AI Portfolio Analysis: Tools and How-To

AI portfolio analysis is the practice of pointing software at the investments you already own and getting back a plain-language read on how they are put together and how they are doing. Instead of exporting positions into a spreadsheet and building your own formulas, you connect an account, the tool reads what you hold, runs the same math a careful analyst would, and a language model explains the result in a sentence you can act on. The output is not a stock tip. It is a diagnosis: where you are concentrated, where two funds secretly own the same companies, how your mix compares to a simple benchmark like the S&P 500, and how much you are quietly paying in fees.

The category exists because the raw work is tedious and most people skip it. Checking whether your top five holdings make up 60 percent of your money, or whether three of your ETFs all lean on the same handful of megacap tech names, is arithmetic, not genius. AI does not make that arithmetic smarter. What it changes is the friction: it reads the account for you, does the counting, and turns the numbers into a narrative, so the analysis actually gets done. For a hands-on walkthrough of the workflow, see how to analyze your portfolio with AI.

Almost every AI portfolio tool works from the same short list of measurements. The wording differs, but the underlying questions are stable, and knowing them lets you judge any tool on whether it answers them honestly.

• Concentration. How much of your money sits in a single stock, sector, or asset. A portfolio that is 40 percent one company behaves very differently from an evenly spread one. This is usually the first thing worth checking, and it is covered in depth in how to check portfolio concentration.
• Holdings overlap. Whether funds you think are diversifying you are actually duplicating each other. Owning an S&P 500 fund, a total-market fund, and a Nasdaq fund can leave you tripled up on the same top names without realizing it.
• Diversification. The spread across sectors, geographies, market caps, and asset classes. True diversification is about how holdings move relative to each other, not just how many line items you have. See how to analyze portfolio diversification for the fuller picture.
• Performance versus a benchmark. How your portfolio has done against a reference like the S&P 500 over a chosen window. A 12 percent gain feels good until you learn the index returned 20 percent over the same stretch. Benchmarking turns a raw number into a judgment.
• Risk and drawdown. Volatility, the size of past declines, and how much a portfolio could fall in a bad stretch. Some tools add a single risk score; the useful ones show the drawdown history behind it rather than a black-box number.
• Fees and cash drag. Expense ratios on funds, advisory fees, and money sitting idle in cash. Fees look trivial in a single year and compound into real money over decades, so a fee read is one of the highest-value things these tools surface.

A tool that covers concentration, overlap, diversification, benchmarking, risk, and fees is doing the full job. One that only shows a total balance and a green or red arrow is a tracker, not an analyzer, which is a distinction worth keeping straight when you shop.

The category looks crowded until you sort it into three types by what they are built to do. Most products lean into one, though the lines blur.

Portfolio analyzers exist to critique what you hold. You connect accounts and get back a structured second opinion: a risk read, concentration and diversification findings, benchmark comparison, and often suggested changes. PortfolioPilot is the clearest example of the AI-native version, generating a full critique with a risk score. Empower (the free dashboard formerly called Personal Capital) runs an Investment Checkup and a fee analyzer across aggregated accounts. SigFig offers portfolio analysis alongside its managed service. Morningstar sits here too, with its long-standing fund and portfolio X-Ray tools that professionals have used for years. For a ranked shortlist of the AI-native analyzers, see the best AI portfolio analyzer roundup.

Portfolio trackers exist to keep a running picture of what you own and what it is worth, across every account. The emphasis is aggregation and monitoring rather than a verdict. Kubera tracks net worth across brokerages, banks, crypto, and even real estate. Sharesight specializes in performance and dividend tracking with tax reporting. Monarch aggregates investments next to budgeting and cash flow. Portfolio Visualizer is the analytical outlier here, a backtesting and factor-analysis tool used to model how a mix would have behaved historically. Trackers answer "what do I have and how is it doing," which is a real need, but they stop short of telling you what it means.

AI chat assistants exist to let you interrogate a portfolio in plain English. Rather than reading a dashboard, you ask a question and get an answer grounded in your real holdings. Magnifi is a conversational assistant for asking about funds and holdings. Mezzi layers tax-aware AI insight across aggregated accounts. Walnut connects the broker you already use and lets you analyze and manage what you hold by talking through Claude, ChatGPT, or a built-in assistant. The chat model is the newest of the three and the one most changed by modern language models, because the explanation step, not the math, is what AI genuinely improved.

It helps to know what is happening behind the friendly summary, because it explains both what these tools do well and where they can be confidently wrong. The good ones follow the same four steps.

1. Connect the account, read-only. You link a brokerage or set of accounts, usually through a regulated aggregator, so the tool can see positions without being able to move money. The safer designs never store your broker login and default to read-only. This step is what separates a grounded analysis from a chatbot guessing.
2. Read the holdings. The tool pulls your actual positions: tickers, share counts, and market values, plus fund holdings where it can look through an ETF to the underlying companies. Everything downstream depends on this being your real portfolio rather than a rough description you typed.
3. Run deterministic math. Concentration percentages, overlap between funds, sector weights, return versus a benchmark, and volatility are all computed with ordinary, repeatable formulas, not a language model. This is the part that must be exact, and in a well-built tool it is code, so the same holdings always produce the same numbers.
4. Let the model explain it. Only after the math is done does the language model step in, to turn "AAPL, MSFT, and NVDA are 48 percent of equity value" into "nearly half your money rides on three tech names, so a rough patch for big tech would hit you hard." The AI is the narrator, not the calculator.

