What the assistant will and won’t do
“Ask about your plan” is the question-answering part of Retire.Simplified. It reads the plan you entered, runs the same calculation engine that powers the Plan and Improve screens, and explains the results in plain language. It is not a financial advisor, it does not know your full financial picture, and it will not tell you what to buy. This page says exactly what it will and won’t do, so you can decide how much to rely on it.
What it will do
It answers questions about your own plan by running calculations, not by guessing. For example:
- Am I on track? It runs your plan through 500 simulated markets and every real market sequence since 1928, and tells you the chance your money lasts, plus the age it runs out in a bad market.
- What if I retire at 62 instead of 65? It reruns the full projection with the change and shows the difference.
- When should I claim Social Security? It compares claiming at 62, 65, 67 and 70: monthly benefit, lifetime total and break-even ages.
- How much tax will I pay? It runs the federal and state tax engine for the year you ask about.
- Should I do Roth conversions? It models a conversion ladder against no conversions, including the Medicare surcharge (IRMAA) effect.
- What should I do first? It chains the projection, Social Security, Roth and withdrawal-order analyses into a ranked list with the dollar impact of each.
It will also answer general questions, such as how a 401(k) works or what an index fund is, from general knowledge. For those, it does not use your numbers.
What it won’t do
Five rules apply to every AI feature in the app: the chat, the health check, the optimizer, the quarterly review and the weekly email.
No funds, tickers, fund companies, brokerages, annuities or insurance products, and never “buy”, “sell” or “hold”. It can explain what a bond or an index fund is in general terms. If you ask what to buy, it will say that this tool doesn’t pick investments and suggest a fee-only fiduciary.
It can propose a change. You decide whether to apply it. It will never say a change “has been made”.
No credentials, certifications or exam scores. No advisor-client relationship is created by using it.
It can show what the tax engine calculates for a scenario. For decisions, it points you to a CPA or fee-only fiduciary.
Every number is framed as “given the assumptions you entered”. A projection is a model, not a forecast.
Where its numbers come from
Every number about your plan comes from the calculation engine, never from the AI’s own arithmetic. The assistant has a set of tools (read the plan, run the projection, run the tax estimate, compare scenarios, and so on) that call the same functions the Plan and Improve screens use. If it hasn’t run a calculation, it says so rather than estimating.
That means you can check anything it tells you:
- A year-specific figure points to the Ledger on the Plan screen (“see age 67 in the Ledger”).
- A dollar impact points to Improve → Show the math, which lists the steps.
- The engine itself is checked against hand-worked IRS and Social Security cases on How we check the math.
If the assistant and a screen ever disagree, the screen is right and the assistant has a bug. Please report it (see the last section).
Changes to your plan
When the assistant suggests a specific change, such as claiming Social Security at 70 or raising a monthly contribution, it shows an Apply button next to the suggestion. Nothing happens until you click it.
- You apply, it doesn’t. The assistant records a proposal; the app changes the plan only on your click.
- Stale suggestions are caught. If you have edited the plan since the suggestion was made, the app checks the current value before applying, so an old suggestion can’t overwrite a newer decision.
- You can undo. After you apply a change, an Undo button stays available for 30 seconds and restores the previous value.
Every suggestion is a hypothetical scenario based on your inputs, not personalized advice.
Your data
There are no accounts. Your plan lives in your browser and is tied to a random ID your browser created on first visit.
- In your browser: your plan, saved scenarios and settings. Clear your site data and they’re gone.
- On our server, keyed to that browser ID: chat history, past health-check reports, and your email preference if you turn on the weekly digest. This is what lets the assistant remember earlier conversations and show your plan’s trend over time.
- Sent to the AI model: the plan figures needed to answer your question, for that request only. The assistant is instructed never to reference your data outside the conversation, and we do not use your plan to train anything.
- Usage analytics (Google Analytics): which pages and screens are visited, and counts of a few actions (finishing setup or the Quick verdict, applying a move, asking a question). It sets Google Analytics cookies in your browser. It never receives your plan’s numbers or what you type, and advertising features are switched off. Analytics is off if your browser sends Global Privacy Control.
- Never collected: your name, account numbers or login credentials. We have no way to link a plan to a person unless you enter an email for the digest.
If you use the app on two devices, they are two separate plans. Export and import on the Tools page moves a plan between them.
How these rules are enforced
The rules above are not just a promise on a page.
- They are written once, in code, and attached to every AI feature. A new feature can’t skip them.
- Automated tests fail if the rules are removed or weakened, so they can’t be dropped quietly.
- The assistant can only get plan numbers through tools that call the real engine, so it can’t invent them.
- The AI model each feature uses is set in one place and checked daily, so a retired or misnamed model is caught within a day.
- The source code is open on GitHub, so anyone can verify this.
The assistant will still get things wrong sometimes. If it names an investment, claims to be an advisor, quotes a number that disagrees with a screen, or does anything else this page says it won’t, open an issue with the question you asked. That is how the rules get tighter.
See also How the math works, How we check it, How it works and What the assistant will and won’t do.
Educational planning estimates, not financial advice.