Using AI for Competitive Advantage
Ask Claude and OpenAI the same question: recommend 10 international utility stocks for my portfolio. I tried this using the same wording and the same constraints. The LLMs overlapped on 4 stocks: E.ON, Enel, Iberdrola, and National Grid, all obvious picks.
Then it hit me. If I get this list, 50,000 other investors get it too. We all buy the same stocks. The price spikes. I become a chaser in someone else’s crowd. This happens, not because they’re bad choices, but because they’re giving me the average recommendation - everyone’s average.
Scaling Up Averages?
That was only my portfolio. Try expanding this scenario to your company, your consulting firms, your SaaS providers, or even your internal teams.
Your consulting firm uses GenAI to analyze the market and recommends entry into a new customer segment. Your competitor’s consulting firm uses the same tools with the same recommendations. You both enter at the same time. Suddenly, that segment has twice as much competition. Customer acquisition costs spike, prices compress, and margins evaporate. You’re not winning with strategy, you’re just running the same race as everyone else.
They all use similar models and similar agentic systems. All of them train on similar data. When they recommend and search online for which market to enter, which customers to target, or how to price products, they offer averages.
Your competitor gets the same average.
Moreover, if you rely purely on GenAI to make a decision, now your competitor has your playbook. They can use it to predict what you will decide.
The Average Answer Offset
Maybe you bring your own data. You try using RAG (retrieval-augmented generation) to incorporate information from your knowledge base rather than relying solely on the model’s training data. Does that avoid the “average answer” problem?
We don’t know yet. The patterns the models learned before remain baked in. Does that lens still pull toward the average? Does your custom data help you fully escape the convergence trap? It’s unclear. But there is emerging evidence that it does not. The mechanism is different, but the underlying question is the same: do you trust it to do something different than your competitors?
A Third Way
Use GenAI to automate, to write boilerplate code, to transform datasets, to generate reports and execute research. Do the work that’s correct by definition.
But for critical decisions, build an intentional game plan. Decide going in which customers matter, which features to build, how to price, and whether to enter a market. These require judgment beyond what the average of public data suggests. Use GenAI to compress work and then use your newfound free time to decide.
If you’re making decisions based on GenAI recommendations - without push-back - you’re betting on the same decision as your competitor.
About Yakov Shkolnikov:
Yakov is an AI systems architect with 20+ years building production AI, from early CNN systems deployed for radar applications in 2010 to enterprise GenAI platforms today. He has advised both internal and external organizations on AI and data strategy, built teams from the ground up, and led work in edge AI, forecasting, and optimization at Fortune 500 scale. His work has driven enterprise-wide impact across products, operations, and strategy.
Yakov holds a Ph.D. from Princeton and is an IEEE Senior Member.
Follow Yakov’s Substack here.
To remind you of this week’s data concept, enjoy Freedom of Choice by Devo, from the Frictionless Data Spotify playlist.