---
title: "How to Build a Portfolio With AI (2026)"
description: "A step by step guide to building an investment portfolio with AI: set a goal, research holdings, stress-test the mix and check the numbers yourself."
url: https://moneda.com/blog/how-to-build-a-portfolio-with-ai
lang: en
---

# How to Build a Portfolio With AI (2026)

> A step by step guide to building an investment portfolio with AI: set a goal, research holdings, stress-test the mix and check the numbers yourself.

Published: 2026-10-23
Updated: 2026-10-23
Author: Federico Kunze Küllmer

Source: https://moneda.com/blog/how-to-build-a-portfolio-with-ai

---
*Federico Kunze Küllmer is the founder of Moneda, one of the tools compared here.*


> **Short answer**: Start with a written goal (what the money is for, when you need it, how big a fall you can sit through). Ask an AI to suggest a mix of assets that fits that goal and to explain why. Then research each holding with sources you can open, stress-test the whole portfolio against past crashes, and compare it with a simple benchmark. The AI does the legwork. You check the numbers that matter and make every buy or sell decision yourself.

AI is good at the slow parts of investing: reading filings, pulling figures, running "what if" scenarios. It is less good at knowing you. So the order matters. You bring the goal and the judgment, the AI brings the research, and you check its work before any money moves.

This guide walks through that process in seven steps, with example questions you can paste into any AI tool. We use [Moneda AI](/financial-agent) as the example in a few places, since it is built for this job, but the method works with general chatbots like [ChatGPT](https://chatgpt.com), [Claude](https://claude.ai) or [Gemini](https://gemini.google.com) too. Where Moneda differs is that research and investing sit in the same app: the AI does the analysis and shows its sources, stress-tests your mix against 2008, 2020 and 2022, benchmarks it against gold or an index, and you buy only with your approval. Most tools do one or the other.

## Step 1: Write down your goal before you open any app

An AI can only build around what you tell it. "I want to make money" gives it nothing to work with. Try to answer three questions in a sentence each:

- What is the money for? A house deposit, retirement, a cushion you might touch in two years.
- When will you need it? Time matters more than almost anything else. Money you need in 18 months should not sit in the same mix as money you will leave alone for 20 years.
- How much of a drop can you live with? If your portfolio fell 30% in a bad year, would you sell in a panic or wait?

Then hand that to the AI as plain text:

- "I'm 34, investing €300 a month for retirement around age 65. I can handle big swings but I don't want more than a third in any single sector. What kind of mix fits that?"
- "I need this money for a flat deposit in about four years. What should I avoid?"

If the reply ignores your time frame or your limit on losses, push back. That is the first test of whether the tool is listening.

## Step 2: Ask for a starting mix, and ask why

Most portfolios are built from a few building blocks:

- **Stocks**: shares in single companies.
- **ETFs** (exchange-traded funds): one product that holds many stocks or bonds at once and trades like a share. A broad index ETF is the simplest way to own hundreds of companies in one buy.
- **Commodities**: things like gold, silver or oil, usually held through an ETF rather than physically.
- **Crypto**: digital assets, which tend to swing much more than the rest.

Ask the AI to propose a split and to explain the reasoning in plain words:

- "Suggest an illustrative portfolio for my goal using mostly ETFs. Explain why each piece is there and what job it does."
- "What's the case against this mix? What would make it a bad fit for me?"

In Moneda, you can describe your goal and Moneda AI builds an illustrative portfolio from it, then tracks it and tells you when something changes. It also sends a daily briefing on your portfolio and the market, so you know what moved without checking every holding. If you already hold investments at another broker or app, you can import them as paper portfolios and have the agent analyse the stocks and your positions. "Illustrative" is the right word for anything an AI gives you here. It is a draft to question, not an instruction.

## Step 3: Research each holding with sources you can open

A portfolio is only as good as what is in it. For every single stock or fund, ask the AI to do the reading and to show where each number came from.

A few terms you will see:

- **10-K**: the annual report a US-listed company files with the [SEC](https://www.sec.gov/), the US market regulator. It is the most reliable source for a company's numbers.
- **P/E** (price to earnings): the share price divided by profit per share. A rough sense of how much you pay for each euro or dollar of profit.
- **Free cash flow**: the cash a business has left after paying to run and maintain itself. Harder to dress up than reported profit.
- **Moat**: a lasting advantage that keeps competitors away, like a strong brand or high switching costs.

Good questions:

- "Summarise the last 10-K for this company. Link the section behind each figure."
- "What has free cash flow done over five years, and does it match reported profit?"
- "Give me the bull case and the bear case, with the evidence for each."
- "Does this company have a moat? What could erode it?"

Moneda AI researches stocks, ETFs, a few commodities through ETFs (gold, silver, oil) and crypto, using company filings and live market data. It works figures out from the data and shows the source behind each one. It also gives both sides of the story, scores a company's [7 Powers](/blog/what-are-the-7-powers) (a framework for judging lasting competitive advantage), checks the quality of earnings (whether profit is backed by real cash) and gives a Moneda Ratio, a descriptive 1 to 7 scorecard. None of these tell you to buy. They are there so you can see the shape of a company quickly.

If you are starting from scratch, describe what you want and let the AI find candidates:

- "Find US-listed companies with growing free cash flow, little debt and a P/E below the market average."

## Step 4: Check the numbers yourself

This is the step people skip, and it is the one that matters most. AI tools can get figures wrong, mix up years or quote a number with no source. Some checks take two minutes:

1. **Open the source.** If the AI cites a filing, click through and find the figure. If there is no source, treat the number as unconfirmed.
2. **Check the date.** Is that revenue figure from the latest year, or two years old?
3. **Recompute one ratio.** Take P/E: look up the price and earnings per share, divide, and see if you get roughly the same answer.
4. **Read the bear case properly.** If the AI could only find weak arguments against a holding, ask again. Every company has real risks.
5. **Look at fees.** For ETFs, check the yearly cost on the fund provider's own page.

