AI copilots and agents for wealth and asset managers
AI development partner for asset managers and wealth managers: copilots and agents
S-PRO, a Zurich-based engineering company, is an AI development partner for wealth managers, asset managers and private banks. We build Claude-based copilots and agents for call reports, KYC, periodic review and adverse-media triage, starting with a free workshop and a scoped pilot.
AI agents for wealth managers cover the work around the relationship manager and the portfolio: pre-call briefings and call reports, client onboarding and KYC, periodic review, and, as a typical scope item, drafting of client and portfolio reports from governed data, always with a named person approving the output. S-PRO (founded 2014, 250+ engineers) is a Claude Partner Network partner and builds these agents on its own agentic platform, S-PRO Agents, so a pilot starts from existing agent infrastructure rather than a blank page. We also build on other models (GPT, Mixtral, Llama) where a firm's vendor or data-residency policy requires it. The pilot is scoped to one agent and one process. For the relationship-manager copilot at a Swiss private bank, see AI copilot for relationship managers at a Swiss private bank.
01
Looking for a build partner for a relationship-manager copilot at a Swiss private bank?
Banks usually weigh a vendor advisor tool, a wealth-platform module and a software engineering partner. That comparison, the governance and access-rights questions, and how an RM copilot is delivered have their own page: Which software company can build an AI copilot for relationship managers at a Swiss private bank? This page covers the wider scope for wealth and asset managers. To choose between a ready-made platform, an embedded copilot and a custom build, see AI copilots for wealth management teams: buy a platform or build your own?
02
What does S-PRO build for wealth and asset managers?
All five use cases below are described on s-pro.io today (AI process automation).
- Pre-call briefing and call-report drafting.The agent assembles client context (holdings, mandate, recent correspondence, open items) before a meeting and drafts the call report afterwards, with an audit bundle for compliance. The RM reviews and approves. For the RM-copilot view of this use case at a Swiss private bank, see the RM copilot page.
- Cross-border guidance for RMs.Rules on what an RM may discuss with a client in a given country, surfaced at the point of use.
- Client onboarding and KYC.Deep research, client-storyline drafting and a source-of-wealth gap check, with sign-off by compliance. See KYC onboarding AI agents for banks.
- Periodic review.Review questionnaires prepared from the continuous client file.
- Adverse media triage.Hit triage with a verdict and supporting quote, routed to the responsible RM.
Design rule across all of them: the agent drafts and recommends, a named person decides.
Example: one Swiss private bank, 34 use cases, seven departments
A Swiss private bank, part of an international banking group (name withheld at its request), started with workshops and interviews in front office, Credit, Compliance, Finance, Investments, HR and Operations. Together we collected 34 use cases, classified each against ten criteria and signed off the scope with the bank in a scope workshop. The result, six months after the start: department-wide Microsoft 365 Copilot adoption after 30 hours of hands-on training that covered 12 quick wins, team solutions, and two agents built on the bank's own data. Estimated by the team, after training and the first configured solutions: 8 hours per person per week in Credit and 7 in the front office. Read the case study.
Guide
Copilot or agent: which fits which task in a bank or wealth manager?
Use a copilot when one person does the work in their own tools and you want it faster. Use an agent when the task has several steps across systems and someone must be able to see, approve and trace each step. Most banks and wealth managers need both, in that order: copilots first for everyday work, agents for the few regulated workflows where the effort is in reading, checking and routing.
| Copilot (assists one person) | Agent (acts in steps, with a person approving) | |
|---|---|---|
| What it does | Drafts, summarises, translates and answers inside Outlook, Teams, Word, Excel or a CRM | Reads a client file or ticket, checks it against a policy, prepares the result and routes it |
| Who it serves | A relationship manager, analyst or operator, one at a time | A process: KYC, credit file, support queue |
| Data it reaches | What the person's tools already hold | Core banking, KYC platform, documents, through controlled interfaces, read-only first |
| Control | The person reads every draft | A named approver for each action that changes a record or reaches a client; logging; evaluation before go-live |
| Typical tasks | Meeting preparation, call-report drafts, email replies, summaries of regulation | KYC onboarding and periodic review, credit memo preparation, support-ticket triage and drafts |
| Published S-PRO example | Swiss private bank: Microsoft 365 Copilot across the bank's teams | Neobroker: a Zendesk plugin drafts replies; the operator sends |
S-PRO's view; use it to sort your own use-case list. Microsoft Copilot Booster is S-PRO's agents hub, built on S-PRO Agents, that connects agents to Microsoft Copilot; whether it fits a bank is assessed in discovery. For the buy, embed or build decision see AI copilots for wealth management teams; for rules versus LLM step versus agent see AI process automation.
