Updated
AI integration for brokers
AI that works inside
your brokerage.
Large brokers already put AI in front of clients: eToro builds its app around its AI agent, Tori; Robinhood launched Cortex; IG Australia opened its platform to ChatGPT in read-only mode. In the UK, the Bank of England and FCA found in 2024 that 75% of financial firms were already using AI.
We build AI into the systems a broker already runs — the client portal and CRM, the trading terminal, payments, and the dealing desk — with the controls regulators expect: PII redaction, audit logs, and a human in the loop for every compliance and money decision.
In short
- Nima Dorostkar's team integrates AI into brokerages: support agents, KYC review assistants, fraud and AML monitoring, dealing-desk copilots, churn prediction, and trader assistants.
- Every AI feature runs with PII redaction, audit logs, and a human in the loop for compliance and money decisions.
- Models can be Claude, GPT, or open-weight models hosted on the broker's own servers.
What we build
AI for brokers:
what's included.
Support agent
Answers account, funding, and platform questions in the client's language around the clock, and hands over to a person with the full conversation.
KYC review assistant
Pre-checks documents, flags mismatches with the application, and summarises each case for the compliance officer who decides.
Fraud & AML monitoring
Finds unusual patterns across deposits, withdrawals, devices, and trading — chargeback risk, bonus abuse, and multi-accounting.
Dealing desk copilot
Scores flow for toxic patterns, summarises exposure, and proposes routing and hedging changes for a dealer to approve.
Retention & churn prediction
Scores which funded clients are drifting away and tells the retention desk who to call and why.
Trader assistant
Market summaries, economic-calendar context, and trade-journal insights inside the terminal — educational, not personal advice.
How we build it
Engineered for
real money, from day one.
- 01Start with one workflow
Begin with the highest-volume task — usually support or KYC review — measure it, then expand.
- 02Your data stays yours
PII is redacted before any external model call, and models can run on your own infrastructure.
- 03Grounded answers
Assistants answer from your documents and live account data through retrieval, and say so when they don't know.
- 04Audited end to end
Every prompt, response, and tool call is logged for compliance review.
Platforms & technology
- Claude
- GPT
- Open-weight models
- RAG
- pgvector
- Python
- FastAPI
- PostgreSQL
Questions
AI for brokers FAQ
Can a broker's AI assistant give trading advice?
A broker's client-facing AI assistant should stay educational: explaining platform features, market events, and the client's own account, not giving personal investment recommendations. We build assistants with guardrails that keep them within the scope the broker's compliance team approves.
Where does client data go when a broker uses AI?
Personal data is redacted before any request to an external model, and models can be hosted on the broker's own servers so client data never leaves its infrastructure. Every request and response is logged for audit.
Which AI use case should a broker start with?
Usually client support or KYC review. Both are high-volume, easy to measure, and keep a person in the loop. Fraud monitoring and dealing-desk copilots follow once the data pipelines they need are in place.
Can AI be added to an existing CRM or trading platform?
Yes. AI connects through the same APIs as the rest of the stack — reading the CRM, scoring live trade events, monitoring payments, or running inside the terminal — so nothing a broker runs today has to be replaced first.
Start a project
Talk to an engineer about
aI for brokers.
Tell us what you run today and what you need. You'll hear back from a senior engineer with questions, a proposed architecture, and a realistic timeline.