You can make the AI do part of this for you:

- "Which of the numbers you just gave me are you least sure of?"
- "Show me the exact line in the filing where that margin comes from."

On [Moneda's own tests](/models/numa-1) in October 2026, Moneda AI scored 97.9% on groundedness (answers backed by the sources it retrieved) and 92% on claim verifiability. Those are Moneda's own tests, not independent audits, and 92% still means some claims need checking. Do the checks.

## Step 5: Stress-test the portfolio against bad years

A portfolio that looks fine in a calm market can behave very differently in a crash. Ask the AI to replay history:

- "How would this portfolio have done in 2008, 2020 and 2022?"
- "What was the worst drop, and how long did it take to recover?"
- "Which holding did the most damage in each of those years?"

2008 was the global financial crisis. 2020 was the sharp Covid crash and rebound. 2022 saw stocks and bonds fall together as interest rates rose. They test different weaknesses, so it is worth running all three.

| Year | What happened | What it tests |
|---|---|---|
| 2008 | Global financial crisis | A long, deep fall led by banks |
| 2020 | Covid crash and rebound | A sudden shock and fast recovery |
| 2022 | Stocks and bonds fell together | Rising interest rates hitting both at once |
 Moneda AI can [stress-test a portfolio](/blog/best-ai-tools-to-stress-test-a-portfolio) against exactly these three years.

If the worst drop is bigger than the number you wrote down in Step 1, change the mix now, not in the middle of the next crash.

Some people also ask for a **DCF** (discounted cash flow), a model that estimates what a company is worth today from the cash it might make in future. It can be useful, but the result depends heavily on the guesses you feed it. Ask the AI to show how the answer changes if growth is a few points lower.

## Step 6: Compare against a simple benchmark

A benchmark is the thing you would have owned if you had done nothing clever. Usually that means a broad index fund. If your hand-built portfolio cannot beat it, or at least match it with less risk, the simpler option may be better.

- "Compare this portfolio with a global index fund over 5 and 10 years."
- "Compare it with just holding gold. Which had the deeper falls?"

In Moneda you can [benchmark a portfolio](/blog/best-ai-tools-to-benchmark-a-portfolio) against gold, an index or any fund.

## Step 7: Invest, then review on a schedule

Once you are happy with the plan, you buy. Some people do this at their usual broker. In Moneda you can [invest in the same app](/invest): US-listed shares and ETFs (including fractional shares, so you can buy part of one) and crypto. Every trade is approved by you with your passkey, using your face or fingerprint. Moneda AI never tells you to buy or sell.

After that, the work is mostly patience. Pick a review rhythm, every three or six months, and stick to it. Between reviews, let the AI watch for changes:

- "Tell me if any holding's debt rises sharply or its earnings quality drops."
- "Has anything changed in my portfolio this quarter that I should read about?"

If you use Claude or ChatGPT already, [Moneda MCP](/developers) connects them to your Moneda account with the permissions you set. You choose what each agent can see and do, and any payment still has to be approved in the app. Premium and Ultra plans also include a REST API and CLI for people who want to script their own checks.

## What AI does well here, and where it falls short

AI tools save hours on reading and number-crunching. They are also good at arguing the other side when you ask. Where they fall short:

- They do not know your full situation unless you tell them: tax, debts, other savings.
- They can state wrong numbers with total confidence.
- General chatbots may not have live prices or recent filings unless connected to a data source.

| | General chatbot (ChatGPT, Claude, Gemini) | Moneda AI |
|---|---|---|
| Live prices and filings | Only if connected to a data source | Built in, with the source shown for each figure |
| Stress tests (2008, 2020, 2022) | You supply the data | Built in |
| Benchmark against gold or an index | You supply the data | Built in |
| Daily briefings on your holdings | Only if you ask each day | Built in |
| Holdings at other brokers | You paste them in each time | Import them as paper portfolios for the agent to analyse |
| Buying | Elsewhere, at your broker | Same app, each trade approved with your passkey |
| Coverage | Anything you paste in | US-listed shares and ETFs, crypto, a few commodities through ETFs |

Moneda has limits too. It invests in US-listed shares and ETFs only, so no European-listed stocks or funds.

On cost, Moneda's [Standard plan](/plans) is free. Premium starts at €9.99 a month and Ultra at €19.99 a month. Higher plans add more AI usage and more commission-free trades each month (1, 5 or 10, depending on the plan).

## FAQ

**Can AI build a portfolio for me?**
It can draft one from your goal, research each holding and stress-test the mix. It cannot know your whole financial life, and it should not make the final call. Treat its portfolio as a starting draft.

**Is it safe to trust the numbers an AI gives me?**
Only once you have checked them. Ask for the source behind each figure, open it, and recompute at least one ratio yourself. Even tools that score well on their own tests get some claims wrong.

**How many holdings should a beginner portfolio have?**
There is no fixed number. Many beginners start with one or two broad ETFs, which already hold hundreds of companies, and add single stocks only once they understand them.

**Which past crashes should I test against?**
2008, 2020 and 2022 cover three different kinds of trouble: a banking crisis, a sudden shock, and rising interest rates. If your mix holds up acceptably in all three, you know more about its weak spots.

**Does Moneda AI tell me what to buy?**
No. It researches, scores and stress-tests, and it can build an illustrative portfolio from your goal. Every trade is your decision, approved with your passkey.

**Can I use ChatGPT or Claude with my Moneda account?**
Yes, through Moneda MCP. You set what each agent can see and do, and payments are always approved in the Moneda app.

*This guide is for education only and is not investment advice.*