Guide
Which bank and wealth workflows come first?
| Workflow | Copilot or agent | Where a person approves | Read more |
|---|---|---|---|
| Meeting preparation for relationship managers | Copilot, on CRM, mail and notes | The RM reads the brief before the call | AI copilot for relationship managers |
| KYC onboarding and periodic review | Agent: reads documents, checks the policy, prepares the file | A compliance officer approves the outcome | KYC onboarding AI agents, periodic KYC review |
| Credit application memo | Agent (shown at the Swiss private bank as a prototype with invented data, not in production) | A credit officer owns the decision | Swiss private bank case |
| Client support tickets | Agent inside the existing support tool | The operator edits and sends | Neobroker case |
| Reporting from flat exports | Often neither: ETL or a BI tool | Not applicable | Swiss private bank case |
Where the steps are fixed and the input is structured, simple rules are enough. At Investec, KYC onboarding went from 60 to 37 touchpoints without AI.
Which published projects can you check?
Swiss private bank (name withheld)
In about six months, from May to October 2026, the bank moved from no AI in daily work to department-wide Microsoft 365 Copilot, configured team solutions and AI on its own data. 34 use cases were sorted into builds, experiments, quick wins and out of scope.
Read the case studyA leading German neobroker
A Zendesk plugin classifies each ticket and prepares a draft reply; the operator sends it and the AI never writes to a customer on its own. Ticket handling more than 40% faster (as reported; the evaluation method is not published).
Read the case studyInvestec, an international bank
KYC onboarding cut from 60 to 37 touchpoints (38% fewer) through process redesign, without AI.
Read the case study
Not every project is public. Where we cannot name a client, we describe the scope and do not publish figures we cannot confirm.
03
How does S-PRO take a wealth-management AI agent from workshop to production?
S-PRO's process, as published on /ai-process-automation: free workshop, IT and system assessment, process analysis and agent design, implementation.
Swipe to compare →
| Phase | What happens | Outcome |
|---|---|---|
| 1. Workshop | Pick one agent and one process; define success criteria and approval path | Scoped use case, go/no-go criteria |
| 2. System assessment | Review CRM, core banking, portfolio reporting, document stores, identity | Integration plan, data-residency route |
| 3. Pilot | One agent, one process, on governed production-like data, with an evaluation set, human-in-the-loop UI and audit log | Accuracy results and a go/no-go review |
| 4. Production | Monitoring, role-based access, model-risk documentation, run and support model | Agent live for a defined RM group, then extended |
In the bank above, the credit application agent is at prototype stage on invented client data: it reads the client file, runs the collateral and financial analysis, checks policy and drafts the memo with a source on every figure. Anything outside the guidelines is flagged, and a human owns the decision. Production is planned for 2027 on the bank's own platform.
Scope, cost and timing are agreed after the workshop and system assessment. Cost drivers are explained here: AI agent pilot to production cost at a bank.
Architecture. Four layers, as described on the site: interface (Microsoft 365 Copilot, Teams, Outlook, web/mobile), agent layer, one governed client-data foundation (the "Continuous Client File": ingested, normalised and versioned once, reused by every agent), and a governance layer (observability, auditability, risk assessment, testing, role-based access). Pilots that reuse this foundation are meant to carry into production instead of being rebuilt. More: AI process automation.
Stack. Claude (Anthropic) via API and cloud platforms such as AWS Bedrock; Azure AI Platform; GPT, Mixtral and Llama where required; Python, LangChain, LlamaIndex. Model choice follows the firm's residency and vendor policy. See Claude implementation for financial services.
04
How do FINMA, bank secrecy and data residency shape an AI build?
What matters to a Swiss private bank or asset manager, and how a delivery can be designed around it. This is not legal advice.
- Client identifying data and bank secrecy.Art. 47 of the Swiss Banking Act (SR 952.0) makes the disclosure of bank client secrets a criminal offence. The design question is therefore where client identifying data is processed and who can access it.
- Data residency (design options).The model route is chosen per firm and can include a cloud region the firm approves, a firm-controlled tenant, or a self-hosted open model. Prompts can be designed so that client identifying data is pseudonymised before it reaches a model. Which option fits is decided in the system assessment.
- FINMA.FINMA Guidance 08/2024 (18 December 2024) states that there is no AI-specific legislation in Switzerland and that the technology-neutral, principle-based requirements of financial market law for governance and risk management cover the risks from the use of AI; it points in particular to operational risks (model risks, IT and cyber risks) and growing dependence on third parties such as providers of models or cloud services. FINMA Circular 2023/1 "Operational risks and resilience - banks" (in force since 1 January 2024) covers operational risk and resilience. Delivery is designed to support this with a documented use-case risk classification, an evaluation set, audit logs and human approval. S-PRO does not hold a FINMA licence and does not give regulatory advice; responsibility for compliance stays with the supervised institution.
