Enterprise AI · For wealth firms

From RM to CIO,
one assistant
for everyone.

Dafin is the AI assistant for private banks, wealth managers and WealthTech platforms — answering every employee's question from your firm's own data, with zero hallucinations and nothing ever leaving your perimeter.

Built for regulated environments
MiFID IIFINMAGDPRSOC 2
D
Dafin
Answered in 1.4s
Is the Schmidt portfolio still within its 25% equity limit YTD?
Yes. Current equity exposure is 23.4%, peaking at 24.1% on 14 March. Within limit throughout the period.
Sources
↳ PMS database · positions EOD
↳ Market data · 252 trading days
↳ IPS-Schmidt-v4.2.pdf · §3.1
Built on trust, not on guesses

Answers you can sign off on.

Whether you're mid-call, mid-rebalance, or mid-review — the data you need is in your firm, it just isn't yet in front of you. Dafin closes that gap with answers you can actually use: calculated from your verified data, every figure cited, never invented.

The Dafin approach
Audit-ready
Verified. Cited. Defensible.
  • Calculated from your verified data
  • Every figure cited to its source
  • Full reasoning trace, logged for audit
  • If the data isn't there, Dafin says so
  • Ready for MiFID and FINMA review
The generic AI approach
Not auditable
Plausible. Confident. Risky.
  • Generated from probability, not data
  • Sources often fabricated or omitted
  • No reasoning trace to review
  • Confidently wrong when data is missing
  • A risk your compliance team can't sign off on
Built for everyone, in every firm

One assistant. Every desk. Every kind of firm.

Not a specialist tool. Not a sidecar bot. The same Dafin works across the front, middle and back office — and adapts to the shape of your firm.

Relationship Manager
8:47 AM · before a client call
Summarise the Müller portfolio YTD vs MSCI ACWI, and flag anything they're likely to ask about.
Portfolio Manager
10:12 AM · monthly rebalancing
Where is allocation drift largest across my Tier 1 mandates?
Risk & Compliance
2:34 PM · quarterly review
Which clients are due for MiFID suitability review this quarter?
CIO
4:55 PM · investment committee
Which mandates are misaligned with the December house view on rates?
Trader
9:03 AM · pre-open
Show my desk's net JPY exposure and where we're closest to limits.
Operations
5:42 PM · settlement
Any reconciliation breaks today over CHF 10k still open?
Client Reporting
All week · Q4 letters
Draft the Q4 letter for the Müller family in our usual house style.
CEO
6:15 AM · before the board
Top three changes in our book this quarter the board should know about.
And built for three kinds of firm
FOR · PRIVATE BANKS
Private banks
Enterprise-grade AI inside regulated banks — for RMs, CIOs, traders, risk and compliance. Built around MiFID II and FINMA from day one.
Read the use case →
FOR · WEALTH MANAGERS
Wealth managers & family offices
For independent wealth managers, EAMs and family offices managing complex multi-jurisdiction structures.
Read the use case →
FOR · PLATFORMS
Portfolio platforms & WealthTech
White-label, REST API, integrate Dafin reasoning into your platform. For PMS vendors and B2B WealthTech providers.
Read the use case →
0
Bytes of client data sent outside your firm
100%
Of answers cited and traceable to source
100%
User adoption across deployed teams
99.9%
Uptime SLA on every enterprise deployment
For your CISO, CTO and compliance team

A wall between your data and the model.

Generic AI tools send your data to the LLM. Dafin doesn't. The language model sits outside your firm and only sees the question — never the client positions, the IPS document, or the answer it generates from them.

Inside your firm
Where your data lives
  • PMS database
  • Custodian feeds
  • KYC & compliance
  • Risk models & documents
  • Dafin reasoning engine
The Wall
question only · no data
Outside
Where the model lives
  • Language model (interprets)
  • Never sees client data
  • Never trained on your firm
  • Never stores anything
  • Returns structure only
Built for regulated environments
MiFID IIFINMAGDPRSOC 2
Frequently asked

Questions teams ask before they sign.

How does Dafin ensure my data is never used for training?
Architectural separation between data access and model training. Your data is used only for retrieval and reasoning at query time, never to train any model.
How long does implementation take?
+
What makes Dafin different from ChatGPT Enterprise or Copilot?
+
Can Dafin be deployed on-premises?
+
How does Dafin avoid hallucinations?
+
20 minutes · with a founder

See it on your own data. In twenty minutes.

Book a focused demo with our founding team — no slides, no SDR, no qualification call.

Book a 20-min demo
Speak with a founder
Live answers, not slides
Walk away with a plan