AI copilots for wealth management teams: buy a platform or build your own?

What does a wealth management copilot actually do?

A copilot supports relationship managers and advisors with work that today sits across several systems: preparing client meetings from CRM and portfolio data, summarising notes and drafting follow-ups, answering questions over research and product documents, and flagging suitability or policy points for the human to decide. It supports the adviser; it does not decide for the client. For the use-case overview see AI for wealth and asset managers; for how an RM copilot is designed and controlled at a Swiss private bank see AI copilot for relationship managers at a Swiss private bank.

Which three types of solution can a wealth team choose from?

The examples below are named only to show what each category looks like. S-PRO has not evaluated these products, makes no claim about them and does not recommend them; check each vendor’s own site. Bank-specific limits in the last column are our reading of the constraints, not a legal opinion.

Swipe to compare →

What should a Swiss private bank check before choosing, whichever route?

• Banking secrecy and client data.Decide which client data classes the copilot may see. Many designs keep client-identifying data out of the model call by pseudonymising it before the call. Article 47 of the Federal Act on Banks and Savings Banks (BankA, SR 952.0) makes the intentional disclosure of a secret entrusted to a bank’s organ, employee, agent or liquidator a criminal offence, and negligent disclosure is punished by a fine. Whether a given data flow to a vendor or model provider is a “disclosure” is a legal question for the bank’s counsel; the fewer client-identifying data reach a third party, the smaller that question. This is not legal advice.
• Data residency and processing location.Ask every vendor, in writing, where inference runs, where data is stored, where logs go and which sub-processors are involved. Residency is an options-per-project question; do not accept “Swiss hosting” without the detail.
• FINMA Guidance 08/2024.In “Governance and risk management when using artificial intelligence” (18 December 2024) FINMA states that there is no AI-specific legislation in Switzerland and that technology-neutral, principle-based financial market law covers AI risks. It names model, data, IT and cyber, growing third-party dependence (for example model or cloud providers), and legal and reputational risks, and describes supervisory findings on governance, inventory and risk classification, data quality, testing and monitoring, documentation, explainability and independent review. It is guidance, not a statute; map your copilot to each risk and record it in your AI inventory. Source: FINMA press release.
• Outsourcing.If a third party runs part of the service, check whether FINMA Circular 2018/3 “Outsourcing – banks, insurance companies and selected financial institutions under FinIA” applies. It applies when a company mandates a service provider to perform a significant function independently and on an ongoing basis; whether a given copilot service is such an outsourcing is the bank’s assessment, per use case. Check finma.ch for a newer version of the circular.
• Human oversight and audit trail.Can you show who approved what the copilot suggested, with the sources it used? See AI pilot to production at a bank.
• Integration.Does it read CRM, portfolio and document data you already hold, with the same access rights as the user?
• Exit and change.What happens to prompts, logs and configuration if you change vendor or model?

Selection checklist: buy, embed or build?

Answer each question with your compliance and IT teams. The pattern of answers points to a route, not a verdict.

1. Process fit.Is your advisory process standard enough to adopt a vendor workflow as-is? Mostly yes: category 1 or 2. Mostly no: category 3.
2. Data perimeter.May client-identifying data leave your perimeter, and to which regions? Not without strict controls: favour category 3 with pseudonymisation, or a vendor deployment you can contract precisely.
3. Proprietary data.Does the value depend on proprietary data (mandates, research, house views)? Yes: category 3, or a hybrid with a custom data layer.
4. Engineering capacity.Do you have, or can you source, engineering capacity to validate and run it? No: category 1 or 2, or a delivery partner.
5. Supervision evidence.Can you evidence supervision (logs, approvals, source citations) to the standard your compliance function expects? Test it in the first pilot, whatever the route.
6. Staged approach.Is a staged approach possible: start with an embedded copilot for drafting and summaries, add a custom layer where proprietary data matters? Often yes; this is the hybrid route.

How does S-PRO help?

S-PRO is a Zurich-headquartered firm, founded in 2014, with 250+ engineers, R&D centers in Zurich and Łódź, and ISO 27001 and ISO 27701 certification. S-PRO is a Certified partner in the Claude Partner Network. Our team includes Claude-certified engineers and architects.

If you are a CIO, COO, Head of Digital or Head of Wealth Management weighing a platform against a custom copilot, S-PRO runs a 45-minute discovery call on one advisory workflow: we map your CRM, portfolio and document systems, apply the checklist above with your compliance team, and scope a pilot with human approval and an audit trail if the build or hybrid route fits. See our AI agent development service. Contact S-PRO, Löwenstrasse 45, 8001 Zürich.

Book a discovery call (45 min)