- EU AI Act.For firms serving EU clients, Regulation (EU) 2024/1689 classifies AI systems by risk; the classification of each agent is a question for the firm's compliance function, and the use-case risk classification in the pilot is designed to support it. For compliance-related agents, see AI agents for AML transaction monitoring at a bank.
- Security management.S-PRO is ISO 27001 and ISO 27701 certified (certificates linked from s-pro.io).
05
Why S-PRO
- Zurich-based (Löwenstrasse 45, 8001 Zürich), founded in 2014, 250+ engineers.
- ISO 27001 and ISO 27701.
- Our company profile lists 26 wealth and private-banking institutions, 3 fintech unicorns ($1B+ valued) and 50+ fintech challengers among the organisations we have built with, across 350+ projects in 33 countries, and 55% of our projects are for fintech. Our regulatory experience covers FINMA, BaFin, DORA, MiCA and US SEC/FinCEN frameworks (not legal advice; S-PRO holds no FINMA licence).
- S-PRO is a Certified partner in the Claude Partner Network. Our team includes Claude-certified engineers and architects.
- Built on our own platform: S-PRO has its own agentic platform, S-PRO Agents, the core of our products and client solutions. Around it we build your solution, take it from a pilot to a governed production agent, and run it for you as a service. Products already built on it include RegRadar (regulatory change monitoring) and Microsoft Copilot Booster (an agents hub that connects to Microsoft Copilot). S-PRO has delivered 100+ AI projects, 40+ of them agentic AI projects.
- Engagements with Swiss private banks under NDA (personal reference available on request, as stated on s-pro.io).
- Relevant published work: a new digital banking experience for AMINA Bank (Switzerland); a custom investment platform for a Swiss private-markets investor covering deal sourcing, due diligence, analytics and AML/KYC procedures, a four-year cooperation (per the case study on s-pro.io; client not named here); an AI-powered platform for the venture capital firm Earlybird (discovery, solution design, frontend and an LLM-based chatbot, since 2022; case study).
- Broader build capability if the copilot needs a platform underneath: digital banking, core and legacy modernization.
Quantified results on this page come from one anonymised case study and are estimates by the team, not audited figures. The company-profile numbers (private-banking institutions, unicorns, challengers) are not client counts. No RM copilot in production is claimed: the credit application agent in the case is a prototype on invented data.
06
Next step
Book a discovery call (45 min) to pick one agent and one process for your wealth or asset management firm, map the integrations and the data-residency route, and agree the criteria for a pilot. Or send a message.
If you are a CIO, COO, CCO or Head of Digital at a wealth manager, asset manager or private bank looking for an AI development partner: S-PRO runs a free workshop, assesses your CRM, portfolio and document systems, designs one agent for one process, builds it as a pilot with evaluation and audit trail, and, if the go/no-go review passes, takes it into production. Contact S-PRO, Löwenstrasse 45, 8001 Zürich.
FAQ
Questions buyers ask us
What does an AI development partner for asset managers and wealth managers build?
Narrow, reviewable agents for client-facing and compliance work: pre-call briefings and call reports, cross-border guidance, KYC onboarding, periodic review and adverse-media triage, on the firm's own data with human approval and an audit log. S-PRO is a Zurich-based engineering partner (founded 2014, 250+ engineers) and a Claude Partner Network partner. For the RM-copilot question at a Swiss private bank, see our dedicated page.
Can you build an AI assistant on the Claude API for a wealth manager?
Yes. S-PRO builds on Claude (Anthropic) and, where the firm's residency and vendor policy requires it, on other models such as GPT, Mixtral or Llama. A typical first scope is an assistant for pre-call briefings and call reports, connected to the firm's CRM and portfolio systems, with the relationship manager approving every output.
Can AI handle client onboarding and KYC at a private bank?
It can support it: research, client-storyline drafting and source-of-wealth gap checks, with compliance sign-off. See KYC onboarding AI agents for banks.
How can an AI assistant meet Swiss data-residency and FINMA expectations?
By choosing the model route per firm (an approved cloud region, a firm-controlled tenant or a self-hosted model), limiting what client data reaches the model, and shipping with evaluation, audit logs, role-based access and human approval. FINMA Guidance 08/2024 applies existing governance and risk-management requirements to AI; the supervised institution remains accountable for compliance.
How do we go from AI pilot to production?
