One platform for claims, fiduciary work and every language clients speak.
The two whiteboard sessions and the meeting notes point to the same product: an agentic, multilingual progressive web app that takes first-line work off two founders and connects clients, repair shops and insurers. This page walks through it in the order you asked: six questions, each answered first and then open to explore.
Strong for intake, documents and routing. Not for advice or coverage decisions. Regulatory status needs a check first.
See chapter 1Pathway A as the core for speed and flexibility, with Graph and Entra where clients already live on Microsoft 365.
See chapter 3Start with U2 for PFR's repair shops, then add U1 claims on top of the same shops and documents.
See chapter 5Who is who, and where the work piles up
Elk Consulting is two people with two client bases. SFA carries the insurance clients and a language gap. PFR carries the fiduciary work and the repair shops. The overlap is the product: repair shops need claims, claims need repair shops.
How the whiteboards read now that the notes exist
The first map flagged several uncertain readings. The notes resolve most of them. Two readings changed: SFA is not employed by Helvetia, and the Zürich row on photo 1 most likely means the insurer.
| On the board | Reads as | Status |
|---|
Is an agentic, multilingual PWA viable for Swiss insurance and fiduciary work?
Yes, for the work around the decision, not the decision itself. Intake, evidence, documents, translation and routing are repetitive and well suited to agents. Coverage, advice and approvals stay with licensed humans.
A capacity bottleneck
SFA receives about 100 calls and emails a day on pension funds, car policies, accidents and liability. He has done it by hand for over 15 years. Fact
A language gap
About 35% of his clients are Arabic-speaking, a segment Helvetia cannot support. He handles those personally, PFR handles the rest. Fact
Manual document chasing
PFR logs into many client portals to download receipts and documents. He knows of no API. Data sits mostly on SharePoint, books in Bexio. Fact
Select a row to see the reasoning. Ratings are my judgement from the notes. Assumption
Elk holds broker agreements with four insurers, which points to untied (ungebunden) status. FINMA's rules say untied intermediaries must be in its public register and must not be bound to a single insurer. Only PFR is entered in the commercial register, and SFA does 80% of his work for Helvetia. Confirm the register entries and the tied or untied question with counsel. FINMA: Versicherungsvermittlung
Six hypotheses the pilots exist to test. Tick them off as evidence comes in. Proposal
What does it save, what should it cost, and what do stablecoins add?
The value is freed expert time, not new revenue on day one. Two workloads carry the case: SFA's daily contacts and PFR's document collection. Cards and TWINT should lead at launch. Stablecoins fit as an opt-in rail, not as the foundation.
SFA · daily contacts
PFR · document collection
Cost and rate
Value is time freed, priced at the loaded hourly cost. It turns into cash only when that time is redeployed to billable or revenue work. The defaults are placeholders to replace with measured numbers in the baseline phase.
SaaS pricing with low upfront cost: one base plan plus modules. Prices are starting hypotheses to test in the pilots. Proposal
Base plan CHF 99 a month includes 3 users, extra users CHF 19. Payment fees are passed through at the processor's rate.
What they add
- No chargebacks: the payer authenticates in their own wallet.
- Cross-border payers settle in USD-denominated USDC while the payout lands in CHF in the Stripe balance.
- One checkout and one reconciliation, because the rail sits inside Stripe.
- Appeals to tech-forward clients and keeps the platform open to future rails.
Where they stop
- Keep insurance premiums outside the platform. Insurers collect them directly.
- Stripe lists Switzerland in private preview for stablecoin payments and caps one payment at USD 10,000. Confirm access before promising it.
- Refunds go back as stablecoins to the original wallet.
- Book the CHF value at receipt in Bexio and keep wallet and sanctions screening with the processor.
| Rail | Best for | Currency | Recurring | Disputes | Watch out for |
|---|
Which rail should this payer start with?
Sources: Stripe stablecoin payments and Stripe TWINT, checked 15 Oct 2026. Provider terms change, so verify before launch.
Today every Arabic request goes through one person. The agent can take the first line in Arabic, prepare the German or French version the insurer needs, and hand SFA only the cases that need judgement. The same capacity then serves Russian and Chinese speakers without hiring.
| Language | Screens | Conversation | Documents | Insurer-facing |
|---|
Ratings are my starting view, to be tested with native speakers in the pilot. Assumption Anything sent to an insurer stays in German or French, with the client's original wording kept next to it.
Portal logins can be automated, but no single method covers every portal. Use the most structured route each portal allows and fall back step by step. WebMCP is for 7ST's own PWA today, because third-party portals do not expose it.
| Method | Robustness | Coverage | Setup effort | Compliance risk | Role here |
|---|
Sources: WebMCP overview (W3C community draft, Chrome early preview since Feb 2026), Playwright MCP.
Open stack or Microsoft stack?
Build on Pathway A and add a Microsoft bridge. The open stack gives the lowest upfront cost, any model provider and any payment rail. Microsoft Graph and Entra cover SharePoint and sign-in where clients already work in Microsoft 365. Azure Switzerland North stays available if an insurer demands residency.
Drag the weights to what matters most to you. The scores are my judgement. Assumption
How do U1 and U2 work, step by step?
U1 turns a damage report into a reviewed repair offer and a routed claim. U2 turns scattered portals into filed, booked documents. Both keep a human at the one step that carries liability or accounting risk.
Which pilot first, and how do we know it worked?
Run two sequential pilots: the collector first, claims second. The collector carries no insurance regulation, solves PFR's problem fastest and builds the shop network that claims then reuse. Each pilot has a gate with measurable exit criteria.
Top risks
Which existing 7ST product should we build on?
Score candidates against eleven criteria, and let three must-haves veto. Where I have no evidence the cell stays blank, so the ranking shows how much of it is backed by facts. Click a cell to cycle its score.
| # | Criterion | What a 2 looks like | Weight |
|---|
Seven decisions to close
None needs code. All of them decide whether the first pilot can start in week one. Tick each as it closes.
Readiness
Sources checked on 15 Oct 2026. Stripe stablecoin payments · Stripe TWINT · FINMA insurance intermediaries · WebMCP · Playwright MCP · Bexio API
Figures marked Assumption are placeholders, not measurements. This page is strategy support, not legal, tax or investment advice. Regulatory points need confirmation by counsel.