Recommendation · MoneyAnekha Sokhal recommends recognizing that when distributing an internal tool or dashboard, adoption is a UI problem, not a data problem — getting someone new to incorporate the tool into their workflow depends on how they interact with it. She also recommends designing data products to push information to users rather than requiring them to pull, citing how Spotify and Instagram hook users by surfacing relevant content they didn't know to ask for.
Endorsed heuristics from the speaker's experience; no measured outcomes cited.
Latent Space · You are the Team — Anekha Sokhal, Moshi
Recommendation · MoneyFelippe recommends that the gates you design need to be fit for purpose, and you need to know the use case really well to put gates where it matters.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Recommendation · MoneyBuild the feedback mechanism on purpose and keep judgment over making sure there is consequential decision-making on the fact that an input may change your whole analysis; you need to be on top of making sure those changes make sense.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Recommendation · MoneyFrancesco says going back and forth with your model really makes a difference, and encourages digging deeper, challenging your assumptions, and answering questions you may not have known you needed to ask yourself when working with the harness.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Recommendation · MoneyFrancesco recommends making sure that any judgment call or disambiguation that is important is captured in gates, while trivial ones can be left to the agent.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Recommendation · MoneyFrancesco advises starting with a chat, starting with a problem, and describing what you want to achieve when getting help from an AI agent.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Recommendation · MoneyFrancesco recommends building the feedback mechanism on purpose, with judgment over making sure there is consequential decision-making when an input may change your whole analysis, and being on top of making sure those changes make sense.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Recommendation · MoneyYou don't need to redo your plan or challenge every decision, but you do want a separate point of reference; something that reviews and gives a rough answer or challenges what looks wrong is extremely helpful.
endorsed advice
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Creator perspective · MoneyAnekha Sokhal describes two practices for handling conflicting information: when major outlets like Reuters, Bloomberg, and the Wall Street Journal disagree (for example on ship or plane counts), the disagreement itself is a data point that may indicate an opportunity because no one knows the truth. And when presenting data-derived claims to users, Moshi attaches a confidence score that contextualizes the claim against everything else happening around the conversation.
Described as Moshi's practice; the ship/plane example is hypothetical in the source.
Latent Space · You are the Team — Anekha Sokhal, Moshi
Principle · MoneyYou always have another point of reference outside of your data — you have a lot of sources of information, you have the internet. Enhance the signal you're getting by looking at resources that are not the ones you're using, to avoid the feedback loop where you only have the information you have available and everything looks right to you. Always try to get a little bit creative about other points of reference you could use to help you make a better decision or to know if something is not quite right.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Principle · MoneyFrom experience and repetition across many use cases, the speakers believe you need to build a harness that verifies things are going in the right direction, not just that things look correct.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Principle · MoneyFrancesco Galletta says the bottleneck in producing financial analysis is no longer speed but how the output connects with people — trust — and that trust must be built into AI systems from the ground up, especially for high-stakes topics like money, which he calls one of the highest-stakes topics for AI systems.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Principle · MoneyFrancesco Galletta recommends providing evidence — showing the receipts — so decisions can be traced back to what led to them, including what information came through and whether things were tested, put front and center.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Principle · MoneyLooking at other implementations of financial analysts in the industry, traceability goes down to the dataset — you know which connector or supplier of data gave you the information, e.g., that Apple had a certain price at a certain point of the day when trading or investing.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Principle · MoneyYou don't need a lot of effort, you don't need to redo your plan entirely or challenge every decision you've made, but you do want to have a separate point of reference. Just having something review and give you a rough answer, or some challenge if something doesn't look right, is extremely helpful.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Principle · MoneyFrancesco Galletta says that even if you are running 100 agents, you are still accountable for what they do; you should keep a good-enough grasp of what the machine is doing, because going full black box mode kills your accountability — though reviewing every change line by line is also unrealistic, as nobody has ever done that.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Creator perspective · MoneyFrancesco argues that asking an AI to build something without knowing the principles and decisions behind it is like asking Claude to build a one-million-dollar company: you are relying on randomness of something good happening by accident.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Principle · MoneyThe fourth design principle is accountability: even if you're running 100 agents, you are still accountable for the results of what they do.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Creator perspective · MoneyThe hosts note that you don't want to see all the logs and the math behind a single number, but just having the trace to figure out 'I can follow back this decision' increases confidence for us humans quite a bit.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
Creator perspective · MoneyFrancesco warns against going full black box mode with an AI system, because then your accountability is gone.
AI Engineer · Stop Chatting With Spreadsheets — Francesco Galletta & Felippe Felisola Caso, Monzo
