The Complaint Machine

"The portfolio underperformed the stated benchmark by 340 basis points over three years while operating outside the agreed risk parameters during months 14 through 19, without documented client consent for the deviation. Fee increases in Q2 2025 were not communicated within the contractual notice period. Two structured products sold in March 2024 fall outside the suitability profile on file."

This complaint was not written by a lawyer. It could be generated in seconds by an AI agent with access to the client's portfolio history, advisory documentation, and risk disclosures.

This is what is coming.

The friction that protected banks

Here is something every head of compliance knows but rarely says aloud: the current volume of client complaints is not a measure of client satisfaction. It is a measure of client effort.

Most clients do not complain. Not because they are satisfied. Because identifying what went wrong, gathering evidence, articulating the issue, and pursuing it through the bank's process costs more energy than the likely outcome justifies.

The friction protects the bank. AI agents remove the friction.

What an agent sees

Give a client's agent access to their portfolio history, transaction records, advisory documentation, and risk disclosures. Give it a simple mandate: identify anything that does not look right.

It compares portfolio performance against the stated benchmark. It checks whether products match the suitability assessment on file. It identifies periods where the portfolio deviated from the mandate without documented consent. It reviews whether fee changes were communicated within contractual terms. It reads risk disclosures and asks whether the risks that materialised were among those disclosed.

Each of these is a routine analytical task. Together, they constitute a comprehensive review that most clients never perform and most banks assume they never will.

Analytical, not emotional

Today, a Swiss private bank handles perhaps a few hundred formal complaints per year. The process is designed for that volume.

Consider what happens when AI agents perform this analysis for a meaningful share of the client base. Not every client will act on every finding. But the ones that do will arrive with structured, evidenced complaints that are significantly harder to dismiss.

The complaints will not be emotional. They will be analytical. That is not a conversation the relationship manager is trained for.

Suitability under a microscope

The area of greatest exposure is suitability.

Under the Swiss Financial Services Act (FinSA, Art. 12–13), investment advice and portfolio management must align with the client's risk profile, investment objectives, and financial situation. In practice, suitability assessments are performed at onboarding and reviewed periodically. Between reviews, products may be sold that stretch the definition of suitability without formally breaching it.

An AI agent does not stretch definitions. It reads the suitability assessment and compares it against every transaction. It flags the structured product sold to a conservative client. The concentration risk that built up gradually. The FX exposure that was never part of the strategy.

Banks with rigorous suitability practices have nothing to fear. Banks that relied on the gap between what was documented and what the client would notice have a problem about to surface.

Get there first

The temptation will be to treat this as a legal risk — tighten complaint handling, prepare for higher volumes, hire more lawyers. That is necessary but insufficient.

The strategic response is to get there first. Audit your own book with the same rigour an AI agent would apply. Identify suitability gaps before clients do. Remediate proactively.

A bank that calls a client to say "we identified an issue and here is what we are doing about it" earns trust. A bank that receives a structured complaint from an AI agent and responds defensively loses it.

The general pattern

The complaint machine is a specific instance of something broader: AI agents will enforce suitability standards, fee disclosure obligations, and risk documentation requirements that the industry relied on clients not to enforce. Not because the standards did not exist. Because the cost of enforcement was borne by the client.

When that cost falls to near zero, every gap between what was promised and what was delivered becomes visible.

The banks that have been disciplined will be validated. The banks that relied on friction will be exposed.

This is not a future risk. The technology exists today. The banks that assume they have time are making the same bet they made about digital banking a decade ago.

Preparation is cheaper than remediation. And significantly cheaper than litigation.