The distinction matters for trust. When the arithmetic is deterministic and only the explanation is model-generated, the numbers are reliable and the prose is the part to sanity-check. When a tool skips the read step and asks a general model to reason about a portfolio you described in a chat box, both the numbers and the story can be invented. That is the difference between analysis grounded in connected holdings and a plausible guess.

A quick map of the named tools in this category by their primary type and account model. Features and pricing change, so treat this as a starting point and verify on each provider's site.

A useful test when comparing these tools: ask whether the output tells you something you could not have worked out by looking at your account. A list of your holdings with green and red arrows is a dashboard. A statement about the shape of the portfolio is analysis, and the clearest example is concentration.

The measure worth knowing is how many positions your portfolio actually behaves like, as opposed to how many you hold. Convert each position to a share of the total, square the shares, add them up, and divide one by the result. Ten holdings at equal weight gives ten. Ten holdings where one has grown to 40% of the money gives about 4.4, because a single name now drives the outcome. The squaring is what does the work: a 40% position contributes a hundred times more to the total than a 4% one, which is a fair reflection of how much each will move your account.

Walnut computes this as Effective Holdings and publishes the full method, because a concentration figure you cannot reproduce is a claim rather than evidence. The underlying measure is not proprietary: it is the reciprocal of the Herfindahl-Hirschman Index, which competition regulators use to assess how concentrated a market is.

Two things it will not do, which are worth demanding of any analyzer that offers something similar. It does not know about correlation, so eight semiconductor companies count as eight holdings even though they move together. And it does not look inside your funds, so overlapping ETFs understate your true concentration. Any tool that presents a concentration score without saying this is overselling the arithmetic.

The right tool depends less on which has the longest feature list and more on what you actually want the analysis for. Work through these questions in order.

• Do you want a verdict, a running picture, or a conversation? If you want a one-shot critique, reach for an analyzer like PortfolioPilot or Empower. If you want to watch everything in one place over time, a tracker like Kubera, Sharesight, or Monarch fits. If you want to ask questions and dig in, a chat assistant like Magnifi, Mezzi, or Walnut is the model.
• Does it read your real holdings? Grounded analysis beats a general chatbot reasoning from a description. Prefer tools that connect your accounts, read-only, so the numbers reflect what you actually own.
• Does the math look deterministic? The concentration and benchmark figures should be exact and repeatable. Be wary of a tool where a language model appears to be inventing percentages rather than computing them.
• What is the account and safety model? Read-only by default, no stored broker login, and a regulated aggregator are the marks of a careful design. Check whether the tool can ever place a trade and under what controls.
• What does it cost, and in what shape? A free tier, a flat subscription, and a percentage of assets compound very differently over the years you plan to invest. Empower's dashboard is free; Walnut has a free tier; several trackers and analyzers charge a flat subscription.
• Is it honest about what it is? Most of these are informational tools, not registered investment advisers. Anything promising to beat the market deserves skepticism. According to S&P Dow Jones Indices' SPIVA scorecard, the large majority of active US large-cap funds have trailed the S&P 500 over long windows, so the honest promise is a better process, not guaranteed returns.

Every analyzer in this category works from the data it is given, and the gaps in that data are consistent across all of them. Knowing the gaps is what separates using one of these tools well from being quietly misled by it.

• Cost basis is usually missing. Most brokerage connections expose current positions and values but not what you paid, which means the return figures you see are window returns over the period being charted rather than your lifetime profit and loss. A holding shown as up 12% may be one you are down on. Any tool presenting a return without saying which of the two it is has made a choice on your behalf.
• Accounts you do not connect are invisible. Concentration computed on one account understates the truth if you hold the same company in a retirement account elsewhere. The arithmetic is only as complete as the connections.
• Fund internals are opaque by default. A tool sees that you own an ETF at a weight. Whether two of your ETFs hold many of the same underlying companies is a separate lookup, and not every tool does it.
• Why you own something is never in the data. A position held for a tax reason, a restricted holding, or one you intend to give away all look identical to the software.

None of these make the tools useless. They make the difference between a tool that describes and one that instructs, which is why the honest products in this category stay descriptive: the software knows the shape of your portfolio and does not know your situation.

If you have never analysed your own portfolio properly, the sequence matters more than the tool. Each step answers a question the next one depends on.

• Connect everything you can, read-only. Analysis of one account is analysis of one account, and most of the surprises live in the gap between accounts.
• Look at concentration before performance. Performance tells you what happened; concentration tells you what will happen to you next. It is also the number most likely to be different from what you expect.
• Check overlap between funds. Two holdings you believe are diversifying often are not, and this is invisible from fund names alone.
• Only then compare returns, and know which return you are seeing. Window return against a benchmark is a fair question. Lifetime profit and loss usually is not available, and conflating them is the most common misreading in this whole category.

Doing it in that order tends to change what you ask. People arrive wanting to know whether a holding is performing and leave having discovered the portfolio is a different shape than they pictured.

To be upfront, since this is our site: Walnut is the AI investing assistant that talks to the broker you already have and places the trades you approve. In this field it sits in the chat-assistant type. It connects your existing brokerage, read-only by default, reads your real holdings, runs the concentration, overlap, benchmark-versus-the-S&P-500, and drift math deterministically, and then lets you talk it through with Claude, ChatGPT, or a built-in assistant. You keep your broker, you pick what you hold, and any trade needs your approval. Walnut is not an investment adviser.

It is one option among many, and it leads only in its own category rather than across the board. If you want a single structured critique, an analyzer like PortfolioPilot is a cleaner fit; if you want to track net worth across everything, Kubera or Monarch will. Walnut is for the person who wants to analyze and manage real holdings by having a conversation, on the broker they already use, rather than reading a dashboard or handing the account to someone else.