Person
2 episodes · 8 quoted claims
Andrea Tricoli described a budgeting workflow in which a team of AI agents working in sequence produces the first draft of a budget.
“The example I wanted to talk about is the budgeting one, where you can have the first draft of a budget put together by a team of agents.”
Andrea Tricoli said McKinsey benchmarks show organizations often lose between one and four percent of value because contract terms such as discounts, rebates and hourly rates are not reflected in invoices.
“Our benchmarks show that often organizations leave on the table between one and four percent of value in leakage because complex terms, discounts, rebates, or specific hourly rates are not always always reflected in invoices.”
Andrea Tricoli said an AI agent for investor relations can be loaded with a company's past market interactions, public data and peers' interactions so that it asks hard questions investors might raise.
“you can effectively load all your previous interactions with the market, public data, as well as your peers, other companies, interactions with the market”
Andrea Tricoli said security for AI tools in finance follows each company's existing security frameworks, as it did for earlier tools.
“The way we think about it is not, it's very similar to to to the past. The appropriate security needs to be followed.”
Andrea Tricoli says chatbots alone were not sufficient and that splitting strategy work into tasks for different AI agents worked better.
“And we have found that chatbot do not do not are not sufficient for that. What really brings this approach in the right place is splitting the work”
Andrea Tricoli describes a setup with researcher, expert and critic agents, where the critic checks the analysis against sources to catch hallucinations.
“There can be the, uh, researcher agent, the the the the expert, the critic that checks that the analysis was appropriate and and checks against sources”
Andrea Tricoli says the first two examples of generative AI in strategy are complex and not available off the shelf.
“in these first two examples, we are sharing, uh, um, ways to apply GENI to strategy that is rather complex. It's not really off the shelf.”
Andrea Tricoli says AI tools increasingly link claims back to sources, though precise numbers and specific stories should still be cross-referenced.
“Increasingly more, there's, uh, an automated source verification, so many claims, uh, are linked back to the source so that the person can go and check.”