Finance Use Cases
๐๐ง ๐ ๐ฉ๐ข๐ฅ ๐ฌ๐ฏ๐ฐ๐ธ๐ฏ ๐ธ๐ฉ๐ข๐ต ๐ฆ๐ญ๐ฆ๐ค๐ต๐ณ๐ช๐ค๐ช๐ต๐บ ๐ช๐ด,
๐'๐ฅ ๐ฉ๐ข๐ท๐ฆ ๐ต๐ข๐ฌ๐ฆ๐ฏ ๐ข ๐ด๐ต๐ฆ๐ฑ, ๐จ๐ฐ๐ฏ๐ฆ ๐ฐ๐ถ๐ต ๐ช๐ฏ๐ต๐ฐ ๐ต๐ฉ๐ฆ ๐ด๐ต๐ณ๐ฆ๐ฆ๐ต,
๐๐ฏ๐ต๐ฆ๐ณ๐ฆ๐ฅ ๐ข ๐ฑ๐ฉ๐ฐ๐ฏ๐ฆ ๐ฃ๐ฐ๐ฐ๐ต๐ฉ ๐ข๐ฏ๐ฅ ๐ฅ๐ช๐ข๐ญ๐ฆ๐ฅ ๐บ๐ฐ๐ถ๐ณ ๐ฏ๐ถ๐ฎ๐ฃ๐ฆ๐ณ,
๐๐ฏ๐ฅ ๐ฉ๐ฆ๐ข๐ณ๐ฅ ๐บ๐ฐ๐ถ๐ณ ๐ท๐ฐ๐ช๐ค๐ฆ, ๐ท๐ฐ๐ช๐ค๐ฆ, ๐ท๐ฐ๐ช๐ค๐ฆ...
๐๐ถ๐ต ๐ ๐ฅ๐ฐ๐ฏ'๐ต ๐ฌ๐ฏ๐ฐ๐ธ ๐ฉ๐ฐ๐ธ ๐ต๐ฉ๐ฆ ๐ด๐ช๐จ๐ฏ๐ข๐ญ ๐ธ๐ฐ๐ณ๐ฌ๐ด,
๐ ๐ฅ๐ฐ๐ฏ'๐ต ๐ถ๐ฏ๐ฅ๐ฆ๐ณ๐ด๐ต๐ข๐ฏ๐ฅ ๐ต๐ฆ๐ญ๐ฆ๐ค๐ฐ๐ฎ๐ฎ๐ถ๐ฏ๐ช๐ค๐ข๐ต๐ช๐ฐ๐ฏ๐ด,
๐ ๐ฅ๐ฐ๐ฏ'๐ต ๐ฌ๐ฏ๐ฐ๐ธ ๐ธ๐ฉ๐ฐ ๐ญ๐ข๐ช๐ฅ ๐ต๐ฉ๐ฆ ๐ฑ๐ฉ๐ฐ๐ฏ๐ฆ ๐ค๐ข๐ฃ๐ญ๐ฆ,
๐'๐ญ๐ญ ๐ฑ๐ณ๐ฐ๐ฃ๐ข๐ฃ๐ญ๐บ ๐ฏ๐ฆ๐ท๐ฆ๐ณ ๐ฉ๐ฆ๐ข๐ณ ๐บ๐ฐ๐ถ, ๐บ๐ฐ๐ถ, ๐บ๐ฐ๐ถ...
Over 40 years ago, the Russian rock band Akvarium sang this and explained ๐๐ต๐ ๐๐ต๐ฒ ๐ฎ๐ฑ๐ผ๐ฝ๐๐ถ๐ผ๐ป ๐ผ๐ณ ๐๐/๐ ๐ ๐ถ๐ป ๐๐ต๐ฒ ๐๐ถ๐ป๐ฎ๐ป๐ฐ๐ฒ ๐๐ฒ๐ฝ๐ฎ๐ฟ๐๐บ๐ฒ๐ป๐ ๐ถ๐ ๐๐ผ ๐ฑ๐ถ๐ณ๐ณ๐ถ๐ฐ๐๐น๐.
Finance professionals have a distinct mindset. Even though theyโve worked in digital environments for over a century and are fluent in numbers, two key obstacles stand in the way of AI/ML integration:
๐ญ. ๐ฃ๐ฟ๐ผ๐ฏ๐ฎ๐ฏ๐ถ๐น๐ถ๐๐๐ถ๐ฐ ๐ง๐ต๐ถ๐ป๐ธ๐ถ๐ป๐ด ๐๐ฒ๐ฒ๐น๐ ๐๐น๐ถ๐ฒ๐ป
Finance is built on ๐ฝ๐ฟ๐ฒ๐ฐ๐ถ๐๐ถ๐ผ๐ป. Professionals are used to dealing with exact figures from the past and expect the same for the future (budgets, forecasts). Concepts like confidence intervals or probabilistic outcomes are fundamentally foreign to them.
Every machine learning model inherently includes ๐๐ป๐ฐ๐ฒ๐ฟ๐๐ฎ๐ถ๐ป๐๐: results are based on probability, not certainty. To a finance manager, an AI-generated number isnโt โthe number.โ So, how should they use it? How should they interpret a fuzzy forecast? Can they trust it? How to ๐ฏ๐ฒ๐น๐ถ๐ฒ๐๐ฒ in the result?ย (Do you see this fuzzy Kandinsky picture - do you trust these blurred circles?)
๐ฎ. ๐๐
๐ฝ๐น๐ฎ๐ถ๐ป๐ฎ๐ฏ๐ถ๐น๐ถ๐๐ ๐ฎ๐ป๐ฑ ๐๐ฐ๐ฐ๐ผ๐๐ป๐๐ฎ๐ฏ๐ถ๐น๐ถ๐๐ ๐ ๐ฎ๐๐๐ฒ๐ฟ๐ (๐ ๐๐ผ๐)
Financial models are simple: +, -, %, and the occasional *.
MS Excel is the perfect tool: it keeps everything traceable.
If a CFO asks why a number is what it is, a financial analyst can walk through every formula, cell by cell and explain the exact reason for every figure on the dashboard. Everything can be traced back.
But ML algorithms are significantly more complex. Even the data scientist may struggle to explain how certain weights or features led to a specific outcome. When the result is wrong - and letโs face it, no AI is perfect - then why and where was the error exactly? Perhaps nobody in the company can give an exact answer. Then, whoโs responsible for this mistake? Somebody should be, and the CFO assumes that this is the finance manager who incorrectly used AI software. So they are not ready to be ๐ฎ๐ฐ๐ฐ๐ผ๐๐ป๐๐ฎ๐ฏ๐น๐ฒ for what they have no understanding.
(Do you understand what's there in the Kandinsky picture? Could you then recommend it to your CFO?)
This is the tension. Until we bridge the cultural and conceptual gap between probabilistic models and the deterministic mindset of finance, adoption will remain slow, not for technical reasons, but for human ones.
However, we have a numerous of examples of the successful adoption of AI in Finance. Check them here:
P&L forecast
...