ways of working instead of redesigning how decisions are made, reviewed, and improved. HBR researchers suggest that this is one of the main differences between companies that are getting real value from AI and those that are simply doing the same small tasks faster.

AI’s real risk is sameness, which is the brand manager’s nightmare

This is where the calculator comparison really falls apart for me. A calculator has no style, and it does not need one. It does not matter if a thousand companies use the same calculator because nobody expects a calculator to make a company recognizable.

A brand is the exact opposite because it needs to be recognized. People should be able to recognize it before they even read the name, just from the colors, shapes, language, tone, or other details that belong to that brand. This is connected to the idea of mental availability and distinctive brand assets discussed by Byron Sharp and Jenni Romaniuk. Brands become stronger when they are easy to notice, easy to recognize, and easy to remember.

The thing is that generative AI can push brands in the opposite direction. AI learns from huge amounts of existing general material, so when you give it a prompt, it tends to give you something that feels familiar and safe. So AI’s safety can become our risk. Ask five companies to use AI to create a campaign from a prompt, and there is a good chance you will get five ideas that sound surprisingly similar. Similar language and structure. Similar concepts. Similar visual direction. Even the same stock-photo feeling and similar colors.

That is dangerous for brands because distinctiveness is already difficult to build. Recent industry analysis on distinctive assets shows that strong and consistent brand assets can improve salience and ROI, while only a relatively small number of brand assets are actually distinctive enough to strongly belong to one brand. If companies start producing large amounts of AI content without strong brand consistency, they may make that problem worse.

There is another side to this in 2026. AI is becoming part of how people discover brands, not only how brands create content. Recent marketing-effectiveness research suggests that a brand’s visibility in AI-generated answers is strongly connected to the brand equity it has already built over time. This makes ownable brand assets even more important. It is not enough for something to simply look different. It should make people think of your brand.

Mastercard’s overlapping circles are a good example. We see them and we know they belong to Mastercard. A color that looks nice but could just as easily belong to five competitors does not do the same job.

The key point is that AI needs to work inside our brand world instead of being asked to create that world for us. Give it your brand guidelines, your approved language, your proof points, your visual rules, and the assets that already make the brand recognizable.

Used this way, AI can actually help protect brand consistency at scale. It can spot the wrong color, an off-brand phrase, or a change in tone across hundreds of pieces of content much faster than a human team could. But if we use it as an endless idea machine without those boundaries, it slowly makes our brand look and sound more like everybody else.

And that is the part of “AI is not a calculator” that matters most to me as a marketer. A calculator can give us a correct and completely forgettable answer. A brand, however, cannot afford to be forgettable. Someone still needs to decide what makes the brand yours.

What this means for how businesses should actually deploy AI

Don’t worry, none of this means AI should be kept out of core business processes. I just mean the calculator framing sets the wrong expectations from the start. A few practical implications follow directly from the research above:

• AI output should be checked before it is used. Since AI output carries a nonzero chance of being wrong even when it sounds confident, workflows need a designated point where a human checks the work, the same way audit checkpoints exist in finance.
• Use deterministic systems where they’re required. Compliance checks, financial calculations, and regulated decisions still belong in rule-based systems with fixed logic. AI can support that work but shouldn’t replace the deterministic layer where auditability is non-negotiable.
• Protect the entry-level work that builds judgment. If junior roles lose the repetitive tasks that used to teach pattern recognition, companies need a deliberate substitute, or they’ll run out of people capable of evaluating AI output a decade from now.
• Measure workload honestly. If AI adoption is quietly extending hours rather than shortening them, that is a signal the tool is being layered onto existing processes rather than genuinely redesigning them.
• Lock your brand inputs before you scale AI content. Feed generative tools your specific distinctive assets, tone, and proof points rather than open-ended prompts, or the output will quietly average itself toward whatever looks like everyone else.

A calculator earns trust because it never needs supervision. AI earns usefulness only when it gets supervision, structure, and judgment built around it. Businesses that skip that step will end up giving AI too much responsibility and trusting decisions that still need human judgment.

Resources

Duncan, D. S. (2026). Help Employees Get Better—Not Just Faster—with AI. Harvard Business Review. https://hbr.org/2026/06/help-employees-get-better-not-just-faster-with-ai

Duncan, D. S. (2026). How Do Workers Develop Good Judgment in the AI Era? Harvard Business Review. https://hbr.org/2026/02/how-do-workers-develop-good-judgment-in-the-ai-era

Ranganathan, A., & Ye, X. M. (2026). AI Doesn’t Reduce Work—It Intensifies It. Harvard Business Review.

Harvard Business Review. (2026). Don’t Let AI Destroy the Skills That Make Your Company Competitive. https://hbr.org/2026/04/dont-let-ai-destroy-the-skills-that-make-your-company-competitive

Harvard Business School, Working Knowledge. (2025). AI Trends for 2026: Building “Change Fitness” and Balancing Trade-Offs. https://www.library.hbs.edu/working-knowledge/ai-trends-for-2026-building-change-fitness-and-balancing-trade-offs

Harvard Business Review Analytic Services & Appian. (2026). New Survey Finds AI Adoption Remains High, Yet Value May Lag Without Modernization and Workflow Integration. https://appian.com/about/explore/press-releases/2026/new-survey-from-harvard-business-review-analytic-services-finds-ai-adoption-remains-high

Axian, Inc. (2025). AI Is Non-Deterministic — And That Matters. https://www.axian.com/2025/10/07/ai-is-non-deterministic-and-that-matters/

Entrepreneur United Kingdom. (2026). Branding, AI and the Risk of Broken Brands. https://uk.entrepreneur.com/technology/mental-availability-ai-era-brand-discoverability

The Brand Algorithm. (2026). How to Build Distinctive Brand Assets (2026 Guide). https://www.the-brand-algorithm.com/how-to-build-distinctive-brand-assets/