Strategy Requires Choice
AI can give us more possibilities than we could ever explore alone. That does not necessarily make strategy easier. It changes where the value sits: in context, connections and the judgment to decide what deserves to move forward.
Category:
Strategy & AI
Author:
Ana Andersen
Read:
4 min
Location:
Norway
Date:
More possibilities do not automatically create better decisions
One thing has changed dramatically in the way I work over the last few years. The cost of exploring an idea has fallen. I use AI throughout my workflow: to interrogate briefs, structure research, analyse information, look for patterns, challenge assumptions, compare scenarios, explore positioning and concepts, work with copy and visuals, and move much faster between thinking and making. I can explore ten directions in the time it previously took to explore two. That is incredibly useful. But the more I use these tools, the clearer something else becomes: The difficult part is no longer generating another option. It is deciding which option deserves to survive. Strategy has never really been about having the most ideas. It is about making choices. What are we actually trying to achieve? What should we prioritise? What are we willing not to do? Which opportunity fits this business, these people and this moment? AI can help me investigate those questions. It cannot take responsibility for the answer. A good answer still needs a context Two organisations can ask exactly the same question and need completely different answers. Their customers are different. Their capabilities are different. Their culture, competitive position, resources and appetite for risk are different. And perhaps most importantly, the people expected to make the strategy real are different. This is where I think it becomes easy to be overly impressed by output. AI can produce something polished, logical and entirely plausible in seconds. But plausible is not the same as useful. A recommendation only becomes interesting when it survives contact with the actual organisation around it. What information went into the answer? What assumptions are hiding inside it? What does the model not know? What would have to be true for this to work? What happens elsewhere in the system if we make this choice? And who is responsible for deciding? One of the questions I now ask AI most often is not: What is the answer? It is: What have I failed to consider? That relationship with the tool is much more useful to me.
AI can increase creativity - and still narrow the landscape
Recent research is starting to show an interesting tension. AI can be remarkably good at generating ideas. A 2026 study of new product ideation found that LLM-generated concepts scored higher on average purchase intent than human-generated ideas and were substantially more likely to appear among the highest-rated concepts. But there was another side to the result: the AI-generated ideas were less diverse as a group. That tension appears elsewhere too. A large-scale comparison published in Nature Human Behaviour, involving more than 9,000 people and over 215,000 LLM responses, found that humans were slightly more creative on average, but the bigger difference appeared at the extremes: human responses showed greater variation and more of the exceptionally creative outputs. Another 2025 study comparing human and GPT-4 writing found that each additional human-written text contributed more new ideas to the overall pool than an additional AI-generated text. The difference became larger as more outputs were added. And experiments with AI-supported brainstorming have found a similar trade-off: individuals can produce stronger ideas with AI assistance while the ideas across people become more similar to one another. This does not mean AI is not creative. It means quality, originality and diversity are not the same thing. And that distinction matters for innovation. Innovation often happens in the connections This is very close to what I keep noticing in my own work. AI can generate an extraordinary number of directions. It can make surprising associations and take a thought somewhere I had not considered. But if I let it lead the process for too long, the thinking often starts to converge. The language becomes familiar. The categories become familiar. The logic stays surprisingly close to the frame it was given. Some of the strongest ideas I have worked with have come from somewhere else entirely: An observation about customer behaviour changes how I think about a brand. A principle from UX reshapes a commercial problem. Something I learned through sustainability or systems thinking changes how I look at marketing. An idea from one industry suddenly becomes relevant in another that appears to have almost nothing to do with it. None of the individual pieces are necessarily new. The value comes from the connection between them. I think this is one of the most interesting parts of innovation. Innovation is not always about inventing something the world has never seen before. Sometimes it comes from seeing two things that have been kept separate and realising that, together, they change the way the problem can be understood. And strangely, the more I use AI, the more valuable my own experience starts to feel. Every project, discipline, industry, mistake and observation adds another piece I can connect to something else. AI gives me access to a much larger landscape. Experience helps me notice which parts of that landscape might actually belong together.
The workflow matters
There is another implication here. If AI-supported thinking can converge, then the answer is not necessarily to use less AI. It may be to use it differently. A recent meta-analysis of 19 studies on human-AI co-creation found a small but statistically significant homogenising effect from generative AI use, with the effect varying depending on the type of creative task. For me, that makes the design of the workflow increasingly important. I do not want AI to simply make the work I already know how to do faster. I want it to widen the space I am thinking inside. So instead of only asking it to generate answers, I increasingly use it to create friction: Give me the strongest argument against this direction. What would a customer see that I am missing? Which assumptions am I treating as facts? Take this idea outside the conventions of the category. What changes if we view the problem from operations instead of marketing? What would have to be true for this strategy to fail? What adjacent industries solve a similar problem differently? Show me three interpretations that contradict each other. That is where I find the tool most useful. Not when it removes thinking. When it gives me more to think with. More output is not the advantage As these tools improve, competent output will become easier to produce. That changes where differentiation lives. Knowing how to generate twenty strategic directions matters less when almost everyone can do it. Knowing why nineteen of them are wrong for this particular organisation becomes more valuable. And sometimes the twentieth direction is still too obvious. That is when we need to leave the frame entirely. For me, this is where strategy, creativity and AI increasingly meet. AI gives me speed, range and a much larger number of possibilities. But the work I value most still happens in the space between those possibilities: Seeing the pattern. Making a connection that was not obvious. Understanding why it matters here. Knowing what to ignore. Making a trade-off. And eventually choosing what is worth building. When generating possibilities becomes cheap, judgment becomes more valuable. I am starting to think our ability to make meaningful connections does too. ________________ Research Wang et al. (2025/2026), A large-scale comparison of divergent creativity in humans and large language models, Nature Human Behaviour. Meincke, Nave & Terwiesch (2025), ChatGPT decreases idea diversity in brainstorming, Nature Human Behaviour. Homogenizing effect of large language models on creative diversity: An empirical comparison of human and ChatGPT writing (2025), Computers in Human Behavior: Artificial Humans. Terwiesch et al. (2026), Artificial intelligence and its impact on creativity and diversity: An empirical study of large language model-generated product ideas.
