What is a robo advisor and how does it manage portfolios?

What is a robo advisor? At its core, it is an automated investment platform that converts a client’s stated objectives, time horizon, and tolerance for loss into a rules-based portfolio—then maintains that portfolio with limited human intervention.

The significance is no longer confined to a lower-cost corner of retail wealth management. Robo-advisory platforms now oversee more than $2.5 trillion globally by recent estimates. That scale reflects a structural re-pricing of basic portfolio construction: diversified exposure, periodic rebalancing, and tax-aware implementation have become software functions rather than labor-intensive advisory services.

The original proposition was straightforward. Replace the conventional 1%-plus advisory fee with a digital workflow charging roughly 0.25% to 0.50% of assets under management, populate portfolios with low-cost exchange-traded funds, and remove operational friction from an activity many investors historically neglected. The result was not a new investment philosophy. It was a new delivery system for established portfolio theory.

For asset managers, custodians, and wealth platforms, that distinction matters. Robo-advisory is not principally about a machine “picking stocks.” It is an infrastructure layer for capital formation, portfolio administration, and client servicing at a scale conventional advisor-led models struggle to support economically.

The algorithmic core: Modern Portfolio Theory translated into software

Most robo-advisory algorithms rest on a familiar institutional foundation: Modern Portfolio Theory. The platform seeks to construct a mix of assets expected to deliver an appropriate balance of risk and return for a defined investor profile. The software does not know the future, and it does not eliminate drawdowns. It formalizes a process for deciding how much portfolio risk a client should own and how that risk should be diversified.

In practice, a robo advisor usually allocates across five to 12 asset classes. The exact architecture differs by provider, but the common building blocks are broad equity, fixed income, and—in more expansive portfolios—international stocks, emerging markets, real estate securities, inflation-sensitive assets, or other listed exposures.

The underlying instruments are typically ETFs. Their role is central to the economics. ETFs offer broad market access, intraday liquidity, and comparatively low expense ratios, often in the 0.05% to 0.20% range for the core exposures used in automated portfolios. A robo advisor can therefore charge a modest management fee without layering the substantially higher internal costs associated with many actively managed mutual funds.

The portfolio process generally follows four decisions:

1. Establish a strategic asset allocation. The platform defines a long-term mix of stocks, bonds, and other exposures based on the client’s profile. This is the dominant investment decision. A 70/30 allocation and a 40/60 allocation may both be diversified, but they embody meaningfully different loss tolerance and return expectations.

2. Select implementation vehicles. The algorithm maps each asset-class target to one or more ETFs. The selection criteria can include expense ratios, tracking quality, liquidity, tax efficiency, and the platform’s commercial architecture. Some providers use third-party funds; others incorporate proprietary funds or affiliated products.

3. Set tolerances around target weights. No portfolio remains at its original allocation once markets move. The platform establishes rules for when deviations become large enough to warrant a trade.

4. Apply the rules consistently. This is where wealthtech automation changes the operating model. The client does not need to notice that equities have appreciated faster than bonds, log in, calculate the drift, and decide whether to act. The system does that work continuously or on a scheduled basis.

The robo model did not automate investment judgment in its entirety. It automated the repeatable mechanics surrounding a strategic allocation decision.

There is a tendency to overstate the sophistication of the algorithmic layer. The proprietary weighting formula at any individual provider is usually opaque, and the market should be skeptical of claims that a secret optimization engine can consistently manufacture superior returns from standard liquid beta. The durable value proposition is more prosaic: disciplined diversification, lower implementation cost, and fewer operational lapses.

That is a considerable proposition in wealth management. It is simply not equivalent to an autonomous investment manager with a unique information advantage.

Onboarding: how a questionnaire becomes an investment mandate

Digital wealth management begins with a client onboarding workflow. The familiar questionnaire asks about age, investment objective, income, assets, time horizon, planned withdrawals, investment experience, and comfort with market losses. A client saving for a near-term property purchase should not receive the same allocation as a younger investor funding a retirement account decades away. The questionnaire is the mechanism that turns that distinction into a portfolio mandate.

This is one of the more consequential parts of the system—and one of the least glamorous. Asset allocation is only as suitable as the information used to construct it.

