The arrival of Deepseek increases the availability of generative AI for those undertaking research and analysis. According to KPMG the growth in use of generative AI products is exponential as organisations embrace it to improve their productivity and efficiency 1. How will the use of generative AI impact the role of the analyst?
There are two opposite views that help us find the middle ground. Some see generative AI as replacing most human creativity, innovation and decision making. The implication is catastrophic social and economic change including high unemployment.
The opposing view is that generative AI is a long way from understanding and conceptualising human intelligence. Humanity does not understand its own intelligence very well at present, so generative AI creators do not have an overall sense of direction and purpose. They can build larger and faster learning and analysis from ever bigger datasets, but the results will soon reach a conceptual incoherence . These AI models then stop being useful to real world decision making.
AI can do analysis, but can it make complex and adaptive decisions? Long before this debate emerged, there was an historical deliberation about the role of analysis in strategic decision making. Having access to data, employees, and consultants to analyse it, does not guarantee better policy and decision-making. In a seminal article in the Harvard Business Review in 1994, Henry Mintzberg articulated this, concluding that analysis alone was not the answer 2. Creative strategists are needed.
The US economist, Stephanie Kelton illustrates one recent example of this analysis-strategy paradox 3. She finds that Deepseek can develop an impressive model of analysis in a few seconds. This is analysis that takes a highly trained economist many hours, if not days, to design. Her example is a model to forecast the impact of 25% Trumpian import tariffs on the Canadian economy. In seconds, Deepseek does hours of analytical work previously done by an economist.
The AI result is a robust demonstration that predicts a quantitative drop in annual GDP. Impressive work that reduces the need for an expensive economic analyst. Economists are some of the best paid graduates. Kelton reminds us that economic models often predict within ranges of error. Analysis does not provide certainty, but probable outcomes and possible scenarios. The Deepseek model says that Canadian economic growth will be impacted by a reduction in annual GDP of between 2% and 8%. What can be done in response? The move from analysis to strategy is deciding what actions policy makers should take in an uncertain world.
Artificial Intelligence, in all its forms, is impacting analysis. Is there a future for those employed as an analyst? What is analysis for? And does it lead to better decision making? These are topics explored in research by Cao, et al, in the Journal of Financial Economics 4. They conclude that rather than replacing analysts, human roles must evolve towards the tasks that AI is less good at. The human-AI interface is critical.
During the 2008 Financial Crisis, historical knowledge and data were not necessarily a good guide to managing new and sudden events. Likewise, during the Covid-19 pandemic, there were clear public health analytics showing the exponential spread of the virus, but countries had differing policy responses and resulting fatality rates.
Artificial Intelligence is doing superfast analysis with a vast range of historical information and data, and this data-based picture is bigger and better than at any time in history. It is still growing in depth, but our understanding of complex social systems implies an excellent historical analysis is not always a continuing good guide to the future. The same dilemma is true of some physical systems like weather forecasting. AI perceives complex pattern changes in a manner that humans never can. Will it know the best way to act in response to such pattern changes? The jury is still out on that next step.
Those working in analytical jobs feel vulnerable. The BBC has recently reported how employees are using AI against their own employer’s regulations and advice 5. The attraction for workers is that it speeds up more tedious aspects of information searches and analysis. It can help to get ahead with tasks and to appear more efficient. This ignores risks about cybersecurity and potential feedback automatically going into the AI provider’s data and algorithms.
The ability of AI to investigate big data is revolutionizing the daily routine of those in analytical roles. There are fears AI can replace analysts. Our understanding of what analysts do, and how they do it, is changed by AI, but there is much more to be done to understand the impact on employment. Organisations are trying to use the human-AI relationship to be more effective and to improve their systems of decision making, policy development and strategy. The learning is just beginning.
References/Footnotes
- KPMG Generative AI Survey https://kpmg.com/us/en/media/news/kpmg-generative-ai-2023.html ↩︎
- Mintzberg, H (1994) The fall and rise of strategic planning. Harvard Business Review 72 (1), 107-114
https://libroweb.alfaomega.com.mx/book/385/free/data/Materiales/Capitulo01/TheFallAndRiseOfStrategicPlanning.pdf ↩︎ - Kelton, S. (2025) The Impact of 25% Tariffs on Canadian GDP: The Bank of Canada vs Deepseek. 27th January. https://stephaniekelton.substack.com/p/the-impact-of-25-tariffs-on-canadian ↩︎
- Cao, S., Jiang, W., Wang, J., & Yang, B. (2024). From man vs. machine to man+ machine: The art and AI of stock analyses. Journal of Financial Economics, 160, 103910.
https://doi.org/10.1016/j.jfineco.2024.103910 ↩︎ - McManus, S ( 2025) Why employees smuggle AI into work – BBC News 4th February. https://www.bbc.co.uk/news/articles/cn7rx05xg2go ↩︎