Most AI consultants have never operated a balance sheet. Most CFOs don't write code. Arthur Kostaras, CPA · CMA · CF · REV · CAFM Cand., does both — thirty years of senior finance leadership, four production AI platforms, and a published methodology. AI counsel for institutions that have run out of patience for theory.
Three decades as a senior finance executive across Canada, the U.S., Europe, and the Middle East. CFO and Managing Director roles in environments where the model has to actually clear.
Not slides. Four working AI-powered platforms on Railway / Node.js / PostgreSQL, designed and operated personally. Code, infrastructure, customers, the whole stack.
The Kostaras Method — a public Claude Code skills library used by practitioners to encode senior-finance judgment into AI workflows. Not theory; an artefact.
The work scales from a single ninety-minute briefing to a full fifteen-week implementation sprint. Every engagement begins with the same diagnostic and ends with operating capability transferred to your team.
Where AI is moving operating margin in 2026 — with the specific drivers, not the hype. The list below is not theoretical; it is the work being done with clients today.
A snapshot of current work across the three practice formats. Each engagement is run personally; no associates, no sub-contractors, no junior delivery.
A half-day live virtual PD course built for the CPABC catalogue. Module structure spans the Profit Formula, the Triangulated Lens, CAS implications, and an audit-defensible Working Paper Documentation Template.
A ninety-minute executive masterclass for approximately fifty commercial lenders at a major Canadian financial institution. Anchored against the OSFI Guideline E-23 effective date of 1 May 2027. The session converts a lending team from passive AI-curiosity into a credit-memo workflow that holds up to internal credit and OSFI review — before competitors do the same.
A fifteen-week AI-driven accountability build for a substantial Canadian construction company. Architecture, governance scaffolding, performance-fee structure, and a Day-105 handover. The system delivers what spreadsheets and dashboards have failed to: durable cause-and-effect visibility on project drift, before it shows up in the cost-to-complete — turning weeks of senior-PM forensic work into a daily, automated brief.
Four AI-powered platforms designed, built, and operated personally on a shared Railway / Node.js / PostgreSQL stack. Not portfolio pieces. Live, instrumented infrastructure.
Why does it matter that a CFO has shipped four platforms? Because the question every client should ask of any AI advisor is — have you ever owned the consequences of an AI decision through to operation? Most haven't. These platforms are the proof that the counsel has been earned the hard way.
An AI-powered matching platform for Canadian CPAs and the small businesses, families, and founders that need them. SEO-instrumented, profile-claim flow, AI bio generator, outreach templates.
A parallel AI-powered platform for Canadian lawyers. Same core stack, distinct content model. Designed for instrument-led claim-to-engagement conversion rather than passive directory listings.
An AI-powered platform for Canadian investors and capital allocators. First real claim conversions in flight. The stage of the lifecycle where capital is put to work, not just sought.
The fourth platform in the lifecycle — matching Canadian founders preparing for transition with the advisors, capital, and counsel that get good outcomes done.
The four platforms form a lifecycle, not a portfolio. Every Canadian professional, business owner, and capital allocator moves through these stages. The infrastructure to support them — instrumented, AI-native, and operating — is already built.
A working estimate of what a single ninety-minute briefing returns to an organization in the first twelve months. Move the sliders. The math is conservative by design.
Returned to the operating business in reclaimed analyst, manager, and partner hours during the twelve months after a single ninety-minute briefing. Conservative model: 48 working weeks, no compounding for downstream automation.
A published Claude Code skills library that encodes senior finance judgment into AI workflows. Not a prompt pack — a methodology with structure, conventions, and a working artefact behind it.
The Kostaras Method is a public, versioned skills library for Claude Code. It is the working memory of a senior finance executive, externalised — how to read an investment memorandum, how to structure a joint venture agreement, how to argue with a DSCR assumption, how to decide whether an AI output is defensible.
The method is what an operator does — the AI just executes it. Theory is the easy half. The harder half — the one that almost no AI consultant teaches — is the judgment that decides whether the answer is the answer.
Finance executive, AI builder, and instructor. Based in Victoria, BC. Operating from Coast Salish territory.
Three decades of senior finance leadership across Canada, the United States, Europe, and the Middle East — carrying roles as Managing Director, Chief Financial Officer, and Chief Executive Officer through capital-markets, royalty-financing, hospitality, real estate, defence, infrastructure, and Indigenous-economic-development environments. Career capital formation track record of $705M+. Operations management track record of $85M+.
The breadth of operating context is the source of the cross-industry intuition the practice delivers. A senior credit memo, a defence-procurement value proposition, a hospitality renovation proforma, and an Indigenous joint-venture agreement do not look the same on the page — but the disciplines that make each of them defensible are recognisably the same. That is the work.
Once frontier AI became operator-grade through 2024, applied it ruthlessly across the senior-finance workflow — investment memoranda, JVA drafting, financial modelling, governance documentation, due diligence. By late 2025, the next step was inevitable: shipping production AI infrastructure. The first platform was CanadaAccountants.app, followed by three more — a lifecycle of four platforms covering practice growth, capital allocation, and business exits for Canadian professionals.
The consulting practice grew organically out of that work. A Canadian financial institution, CPABC, and a substantial Canadian construction company were the first three institutions to engage Arthur personally as an AI Specialist. The practice exists to extend that work to a small, selective roster of additional institutions per year.
What the practice actually brings to the room. Each item is grounded in years of executive-grade delivery — not skills picked up for a brochure.
Aristotle distinguished three kinds of knowing in the Nicomachean Ethics. Sophia is theoretical wisdom — what academics deliver. Techne is technical craft — what AI tools deliver. Phronesis is the third thing: practical wisdom — the wisdom that knows how to act well in the specific situation.
That third thing is exactly what is missing in most AI advisory conversations. Theory and technical capability are everywhere; practical wisdom in their application is rare. The practice is named for what it installs.