A typical robo advisor will assess:

• Investment horizon: When capital is likely to be needed. A short horizon creates a duration mismatch if the portfolio carries substantial equity risk.
• Return objective: Whether the account is intended for capital preservation, income, long-term growth, or a defined liability.
• Risk capacity: The financial ability to withstand losses, which is different from emotional willingness to do so.
• Liquidity requirements: Whether the investor may need to withdraw funds during a market drawdown.
• Tax status and account type: Taxable accounts, retirement accounts, trusts, and other structures carry different implementation constraints.
• Existing holdings: Some platforms can account for outside assets; others optimize only the capital held within their own custody environment.

The best digital onboarding systems recognize that risk tolerance is not a single number. It is an imperfect proxy for the client’s capacity to absorb a permanent or temporary impairment of capital. A questionnaire can identify broad portfolio bands. It cannot reliably resolve the harder questions around concentrated business ownership, intergenerational wealth transfer, executive compensation, illiquid private assets, or complex tax obligations.

This is why the notion that robo advisors will simply replace advisors has always been analytically weak. They replace—or more accurately, compress—the economics of standardized portfolio management. Complex planning remains a different business with a different labor model.

The table is less a contest than a division of labor. A $25,000 diversified account and a multigenerational balance sheet should not be priced or serviced as if they pose the same operational burden.

Automated rebalancing: the quiet discipline of the model

Automated portfolio rebalancing is the operational heart of a robo advisor. Markets move; target weights drift. If global equities rise sharply while bonds remain flat, an initially balanced portfolio can become materially more equity-heavy than the investor intended to own.

Rebalancing restores the original strategic allocation by selling portions of overweight assets and buying underweight ones. The process is inherently countercyclical. It asks the portfolio to trim exposures that have appreciated and replenish those that have lagged.

That behavior sounds elementary. It is also where conventional client behavior routinely breaks down. Investors often allow winning assets to become oversized because selling them feels premature, while underperforming assets are abandoned because adding to them feels uncomfortable. A rules-based system does not eliminate market risk, but it can reduce the behavioral leakage created by delayed decisions and emotional timing.

Platforms use different triggers. Some rebalance at fixed intervals; others monitor portfolios and trade only when an asset-class weight breaches a defined tolerance band. The latter approach can reduce unnecessary turnover, which matters in taxable accounts and in environments where transaction costs, bid-ask spreads, or market impact remain relevant.

The practical sequence is simple:

1. A client begins with a target allocation, such as 60% equities and 40% fixed income.

2. A market move causes the portfolio to drift, perhaps to 67% equities and 33% fixed income.

3. The robo-advisory algorithm compares actual weights with target weights and its permitted deviation ranges.

4. If the threshold is breached, the platform sells sufficient equity exposure and buys fixed income exposure—or directs new cash flows toward the underweight sleeve.

5. The account returns closer to its intended risk posture.

The value here is not tactical prescience. Rebalancing does not forecast a reversal in equities or bonds. It protects the integrity of the client’s original allocation mandate. In institutional language, it is portfolio governance embedded in trading infrastructure.

Tax-loss harvesting: a scaled version of a private-wealth technique

Tax-loss harvesting is the feature that moved robo advisors beyond basic digital brokerage. In taxable accounts, the platform may sell an investment trading below its purchase price, realize the loss, and use that loss to offset realized capital gains where applicable. The proceeds are then reinvested in a similar exposure so that the portfolio remains aligned with its strategic allocation.

Historically, this kind of tax-aware trading was associated with high-net-worth mandates, where advisors and portfolio managers had sufficient fee revenue to justify account-level oversight. Wealthtech automation made the process scalable across much smaller balances.

The mechanics require care. A platform cannot simply sell a broad-market fund, hold cash, and call the exercise tax optimization. The portfolio must maintain market exposure, and the replacement holding must be chosen with attention to applicable tax rules. That is why platforms typically pair a sold ETF with a comparable, but not identical, alternative exposure.

There are three points that deserve emphasis.

First, tax-loss harvesting is not a source of investment return in the conventional sense. It is a tax-management technique. Its value depends on the investor’s taxable gains, holding period, jurisdiction, tax rate, available losses, and future ability to use those losses.

Second, it is most relevant in taxable brokerage accounts. Its utility is materially different in tax-advantaged retirement accounts, where realizing losses generally does not produce the same benefit.

Third, the feature does not repeal the economic reality of a declining asset. A harvested loss means the investment fell in value. The tax offset can soften the after-tax result; it cannot convert a loss into a gain.

Tax-aware automation is valuable precisely because it is unromantic: it turns hundreds of small, administratively burdensome decisions into a consistent process.