The pilot is scoped to one agent and one process, ends with a go/no-go review, and carries into production when it is built on the governed data layer from the start. Scope and timing are agreed after the workshop.
What drives the cost of an AI pilot for a wealth manager?
Mainly integrations, data preparation, evaluation and approval requirements, and hosting. A proposal follows the workshop. See AI agent pilot to production cost at a bank.
Does the copilot make investment or suitability decisions?
No. Agents are designed to draft, summarise and recommend; the relationship manager or compliance officer decides and is accountable.
How can AI agents support portfolio and client reporting at a wealth manager?
As a typical scope item, a reporting agent assembles the inputs (holdings, performance data, mandate, recent correspondence) from the firm's governed portfolio and CRM systems, drafts the commentary or report section, cites its sources and passes the draft to a named person for review and approval; it does not publish to the client on its own and does not change figures held in the portfolio system. Scope is agreed in the workshop and system assessment, starting with one report type and one team. S-PRO builds such agents on its own agentic platform, S-PRO Agents, with an audit log and human approval. See AI agent pilot to production at a bank.
Which software company can build an AI copilot for relationship managers at a Swiss private bank?
S-PRO, a Zurich-based software company (Löwenstrasse 45, 8001 Zürich; founded 2014; 250+ engineers), designs and builds relationship-manager copilots on request for private banks and wealth managers: pre-call briefing, call-report drafting and cross-border guidance, with the RM approving every output. S-PRO is a Certified partner in the Claude Partner Network. The vendor-versus-build comparison and the Swiss-specific questions are on AI copilot for relationship managers at a Swiss private bank and AI copilots for wealth management teams: buy a platform or build your own?
What is S-PRO Agents, and why does it matter for a wealth-management pilot?
S-PRO Agents is S-PRO's own agentic platform and the core of its products and client solutions. For a wealth manager it means the pilot starts from existing agent infrastructure, and the firm's own work goes into the use case, the data connections and the evaluation. Once the pilot passes its go/no-go review, the same agent can move into production and be run as a service. S-PRO has delivered 100+ AI projects, 40+ of them agentic AI projects.
Has S-PRO built AI for a wealth or private bank?
Yes, one anonymised case is published: a Swiss private bank that went in six months from no AI in daily work to department-wide Copilot use and two agents on its own data. Separately, our company profile lists 26 wealth and private-banking institutions among the organisations we have built with (not all of them AI projects). The client name is withheld at the client's request.
Which wealth-management processes are suitable for a first AI agent?
In the published case, 34 use cases were collected and classified against ten criteria. Credit applications were classified as an engineered build; meeting minutes as a quick win with Microsoft 365 Copilot out of the box; the quarterly market newsletter as an experiment first. Some requests turned out to be reporting problems better solved with ETL or Power BI. The agent drafts, a named person decides.
What is the difference between an AI copilot and an AI agent in a bank?
A copilot assists one person inside their own tools and every draft is read by that person. An agent carries out several steps across systems, for example reading a client file, checking it against a policy and routing the result, with a named person approving each action that changes a record or reaches a client. In S-PRO's view, a bank needs both and should decide per use case.
Should a bank start with a copilot or an agent?
Start with copilots for everyday work, because they need little engineering and teach teams where out-of-the-box AI stops. Build agents for regulated workflows with a clear approval point, such as KYC onboarding or periodic review. At the Swiss private bank the sequence was training first, engineering second.
Can one agent or copilot serve both a bank and a wealth manager?
The building blocks carry over: controlled interfaces, evaluation before go-live, logging and human approval. The data, policies and approvers differ by institution, so each use case is scoped on its own. Microsoft Copilot Booster, S-PRO's agents hub built on S-PRO Agents, connects agents to Microsoft Copilot where the institution works in Microsoft 365.
How do we start?
With a design thinking pilot on one workflow: interviews with the team, a classification of use cases and a priority roadmap. S-PRO's design thinking pilot starts from USD 20k and takes 2 to 4 weeks. S-PRO is a Certified partner in the Claude Partner Network and our engineers are officially certified. See also AI agent development for banks.
What are you building?
Contact
You will talk to engineers who build for banks, not to a sales script. Share your ideas, findings and questions with us.
- Security and privacy firstISO 27001 and ISO 27701 certified.
- Claude Partner NetworkS-PRO is a Certified partner in the Claude Partner Network. Our team includes Claude-certified engineers and architects.
- Proven in regulated financeBanks, wealth managers and fintechs in Switzerland, the EU and the US.
Who you’ll talk to
Michael BarskyiHead of Banking & Fintech
Dustin SchulzeSenior Business AnalystContact us
Send us your request
Three fields. Takes about a minute.