Hub
All tools in one workflow
Cases, reports, redaction, portfolio analysis and documents in one place, set up around your workflow. Provided with a one-off set-up and an ongoing fee.
We are financial planners ourselves and built the infrastructure we work with every day for our own company. You can adopt it without any technical effort: the AI does the groundwork, calculations are traceable, and you approve.
hub.your-company.local
Hub
Server in the office
Cases
Case 2026-014 Keller08:40 Report draft v3 created with Assumptions_2026.json
Sample view with made-up data
The scientific background
In his bachelor thesis at the ZHAW School of Management and Law, Lukas Bachmann examined how well an AI assistant solves real financial planning cases. The basis were the two official exam cases of the Swiss federal diploma in financial planning ("eidg. Finanzplaner/in") from 2021 and 2024. They were solved by a language model with a curated knowledge base on Swiss pensions and taxes on one side, and by an experienced financial planner on the other. Both were scored on five dimensions of 10 points each.
LLM-Based Reasoning Assistants in Financial Planning – A Decision Quality Comparison with Professional Planners. BSc Business Administration, Banking & Finance.
On average, the AI assistant scored 8.3 out of 10 points, the planner 7.7. Per case it needed 5 to 7 minutes, the planner 79 and 92 minutes. In return it made more calculation errors: 5 incidents against 3.
Among other things, the assistant did not apply the BVG reduction for over-insurance and used parameters from the wrong year.
ScoreScale 0 to 10 points
- AI assistant
- 8.3
- Planner
- 7.7
Time per caseScale 0 to 90 minutes
- AI assistant
- 6 min
- Planner
- 85.5 min
Bars to scale per measure. Time ratio around 1:15 (2021 case) and 1:14 (2024 case).
Limitations
Two cases and a qualitative comparison with one planner: no statistical conclusion can be drawn from this. In addition, the solutions were scored by a language model acting as assessor (GPT-4o), with the official model solution as reference, not by an exam board. The results show a direction, not proof.
What we take from it
Fast and strong on content, but not free of errors. That is why we use AI as a sparring partner: it does the groundwork, calculations run through fixed methods, and the planner reviews before anything goes to the client.
Summary (PDF, 3 pages)A sparring partner, not a replacement
Nobody hands their job to the AI. Four parties, each with a clear task: we run the infrastructure, the AI suggests, the program calculates, and you decide.
Tap or hover over a role to see its tasks.
We set up the infrastructure, operate and maintain it. When your firm has a specific request, we build the solution for it.
- Set-up of hub, website or local server
- Operation, updates and backups
- Maintenance and support
- Further development on specific request
New report template built on request
You decide what the calculations are based on and how texts are written, and what goes to the client. The client relationship stays with you.
- Set assumptions, templates and tone
- Review suggestions and drafts
- Decide and approve
- Lead conversations and the client relationship
Report moved from draft to approved
The AI reads documents, recognises client data and drafts texts. What it delivers remains a suggestion until you approve it.
- Read and file documents
- Suggest client data for redaction
- Draft report texts in your tone
Suggestion: redact 4 items
Figures come from fixed calculation methods, not from the language model. The same inputs give the same result, the workings are in the appendix.
- Asset projections and pension gaps
- Tax effect of buy-ins and pillar 3a
- Overlaps and concentrations in portfolios
- Workings to read through
Same inputs, same result
Various modules, individually or as a package
Start with what you need and add more later.
Price on request: a one-off set-up and an ongoing fee.
Local AI
AI in your own office, no cloud
A local server with open language models and text recognition that runs offline. Client data stays in your office and is not sent to OpenAI, Anthropic or any other provider.
Website
Your site on our technology
A website for your firm on the same technical foundation as innovorsorge.ch, with a blog, two languages and embedded calculators.
Tools
Calculators and tests for your clients
Calculators and tests for your clients, such as a compound interest calculator, budget calculator, pension fund test or retirement self-test.
Custom
Whatever else you need
Whatever does not fit into one of the modules, we build with you as a dedicated solution.
What the work looks like
Four tools from our own daily work, with made-up sample data. Each demo plays once and can then be used directly.
Capture the case, set the assumptions, calculate deterministically. The AI drafts the text, you review and approve.
Hub / Reports / Case Andrea Muster
- Case captured
- Calculated
- Draft
- Approved
Case captured. Not calculated yet.
Case captured. The assumptions on the left can be adjusted.
Client data stays in your office
With the Local AI module, the language model and text recognition run on a server at your premises. Working with client documents needs no connection to an AI provider.
Your office network
- Workstation
- Local server with language model and text recognition
- Result in the hub
Documents do not leave the office.
Your office network
- Workstation
- Transfer over the internet
- Data centre of the provider
For example
- ChatGPT
- Claude
- Gemini
- Microsoft Copilot
Storage location and retention follow the provider's terms.
The website and tools for your clients run on the internet like any website. What stays local is the processing of client documents.
Built by financial planners, in use in our own office
innoVorsorge is Alessandrina Gull and Lukas Bachmann. We advise private clients on pensions, retirement and investments, and use the tools shown on this page every day.
Whatever does not hold up in daily work, we rebuild. Because you use the same infrastructure, these improvements reach your version as well.
More about usHow the introduction works
Conversation
We look at your workflows and clarify which modules fit. You then receive an offer with set-up and ongoing fee.
Set-up
We set up the hub, website or local server and apply your defaults, such as assumptions, templates and report tone.
Training
You and your team learn the tools on cases from your own daily work.
Operation and development
We run and maintain the infrastructure. New features from our own use are added continuously.
Frequently asked questions
Who calculates, the AI or a program?
Calculations run through fixed, deterministic methods. The AI captures, structures and writes. Figures such as asset projections or tax effects come from the calculation engine and can be traced.
Does client data leave my office?
Not with the Local AI module. The language model and text recognition run on a server at your premises, nothing is sent to OpenAI, Anthropic or other providers. The website and tools for your clients, on the other hand, run on the internet like any website.
Do I need technical knowledge?
No. We handle set-up, operation and maintenance. You work in the hub like in any other browser application.
Can I adjust assumptions and templates?
Yes. Assumptions such as return, inflation or retirement age, the scope and depth of reports and their tone are yours to set, as a default for the whole firm or per case.
Does the AI replace my advice?
No. It is a sparring partner: it makes suggestions and prepares, reviewing and approving stays with you. In the study behind this offering, the AI assistant made more calculation errors than the planner. Review is therefore a fixed part of every workflow.
What does it cost?
The price depends on the modules and the scope and is available on request. The model consists of a one-off set-up and an ongoing fee. We clarify both in the demo call.
Book a demo
In the demo we show you the tools on sample cases and clarify which modules suit your firm.
Price on request: a one-off set-up and an ongoing fee.
You can also reach us directly by email or phone: