Strategic Acquisition Brief · Confidential
Anthropic is winning the enterprise — and financial services is the beachhead vertical. Every bank, insurer, and platform deploying Claude agents converges on one question: can the agent's money answer be warranted? Today no lab can say yes. MaxiFi is the deterministic engine that computes each household's lifetime plan — provably, reproducibly, auditable — and the only foundation on which an accuracy guarantee can be underwritten.
The growth case →Request the briefingClaude for Financial Services put Anthropic's flag in the vertical where correctness is priced: banks and insurers buying agents must answer their examiners for “the reliability and accuracy of the AI model,” and under Reg Notice 24-09 the obligation reaches embedded third-party tools. Every FS deal Anthropic wins ships that question to a compliance department. The lab that can hand its enterprise customers a deterministic, auditable computation layer under the conversation wins the RFP — and the category's own tests, CBS's among them, show why no model alone can: the category estimates. That is not a Claude weakness; it is the physics of language models. It is also the gap one asset closes.
MaxiFi solves the lifetime plan for a household's facts and assumptions — Social Security, full federal and state tax code, Roth sequencing, withdrawal order, insurance sizing — same inputs, same answer, every time, with an audit trail an examiner can walk. Thirty years of Laurence Kotlikoff's economics, taught with at MIT Sloan by Nobel laureate Robert Merton. Un-scrapeable, patented, and — because the math is exact — the only foundation for a bounded Accuracy Guarantee.
That changes what the claim is. Backed by the pedigree — thirty years of Laurence Kotlikoff's economics, taught with at MIT Sloan by Nobel laureate Robert Merton — and by the reproducible computations themselves, the accuracy claim stops being puffery and becomes a substantiated statement of fact. And determinism unlocks what a claim alone never could: a bounded Accuracy Guarantee with a defined remedy — the play that built TurboTax's franchise, never before available in planning, insurable only because the math is exact.
The substantiation regime that polices financial advertising — FINRA 2210's fair-and-not-misleading standard, FTC substantiation doctrine — protects this claim. Rivals can run vague accuracy language; what they cannot run is your claim: the specific, falsifiable, guaranteed one. Copying it without the engine is a false claim regulators, NAD panels, and Lanham Act suits will punish.
“Claude agents, with computed and warrantable money math” — the differentiator no rival lab can match in the sales cycle that decides the vertical.
Your enterprise customers carry the exam; MaxiFi is the answer they hand the examiner — deterministic outputs, versioned law tables, audit trail.
The assistant whose money answers are computed and guaranteed — trust as a product property on the hardest consumer domain.
MaxiFi generates verified ground truth for training and remains the runtime verifier — the economics in the weights, the warranty on the output.
| OpenAI | Anthropic + MaxiFi | ||
|---|---|---|---|
| The agent's money answer | Estimated | Estimated | Computed — and guaranteeable |
| The enterprise exam question | Open | Open | Answered: deterministic core, audit trail |
| Can rivals copy it? | The words, not the proof | The words, not the proof | One engine exists; imitation is a false claim |
One quarter of the computed core inside the FS agent stack — RFP wins, examiner answers, the guarantee — answers what no forecast can. Owning MaxiFi is the exclusive right to run that play, and to deny it to OpenAI and Google permanently. It is a revenue line, not a legal reserve.
MaxiFi (Economic Security Planning, Inc.) uses consumption smoothing and dynamic programming to compute the single, mathematically optimal lifetime plan — solving simultaneously across Social Security strategy, federal and state taxes, Roth-conversion sequencing, withdrawal order, insurance sizing, and upside investing. For a household's facts and assumptions it solves — not guesses: same inputs, same answer, every time, with an audit trail.
Prof. Laurence Kotlikoff — William Fairfield Warren Professor at Boston University; Harvard Ph.D.; former Senior Economist, President's Council of Economic Advisers; named by The Economist among the 25 most influential economists.
Taught with at MIT Sloan by Nobel laureate Robert Merton as an “outstanding science-based lifecycle and retirement management platform” (Merton does not endorse products); featured in Bankrate's “Best financial planning software of 2025” roundup. The economics trace to Nobel-recognized lifecycle work.
Patented algorithms and thirty years of continuously maintained federal/state tax, Social Security, and benefit rules with a validation record — exactly the IP a language model cannot reverse-engineer and a build team cannot shortcut.
Larry Kotlikoff intends to stay on with the acquirer — to integrate the engine, validate the training and guarantee programs, and continue as spokesperson. The acquirer buys the engine and keeps the economist who built it.
Claude keeps the reasoning and the relationship; MaxiFi computes the number underneath and stamps it with the audit trail. SCF-perturbed households become verified training pairs; at runtime the engine verifies what the agent speaks. The conversation stays Anthropic's; the answer becomes warrantable. Larry Kotlikoff stays on to integrate and as spokesperson.
This is the first FINRA examination cycle to treat generative AI as a standalone topic, and banking supervisors are asking the same question of every deployed agent. The lab whose stack ships with a deterministic, auditable computation layer converts its customers' examination liability into an examination exhibit — and makes that the reason to choose Claude.
And the engine ships with the architecture that keeps the floor solid under an advertised claim: assumptions and law-table version disclosed on every output, customer input attestation, versioned rule tables with re-run notices on law changes, and the Accuracy Guarantee's defined remedy. The audit trail proves each customer was told exactly what was — and wasn't — promised.
We price the asset on the growth case above. The defense beneath it is a term of the deal, not the deal — and, like the claim itself, it is denied to every competitor the day it is yours.
A frontier model's retirement “smile” ran 13% too low in each of a real household's 40 remaining years against MaxiFi's computed path — dated, dollar-specific, reproducible.
Four frontier AIs sized the same father's coverage at $1.3M, $1.4M, and $3.8M — against MaxiFi's internally consistent $2.09M. Every shortcut the AIs used is programmable — and wrong.
One retirement question, three frontier engines, three different verdicts — with MIT's Andrew Lo noting these tools carry no best-interest duty. The category estimates; the divergence is the proof.
The tests publish to 145,000+ subscribers and counting — credibility no rival in the category can match, and it conveys with the acquisition.
Larry built this over thirty years for the households your agents will advise. We are running a deliberately narrow process for the home where computed correctness does the most good — and where safety is already the brand. That is why this conversation.
The next step: a 30-minute briefing — MaxiFi solves a real household's lifetime plan, live, while a frontier model is asked to match it. The gap is the thesis; the funnel is the price.
Michael Kane, Ph.D., J.D. · Managing Partner, Kane & Company · FINRA / SEC / SIPC–Registered Investment Bank
Commerce@kaneco.com · 310-441-5263 · Representing Economic Security Planning, Inc.