Recommendation · MoneyHarish Gaggar recommends using the LangGraph framework from the LangChain community as the agent framework for building task-specific nodes, saying they use it for most cases. LangGraph provides a graph-based architecture supporting loops, branching, and state management for complex multi-agent AI workflows.
AI Engineer · Paving the Road to Production: Governed Enterprise AI Agents — Harish Gaggar, Credit Karma
Recommendation · MoneyOne of Replit's actual shipped fixes: a planning-only mode that enforces agent strategization without allowing it to touch the code base.
AI Engineer · Your AI Isn’t the Problem. Your Infrastructure Is. — Jiachen Jiang, Coder
Recommendation · MoneyOne suggestion is to run a premortem with your team assuming one of the three incidents happens to you next month. Either ask people what they actually do, or pretend that it's actually happening and see what people actually do.
AI Engineer · Your AI Isn’t the Problem. Your Infrastructure Is. — Jiachen Jiang, Coder
Recommendation · MoneyOne of Replit's actual shipped fixes: a one-click restore covering full project state, enabling recovery from agent-caused damage.
AI Engineer · Your AI Isn’t the Problem. Your Infrastructure Is. — Jiachen Jiang, Coder
Recommendation · MoneyPaden Gayle recommends that before designing retrieval, you find the person accountable for each data source, and confirm the allowed use and the environment where processing can actually happen. If those answers change, your design changes with them.
AI Engineer · Deploying AI Inside a Bank Is Not an Engineering Problem — Paden Gayle, Applied AI Engineers
Recommendation · MoneyPaden Gayle recommends leaving every discovery conversation with names, decisions, or next conversations. If someone says security will be fine, ask who will actually review the design and when you can meet them. Then close the loop: send each owner a short description of the proposed use and confirm what they need from you. By Friday of discovery week, the people whose decisions the deployment depends on should know the project exists.
AI Engineer · Deploying AI Inside a Bank Is Not an Engineering Problem — Paden Gayle, Applied AI Engineers
Recommendation · MoneyPaden Gayle recommends that for the data decision, you find who owns the source, who can grant access, and who can approve using it in the workflow. Responsibilities may belong to different people, and being able to create a dataset does not settle every question about how the application may use it.
AI Engineer · Deploying AI Inside a Bank Is Not an Engineering Problem — Paden Gayle, Applied AI Engineers
Recommendation · MoneyPaden Gayle recommends bringing a set of questions to a discovery call: where does the data live and who can approve reading it; who has stopped a similar project before and why; which existing controls could cover this workflow and who owns them; where will the production run happen and who are the administrators for that environment; and who signs the relevant risk review and what have they signed before.
AI Engineer · Deploying AI Inside a Bank Is Not an Engineering Problem — Paden Gayle, Applied AI Engineers
Recommendation · MoneyFor the hypothetical document exception assistant, Paden Gayle would put three lines in front of the sponsor — review the documents, draft the exceptions, analyst submits — and then make each line specific: which documents can the analyst access, is that access approved for this processing, what should the model produce (highlighted passage, draft form, or both), and who checks the draft and who submits it.
AI Engineer · Deploying AI Inside a Bank Is Not an Engineering Problem — Paden Gayle, Applied AI Engineers
Recommendation · MoneyPaden Gayle recommends that when a sponsor wants AI for onboarding, you sit with an analyst and watch which systems they open, where they make a judgment, and when they ask for permission. Ask what happens when the document is missing. Write down the actual sequence of the analyst's workflow — that is the beginning of your specification.
AI Engineer · Deploying AI Inside a Bank Is Not an Engineering Problem — Paden Gayle, Applied AI Engineers
Recommendation · MoneyPaden Gayle recommends that for third-party risk, you draw the actual dependency chain — including the model provider, host arrangements, and your role in the operating service — and find out which third-party relationships have already been reviewed and which proposed changes require another decision.
AI Engineer · Deploying AI Inside a Bank Is Not an Engineering Problem — Paden Gayle, Applied AI Engineers
Recommendation · MoneyPaden Gayle recommends that for audit, you ask what someone would need to reconstruct a consequential action: which input was used, which configuration produced the output, what the reviewer saw, and who approved the action. These questions determine the records you need to capture from the beginning.
AI Engineer · Deploying AI Inside a Bank Is Not an Engineering Problem — Paden Gayle, Applied AI Engineers
Recommendation · MoneyPaden Gayle recommends starting at the lowest step that creates value, then collecting evidence from there: are reviewers overriding the drafts? Which failures keep recurring? Can people recover when something breaks? Use those observations to make the next authority decision.
AI Engineer · Deploying AI Inside a Bank Is Not an Engineering Problem — Paden Gayle, Applied AI Engineers
Recommendation · MoneyPaden Gayle asks which increment creates the useful value today — could you draft inside the analyst's current workflow to solve enough of the problem? If so, start there, confirm the existing control really fits, and build the smallest useful version (MVP) with a credible path of approval. He starts with the least authority that solves the problem, then asks what tasks can be proven as evidence to give the system more.
AI Engineer · Deploying AI Inside a Bank Is Not an Engineering Problem — Paden Gayle, Applied AI Engineers
Recommendation · MoneyPaden Gayle recommends proposing a success criteria such as less review time with no increase in missed exceptions, and notes that you still need to agree on how to measure that against the current level.
AI Engineer · Deploying AI Inside a Bank Is Not an Engineering Problem — Paden Gayle, Applied AI Engineers
Recommendation · MoneyPaden Gayle recommends ensuring that by Friday, the people whose decisions the deployment depends on know the project exists; leaving them unaware means scoping lacks a firm starting point.
AI Engineer · Deploying AI Inside a Bank Is Not an Engineering Problem — Paden Gayle, Applied AI Engineers
Recommendation · MoneyPaden Gayle recommends that for model risk, you find a person who can tell you which review applies to the intended use, and ask what a previous model risk submission looked like.
AI Engineer · Deploying AI Inside a Bank Is Not an Engineering Problem — Paden Gayle, Applied AI Engineers
Recommendation · MoneyAlvin advises to stop panicking about raw differences, pull the diffs, and manually review some of them. When reviewing, ask: was the candidate better, was production better, or were they just two different equally acceptable ways to handle an ambiguous input? For HarborByte, the completed diff review confirmed 443 regressions but also found 2463 outright improvements — almost six improvements for every one regression from a prompt half the size.
AI Engineer · How to Actually Tokenmaxx and Save Millions — Alvin Pane, OutRival
Recommendation · MoneyJemima Abu recommends the Modern Web Guidance skill from Chrome, which she likes to use; it ensures agents use the newest HTML and CSS features as opposed to custom JavaScript implementations.
AI Engineer · The Cost of Accessibility in AI — Jemima Abu, CAIS
Recommendation · MoneyJemima Abu advises that when providing prompts to your agent, you should focus on ensuring it uses existing semantic elements rather than custom reimplementations.
AI Engineer · The Cost of Accessibility in AI — Jemima Abu, CAIS