For asset managers, this capability has strategic consequences. It shifts competition away from fund selection alone and toward account-level outcomes. A low-cost ETF is useful; an ETF embedded in automated tax management, rebalancing, reporting, and digital onboarding is a more durable client proposition. That is where platforms seek to earn their liquidity premium in the client relationship.

The fee model: where the disruption actually occurred

The original robo-advisory challenge to incumbent wealth managers was economic. A traditional advisor charging 1% or more of AUM had to demonstrate value beyond allocating a client across a handful of diversified funds. For straightforward portfolios, technology exposed how much of the historical fee stack paid for manual administration rather than scarce investment expertise.

A typical robo advisor charges an annual management fee of roughly 0.25% to 0.50% of assets, often with account minimums ranging from zero to $500. Fund expenses sit beneath that fee, but are generally modest because portfolios tend to use low-cost ETFs.

The all-in cost is not always as simple as the headline management fee suggests. A platform offering a zero management fee may still generate revenue through cash allocations, proprietary products, securities lending arrangements, premium planning tiers, or other elements of its balance-sheet and product model. Fee compression does not mean the economics disappear. It means the economics migrate.

That migration is one of the defining themes in wealth management. The revenue pool is moving from a standalone advice charge toward a broader ecosystem of custody, banking, lending, cash management, portfolio products, and technology services.

The major incumbents understood this early. Betterment’s public launch in 2010 and Wealthfront’s subsequent investment-service launch established the category’s consumer identity. But the more consequential industry development came when large firms integrated automated portfolios into existing distribution networks. Charles Schwab’s 2015 launch of Schwab Intelligent Portfolios was emblematic: robo technology became a strategic component of a full-service platform, not merely a venture-backed challenger’s product.

The pandemic-era acceleration in digital adoption reinforced the trend. Clients became more willing to open accounts, transfer assets, and receive portfolio guidance through remote workflows. The question for established managers was no longer whether digital advice would exist. It was whether it would be owned as a direct capability or ceded to platforms controlling client data, cash balances, and account access.

Where automated systems stop

Robo advisors are well suited to standardized wealth-management tasks. They are less suited to ambiguous, interdependent decisions where a client’s finances cannot be reduced cleanly to an ETF allocation.

The limitations are not incidental. They define the boundary between automated portfolio management and comprehensive advice.

A robo advisor may struggle to handle:

• concentrated stock positions created by founders, executives, or inherited wealth;
• stock-option planning, restricted shares, and liquidity events;
• estate structures, trusts, charitable vehicles, and cross-generational transfer planning;
• private-market commitments and the liquidity planning required around capital calls;
• complex tax jurisdictions, business ownership, or partnership income;
• family dynamics that shape whether a theoretically sound recommendation will actually be followed.

There is also a market-structure limitation. Most robo-advisory portfolios are designed around liquid, listed securities. Their rebalancing logic works most cleanly when prices are observable and assets can be traded efficiently. The model becomes less straightforward when portfolios include real assets, private credit, private equity, direct real estate, or other holdings with infrequent valuations and constrained liquidity.

That does not diminish the utility of automated investment platforms. It places them correctly in the wealth-management stack.

The central risk is not that an algorithm will inevitably make a visibly irrational trade. The larger risk is false completeness: the client assumes a polished digital experience has solved a planning problem that remains unresolved. A well-designed platform can make portfolio management more efficient. It cannot infer a client’s unstated liability, family obligation, or impending liquidity event.

The strategic endpoint: automation becomes the baseline

Robo advisors have matured from a category novelty into a baseline capability for wealth platforms. The essential functions—digital onboarding, risk profiling, ETF portfolio construction, automated portfolio rebalancing, and tax-aware trading—are increasingly expected components of a modern advice architecture.

This does not mean every asset manager must become a pure-play robo provider. It means the unbundling of advice has become permanent. Portfolio implementation is becoming cheaper, more automated, and more tightly integrated with custody and account infrastructure. Human advice, where it remains valuable, must concentrate on judgment, complexity, behavioral counsel, and the coordination of a client’s full balance sheet.

For firms, the margin pressure is real. The standardized portion of wealth management will continue to experience fee compression because technology has made it scalable. But the product opportunity is equally real: automated portfolios can serve as the entry point for banking, planning, lending, retirement, and higher-touch advisory relationships.

The robo advisor is therefore best understood not as a substitute for every advisor, nor as a black box that outsmarts markets. It is the industrialization of routine portfolio management. And once clients experience that service at a fraction of legacy pricing, the rest of the wealth-management industry has to explain exactly what its additional fee is buying.