Key Takeaways
- AI can improve performance while making ideas and decisions more alike.
- Strategic convergence can narrow the options companies consider.
- Standardization helps in some areas but can weaken differentiation in others.
- Productive friction helps preserve independent judgment and competitive advantage.
Nobody Decided to Become the Same
In the United States, many landlords use revenue-management software from RealPage to help set rents. The software draws on market data to recommend prices and includes an “auto accept” option that can implement recommendations with little or no human intervention.
In 2024, the U.S. Department of Justice sued RealPage, alleging that competing landlords shared nonpublic, competitively sensitive information through the system and that the software helped align rents rather than letting landlords compete independently. The DOJ also said a large number of landlords used “auto accept”.
RealPage is an antitrust case, not a generative-AI experiment. But it captures the central question: what happens when competitors delegate similar decisions to similar systems?
Normally, convergence should create its own correction. If everyone reaches the same competent answer, being different becomes more valuable. The problem is that companies may remove the very friction that once allowed those differences to emerge.
Differences Nobody Planned
A 2026 Harvard Business Review article on the “Agentic Convergence Trap” makes this point directly. Human judgment naturally introduces variation: one manager overrides a recommendation, one team moves more slowly, one regional leader interprets the market differently. Those frictions can look inefficient inside a company, yet they also produce strategic diversity.
When organizations automate, some of that variation disappears because removing delays, inconsistency, and human review was part of the goal in the first place.
So, some of what makes companies different was never planned. Similarity can arrive as a side effect of becoming faster and more consistent.
Markets can still reward companies that break away from the pack. But that correction only works if someone notices the convergence and still has the authority to choose differently.
The Gap Closes in Both Directions
A 2024 study by Anil Doshi and Oliver Hauser shows the trade-off clearly.
They asked 293 people to write short stories. Some wrote alone; others first received story ideas generated by GPT-4. Separate evaluators then rated the stories without knowing who had used AI.
AI-assisted stories were rated as more novel, useful, and enjoyable, with the biggest gains concentrated among less-creative writers. At the same time, the AI-assisted stories became measurably more similar to one another.
The benefit was real: AI raised the floor. Competitively, the same mechanism can help weaker rivals close part of the gap. That does not mean every company will converge, but it changes where advantage has to come from.
There is another revealing detail. The writers’ own ratings showed no significant difference in novelty or usefulness between the AI-assisted and human-only groups. The convergence was easier to measure from the outside than to feel from the inside.
Short-story writing is not business strategy. But it demonstrates a mechanism worth watching when AI begins shaping more consequential work.
When AI Narrows the Choices
The stakes rise when AI stops helping with the work and starts advising on decisions.
A 2026 Journal of Business Research study compared 2,400 strategic recommendations from four leading AI models with a benchmark of 218 managers facing the same kinds of institutional pressures.
Across the AI recommendations, 88.2% favored “acquiescence”: broadly adapting to outside expectations rather than resisting, avoiding, or trying to reshape them. The researchers describe the models as potential “institutional amplifiers” because they compressed the range of strategic responses.
Compliance is often the right answer. The study did not establish that conformity was necessarily wrong. The more interesting issue is what happens before the final decision: if the AI strongly favors one class of response, other strategic options may receive less attention before a manager ever evaluates them.
That problem becomes sharper if AI is used twice: first to generate the options and then to evaluate them. A list can appear diverse while both the generator and the filter share similar assumptions. What looks like independent validation may not be independent at all.
Where Similarity Pays, and Where It Costs
Convergence is not automatically bad.
In documentation, reporting, compliance, routine support, and many operational processes, standardization can improve quality, safety, and efficiency. Companies do not need to be different everywhere.
The question is where similarity becomes costly. In strategy, product, brand, and customer experience, convergence can weaken differentiation. In critical infrastructure and financial systems, the risk can be more serious: similarity can reduce resilience.
When Everyone Is Wrong at the Same Time
The Financial Stability Board warned in 2024 that AI adoption could amplify vulnerabilities including third-party dependencies, service-provider concentration and market correlations.
If institutions depend on overlapping models, providers, and signals, they may react to the same error in similar ways. A decision that appears reasonable inside each bank can become dangerous when repeated across the system.
The danger is not that the banks look alike. It is that they may be wrong together, with fewer genuinely independent checks left.
Friction, on Purpose
A 2026 Microsoft Research paper on AI “vibe-coded” websites offers a useful way to think about the solution. The researchers examined where design homogenization could emerge in AI-mediated web creation and argued that frictionless generation can reinforce dominant conventions.
Their proposed response is “productive friction”: deliberately preserving moments where creators have to question a default output rather than simply accepting it.
The same idea applies to business. Keep human accountability around decisions where variation matters. Give AI objectives your competitors may not share. Ground it in knowledge and data they do not have. And do not assume that the first polished answer, or an AI-generated evaluation of that answer, is independent judgment.
Before automating a decision, ask:
If every competitor made this exact choice today, what would happen to us?
If the answer is “not much,” standardization may be exactly what you want. If the answer is “we become interchangeable”, that is where productive friction belongs, and where the decision becomes yours to make, not the software’s.
