One of the most popular sessions from the most recent edition of FP&A Week was AI & Finance: How AI Will Reshape the Finance Organization.
The full session is available on demand here and well worth watching, but we’ve summarized some key takeaways here:
Treat AI outputs as opinions
AI doesn’t know everything, and it can’t answer all your questions. But opinions are valuable, which is why Anders suggests treating AI as another source of opinion while making decisions. There was some interesting discussion on the ‘wisdom of the crowd.’
Craig mentioned that the toughest challenge he has faced is related to security, especially in a component-based architecture. Securing and authorizing data across various systems can be complex, leading to potential issues like authentication problems that can delay production.
Early engagement with InfoSec
Organizations should involve their information security teams earlier in the process of implementing AI-based solutions to avoid late-stage challenges, especially around passwords, encryption, and tokenization.
Building trust in emerging technologies
Anders emphasized that while emerging technologies offer a lot of promise, it's essential to test and verify their outputs. For instance, when he tested GPT-4, it provided unexpected advice on creating a fake invoice, highlighting the importance of vigilance while we explore uncharted waters.
Advice on adoption
Anders advises organizations to embrace the spirit of experimentation: identify problems and look for ways to innovate. At this stage, it’s OK to make mistakes.
- Vision and strategy
Craig believes in setting a clear vision of what success looks like and then establishing resolutions or actions to achieve that vision. This approach should be embedded in the organization's culture and strategy.
Upskilling and buy-in
It's vital for leaders to lead by example when it comes to training: showing commitment to learning will incentivize their teams to follow. Also, it's essential to communicate the benefits of new technologies, focusing on how they can enhance roles, improve efficiency, and reduce tedious tasks.
Moving from POC to Reality
Operationalizing AI can be difficult. To successfully manage the transition from a proof of concept (POC) to a fully-fledged implementation, organizations should:
- Define clear use cases
- Focus on a few critical outcomes
- Understand that AI outputs are ‘opinions’ and should be treated as such
You can catch up on this session, as well as all the others from FP&A Week, by visiting this page.