Which AI models does Quantum use?

Anthropic's Claude models, and only those. A build-blocking test fails the server build if any other model provider is imported.

Can the AI change my client's data?

Your entries are never overwritten by AI. The main Q Assistant chat reads client data and never writes it. Uploaded documents can pre-fill empty client fields under validation and confidence rules, and any AI-proposed edit on the assistant's action surface is confirmed by you field by field before it is saved.

What happens if the AI is unavailable?

Every number is unaffected, because no number depends on the AI. Narrative surfaces fall back to deterministic templated text built from the same engine figures.

Methodology · AI

Deterministic Financial Modeling. AI-Powered Client Intelligence.

Quantum Portfolio Analytics draws a hard line between computing and explaining. Deterministic engines compute every score, projection, tax figure, allocation, and conversion schedule. AI — Anthropic's Claude models — explains those results to advisors, answers questions grounded in engine output, runs what-if scenarios through the engines, and drafts client-ready language that the advisor reviews before anything reaches a client.

In industry terms, Quantum uses large language models only as an interpretation and drafting layer over a deterministic calculation core; every numeric output is produced by tested, version-stamped engines that contain no model call.

The claim, exactly

Quantum's position on AI: Quantum's AI does not compute, alter, or determine any financial projection, and it does not select or size strategies. Every number and every strategy choice comes from Quantum's deterministic engines. AI helps advisors interpret modeled results, answer questions, explore scenarios, and draft client-ready language — which advisors review.

What the engines own

Each of these is a pure function of advisor-entered data: no model call, no randomness beyond a fixed seed, no live market data.

The Quantum Retirement Score and its simulation.
The Portfolio Durability Test.
Federal and state tax, Social Security taxation, RMDs, and Medicare IRMAA.
Modeled Roth conversion options, their pricing, and the lifetime comparison.
Income strategies, Five-Bucket allocation, withdrawal ordering, and suitability checks.
Every figure that appears on a client presentation.

What AI does

Explains the score. After the Quantum Retirement Score is computed, a narrative is generated from a packet of already-rounded engine figures; the narrative cannot feed back into the score.
Answers advisor questions. Q Assistant answers with numbers produced by engine tools — tax analysis, conversion scenarios, the score, the tax horizon, stress scenarios.
Explores scenarios. A what-if request is run through the same engines; the AI relays the engine's result.
Drafts client-ready language. Plan summaries and presentation commentary are drafted from engine output for the advisor to review and edit before delivery, and are labeled as drafted with AI assistance.
Reads uploaded documents. Statements and intake forms are read to pre-fill empty client fields, subject to validation rules and confidence thresholds; an advisor-entered value is never overwritten, and the advisor is told what was filled.
Stays available. When a model is unavailable, every narrative surface renders deterministic templated text built from the same engine figures.

How the boundary is enforced

A build-blocking test scans every server module and fails the build if any model provider other than Anthropic is imported. Narratives are generated from personally-identifiable-information-free packets of engine figures. Client-facing summaries carry the disclosure that portions were drafted with AI assistance and reviewed by the advisor, and Q Assistant answers carry an advisor decision-support notice. The engines themselves are covered by golden-master and parity test suites that pin their outputs byte for byte.

Why this is the trustworthy place for AI in financial planning

Financial projections must be reproducible, auditable, and explainable to a regulator. A language model is none of those things for arithmetic; it is very good at explaining arithmetic that something else did. Quantum's design keeps every number inside engines whose inputs and assumptions are shown beside the result, and uses AI where its strength lies — turning a correct result into language an advisor and a client can act on.

Assumptions

The engine / AI boundary by responsibility
Responsibility	Deterministic engines	AI
Scores, projections, tax figures, schedules	Compute them	Never
Strategy selection and sizing	Select and size	Never
Explanation and narrative wording	Supply the figures	Drafts; the advisor reviews
Client presentation and summary prose	Supply every number	Drafts commentary; the advisor reviews
Answering advisor questions	Run the tools	Composes the answer from tool results
Document intake	Validation rules and confidence gates decide what is applied	Reads the document and proposes values for empty fields
Market data	Run on advisor-entered data and the bundled, version-stamped historical return series	May run a bounded web search to answer a question; calculations never read it

Questions advisors ask about AI in Quantum

Does AI calculate the numbers?: Quantum's AI does not compute, alter, or determine any financial projection, and it does not select or size strategies. Every number and every strategy choice comes from Quantum's deterministic engines. AI helps advisors interpret modeled results, answer questions, explore scenarios, and draft client-ready language — which advisors review.
Which AI models does Quantum use?: Anthropic's Claude models, and only those. A build-blocking test fails the server build if any other model provider is imported.
Can the AI change my client's data?: Your entries are never overwritten by AI. The main Q Assistant chat reads client data and never writes it. Uploaded documents can pre-fill empty client fields under validation and confidence rules, and any AI-proposed edit on the assistant's action surface is confirmed by you field by field before it is saved.
What happens if the AI is unavailable?: Every number is unaffected, because no number depends on the AI. Narrative surfaces fall back to deterministic templated text built from the same engine figures.
Is AI-drafted text disclosed to clients?: Yes. Client-facing summaries state that portions were drafted with AI assistance and reviewed by the advisor, and client presentations carry an AI-assisted analysis disclosure.

Numbers from engines, language from AI, decisions from you

See it on a real client

Run the engines on your own clients and let Q Assistant explain the results. Licensed to financial professionals.