This Week's Term: AI Ensemble - getting several independent answers to the same question, from different AI models or from repeated runs of one model, and then combining them by voting, averaging, or having another model act as a judge that compares them.
The term comes from machine learning, where combining several models into an "ensemble" has been a standard technique since the 1990s. One well-known example is the Netflix Prize, a $1 million competition to improve Netflix's movie recommendations. The leading teams reached their results by blending many models, and the runner-up team even named itself The Ensemble.
An AI ensemble works much like the wisdom of crowds, the subject of this issue's leadership post, Are Right, A Lot: crowd wisdom with humans and AI: several answers combined are often more reliable than any single one. Where you can, use models from different companies rather than asking the same model again and again.
How can you start using an AI ensemble?
No special tool is needed to start. Ask ChatGPT, Claude and Gemini the same question, paste the three answers into one of them, and ask:
Where do these answers agree, where do they disagree, and what would I need to check to know which one is right?
Some tools now run the ensemble for you:
- Perplexity Model Council sends your question to three models at once and shows you where they agree and where they disagree. I reviewed Perplexity earlier in Perplexity for enterprise research.
- Microsoft 365 Copilot Researcher offers a Model Council option: GPT and Claude research in parallel, and the summary highlights where they agree and where they diverge.
- Base44's Compare models builds your next app change with two different models, so you can try both versions and keep the better one.
The same idea is behind this issue's framework post, Ten meetings in an afternoon, which uses several simulated perspectives to test a business decision.
Recommended resource
To dive deeper, watch Mike Tholfsen, a Group Product Manager at Microsoft, walk through the new multi-model capabilities of Researcher in Microsoft 365 Copilot.
Your action step
Pick one question you are working on this week and ask it in three models from different companies. Use the prompt above to compare the three answers, and check the point where they disagree before you rely on any of them.
If you are deciding how your teams should combine and compare AI models, that is the kind of question I work through in AI strategy advisory engagements and in sessions as an AI keynote speaker across Europe.