Person
1 episodes · 8 quoted claims
The guests identify five emerging roles for AI in strategy: researcher, interpreter, thought partner, simulator and communicator.
“We see the researcher, uh, the interpreter, the thought partner, and the simulator, and then communicator as these five emerging roles.”
The worked example follows a European financial services company looking to expand into Vietnam, including possibly acquiring a local company.
“we consider European player financial services, which is, um, eager to expand in Asia or in emerging markets, and in this case, in in Vietnam.”
Andrea Tricoli describes AI tools that read annual reports, news, patents and marketing material to build a graph of companies, then filter it into likely acquisition targets.
“By by going through all these data points, the AI is able to create a graph of all these companies.”
Bruce Delteil found the AI-generated Vietnam target list was reliable but sometimes confused companies with their parent companies.
“Sometimes the the companies are actually, um, referring to a different parent company.”
Bruce Delteil says foreign ownership limits are the most important regulatory risk for entering Vietnamese banking, yet the AI tool did not flag it.
“I I I take an example, um, regulatory risk, the foreign, uh, ownership limit is the most important one.”
Bruce Delteil says attracting and retaining Vietnam's digital talent is a key issue for any institution entering the market.
“Vietnam is known to have a lot of great digital talent, attracting and retaining them becomes very important.”
Bruce Delteil says that when he started experimenting with AI, he lacked knowledge of what the model was trained for and relied on data scientists to explain it.
“From my own experience, what was very much missing for me at the beginning, when I when we started leveraging and experimenting, is I didn't know what the model was trained for.”
Bruce Delteil suggests AI could eventually assess how successfully a company could integrate an acquisition target, using its historical data, but says this had not yet been presented.
“you could think on our M and A example of a case where you start asking the model to what extent you can be successful at integrating a particular target”