I've started NESTPILOT — AI-native operating tools for the small end of financial services

After 19 years at BNY Mellon — leading cloud and domain architecture and helping build an award-winning advisory platform — I've gone out on my own. NESTPILOT LLC is a Florida-based company building AI-native operating tools for the small end of financial services: solo CPAs, small advisory firms, independent retirement planners, and the households they serve.

This post is the public version of the conversation I've been having privately for the last few months. Here's the thesis, what NESTPILOT is, what we're shipping, and what I'm asking for.

The thesis: the small end of financial services gets the worst tools

Two decades inside one of the largest custodians and advisory platforms in the country taught me one uncomfortable thing: the tooling gap in our industry runs in the opposite direction most people assume.

Enterprise gets the white-glove software. Consumer gets the polished apps. The middle — solo practitioners, small firms, and the households who can't afford a $5,000/year advisor — gets the leftovers.

• The solo CPA running 200 returns through ATX or Drake is patching together the same document-chase workflow she ran in 2010, except now her clients send PDFs and screenshots over five different channels instead of two.
• The independent retirement planner serving 30 households is paying for eMoney or MoneyGuidePro, both designed for advisor firms, not solo practitioners.
• The household saving in a 401(k) and trying to make Medicare and Social Security decisions on their own is using either a free toy calculator or paying $5,000/year for advice on decisions they could understand themselves if anyone bothered to explain the math.

This middle band is enormous. It's also where I think AI is finally good enough — and cheap enough — to change the economics. Not by replacing the practitioners. By giving them the leverage that enterprise has had for 20 years.

Why now, and why me

Two reasons converged.

First, the AI inflection. Until 2023-2024, the dream of "an AI that reads a stack of tax documents and extracts the data the way a human would" was a research project. By 2026, it's a Tuesday-afternoon engineering task. The same is true for retirement-decision reasoning, document classification, and a dozen other workflows that previously required deterministic rules + human review. The cost curve on the model side dropped fast enough that the small end is now economically viable as a market for AI-native tooling, in a way it wasn't even 18 months ago.

Second, I had run out of patience for institutional pace. Two decades inside BNY taught me how to ship at scale, how to defend a complex platform in front of regulators, and how to build engineering culture in a regulated environment. It also taught me that the cycle time for "notice problem → ship solution to a real user" inside a bank is measured in quarters, sometimes years. The cycle time for the same loop outside a bank, with a small team and AI as a force-multiplier, is measured in days. I wanted that cycle time back.

So I left. The transition I expected to take two weeks took two months — clean exits from 19-year roles are not 2-week affairs, it turns out — but as of May 2026 I'm full-time on NESTPILOT.

What we're building

NESTPILOT is the umbrella. Two products operate under it today, with deliberately different audiences.

TaxPilot — operating tools for solo and small CPA firms

Audience: solo and small-firm CPAs, EAs, and tax preparers.

The wedge: the chaos before the return.

Tax preparers love their tax-prep software. Most don't love their tax-prep software, but they know it well, they've invested in workflows around it, and they're not switching. What they hate is everything that happens before the return gets into ATX or Drake or UltraTax or ProConnect:

• Chasing clients for documents across email, text, WeChat, and printed packages mailed to the office
• Sorting 47 PDFs into "W-2", "1099-INT", "1099-NEC", "K-1", "Schedule C support", "not actually a tax document"
• Re-typing data from PDFs that should have been a structured form
• Reconciling what the client said they have against what the IRS transcript shows
• The 47-email back-and-forth that follows every missing-document discovery

This is the workflow we're automating. AI-native document intake. AI-based extraction into a canonical tax JSON that's tool-agnostic. Import adapters into the four major tax-prep tools so the CPA's actual return-prep workflow doesn't change — the upstream chaos gets compressed into a single review queue. Our CPA Bridge product page goes deeper on what this looks like in practice.

We're in early design-partner mode with TaxPilot — a first CPA firm has agreed verbally; pilot work is underway. We're looking for 3-5 more design partners in the next 90 days.

