Building in public
Bill and WilliamFrank build agentic AI rails for wealth management — not a chatbot bolted onto a legacy stack, but the operating layer underneath research, planning, portfolio work, and client service. This page is a running account of what that build actually involves, for the people who might want to work on it.
Four kinds of work are in progress at any given time. None of them are secret; none of them are finished.
Grounded content generation. Producing advisor- and client-facing writing that is actually anchored to an evidence base — market data, academic research, a firm's own methodology — rather than pattern-matched prose that sounds right and checks nothing. The hard part is not the drafting. It is building the corpus and the citation discipline underneath it.
Lead and client agents. Systems that can hold a real conversation with a prospect or client, know what they do not know, and hand off cleanly to a human the moment the conversation needs judgment rather than information. Streaming responses, session state, a compliant record of every exchange, a human-in-the-loop handoff that actually works under load — and fee structure and standard-of-care disclosure surfaced plainly and automatically, not just once at onboarding.
Portfolio and tax-aware automation. Turning the spreadsheet logic behind personalized indexing, tax-loss harvesting, and rule-based rebalancing into an agent-run engine wired to live data feeds — index constituents, custodial positions, order generation — with a human approving every trade before it goes out.
Operational rails. Custodial and portfolio data flowing into a rules layer that turns drift, cash balances, RMD deadlines, and harvest windows into tracked tasks with owners — the plumbing that keeps a growing book of households from depending on any one person's memory.
Small team, agent-augmented, building from scratch — no licensed platform underneath this, no legacy codebase to work around. Every build is deterministic where it needs to be, versioned, explainable, and human-gated on anything client-facing or money-moving. That is a design constraint, not a slogan: a fiduciary business cannot ship an agent that cannot show its work.
The wealth management industry is short on advisors and getting shorter — the case is laid out in Agentic wealth. The leverage problem is the same reason it is a good time to build the underlying rails: an industry that has to do more with the people it has is an industry where good infrastructure compounds fast.
We are not running a job board here, and this is not a pitch. If the problems above are the kind you think about anyway — agent orchestration, evidence-grounded generation, compliance-aware automation, the plumbing under a fiduciary business — send a note through contact and mention what you are working on now. Thought leadership is the primary point of this site; this page exists because some of the people who read the essays are the people we would want to build with.