The most common question in strategy and research circles right now is some version of: what does AI change about how we do this work? It is a reasonable question, and it deserves a more precise answer than the field has generally produced.
Here is what changes. The mechanical work of research, gathering sources, summarising documents, identifying surface-level patterns across large datasets, producing first drafts of structured analysis, can now be done faster and at lower cost than at any previous point. What took a junior analyst three days can, with the right tools and prompting, be compressed into hours. This is a genuine and significant shift.
The scale of adoption reflects this. According to McKinsey's State of AI 2025 report, 88% of organisations are now regularly using AI in at least one business function, up from 78% just a year earlier. In financial services specifically, 58% of institutions directly attribute revenue growth to AI. The technology has moved from experimental to operational across much of the economy.
Here is what does not change. The judgment calls that make research valuable, what to look for in the first place, which sources to trust, how to weigh conflicting evidence, what the data actually means in the context of a specific decision, remain stubbornly human. Not because AI is incapable of producing outputs that look like judgment, but because the accountability for those judgments, and the contextual understanding that makes them reliable, cannot yet be delegated.
The distinction matters because the temptation, when a new tool dramatically accelerates output, is to treat speed as a proxy for quality. A competitive analysis produced in two hours feels like it must be less rigorous than one produced in two weeks. Sometimes it is. Often, the two-hour version is simply less padded, and the two-week version contained ten days of activity that did not improve the underlying thinking.
What AI genuinely demands of researchers and strategists is a higher standard at the judgment layer. If the mechanical work is handled, the value has to come from somewhere else: from the quality of the questions being asked, the precision of the research frame, the ability to evaluate and critique AI-generated output rather than accept it, and the capacity to translate findings into recommendations that are specific, defensible, and actionable.
These are the skills that distinguished good research from mediocre research before AI arrived. The difference is that they are now the primary differentiator, the mechanical work having been largely commoditised.
Research is not going away. The version of it that required volume rather than judgment largely already has.