NestPilot — retirement-decision tools for households

Audience: households approaching or in retirement, plus the advisors and CPAs who serve them.

The wedge: the five irreversible retirement decisions that cost $30K-$300K each.

We wrote about this thesis in detail yesterday. The short version: most retirement wealth isn't lost in budgeting. It's lost in a small number of narrow-window decisions — Medicare enrollment timing, IRMAA bracket cliffs, Social Security claiming age, Medigap underwriting, Roth conversion gap-years. Each one happens once or twice in a lifetime. Each one has a window that closes. And each one, when made wrong, costs tens of thousands to hundreds of thousands of dollars.

NestPilot is built around these decisions. It's not a budgeting app. It's a Decision Copilot for retirement — AI-native, retirement-specialist, with persistent context across decisions over time so each subsequent decision benefits from accumulated knowledge of your specific situation.

We operate a 501(c)(3) Foundation as the educational layer of NestPilot — free anonymous tools, plain-language explainers, and the Medicare Enrollment Guardian wizard, all hosted on the Foundation site. NESTPILOT LLC operates the paid Pro / Advisor / CPA Bridge tiers on top. The trust-first split means we don't need to monetize household data to fund the educational work, and the paid tiers cover advisors / CPAs / multiplayer scenarios on the same engine.

NestPilot product · Pricing · Foundation site

How the two products fit together

They aren't separate companies pretending to be related. The architecture is shared:

• A canonical tax / financial-profile schema that both TaxPilot (extraction-side) and NestPilot (planning-side) speak natively.
• A sanitized handoff layer — when a CPA using TaxPilot also serves a household using NestPilot, the CPA can hand off a sanitized, consent-bound profile from one product to the other without re-entry.
• Reactive (Spring WebFlux) backend, Next.js frontend, AI/LLM-based extraction and reasoning, Postgres + R2DBC for persistence. Stack consistency across products means lower maintenance burden for a small team.

The shared architecture is the long-term moat. A CPA-bridge tool that's also the household's retirement-planning workspace, with consent-bound data flow between them, is a category that doesn't exist yet in the small-end-of-finserv market.

What I'm asking for

Three asks, ordered by which would help most right now.

1. CPA design partners — solo and small firms, 1-15 preparers

If you run a solo or small CPA / EA / tax-prep firm, I'd love a 30-minute conversation about how your tax-season workflow actually breaks down. Not a sales call. Genuinely just want to understand the friction points before I build the wrong thing.

Specifically interested in firms where:

• You handle a mix of 1040s and small-business returns (Schedule C, K-1 distributions, etc.)
• Document intake currently happens across more than two channels
• You've looked at the "tax workflow automation" vendors and either bounced off the price or the fit

Design-partner slots include free access to TaxPilot for the pilot season, deep input into the roadmap, and named-partner status when we GA (if you want it). Open a CPA Bridge conversation.

2. Retirement-planning advisors evaluating partnership

If you're an independent RIA, CFP, or retirement planner serving 20-200 households, the multiplayer mode we're building — same engine the household sees, with advisor-facing controls and proposal workflows — may be relevant. The model is partner with, not replace. Specifics on our For Advisors page.

3. Conversations with builders in adjacent spaces

If you're building anything AI-native in fintech, regtech, tax-tech, or wealth-tech, I'd love to compare notes. Solo founders in this space tend to under-share — partly out of stealth instinct, partly because the natural support network is small. I'd like to fix that for the people in our cohort.

Building in public

I'm writing here weekly going forward. The cadence will be:

• What we're shipping — product changes, pilot lessons, technical decisions worth surfacing
• What we're learning — from CPA discovery conversations, advisor partnership conversations, household pilots
• Industry POV — where the AI-fintech space is going, what we're positioning for and against, what we got wrong

If any of the above resonates, subscribe to the blog feed, open a conversation, or find me on LinkedIn.

The decisions that compound over decades — both for CPAs running their practice and for households planning their retirement — deserve more than what the current generation of tools offers. That's the bet. Glad to have you reading.

— John Pan Founder, NESTPILOT LLC