ChatGPT and Claude are excellent tools for internal team productivity, prototyping and drafting. They are not adequate for regulated broker workflows — dealing desk, retention, KYC orchestration, surveillance. Firm-owned AI systems are non-negotiable for anything that touches client-facing regulated activity, produces evidence a regulator will pull, or represents a competitive moat.
The right answer is both
Every broker should be using frontier LLMs (ChatGPT, Claude, Gemini) for team productivity — research, drafting, code assist, meeting notes. That is table stakes.
For any AI touching a regulated workflow — dealing, retention, KYC, comms, marketing — a firm-owned system is required. Not because ChatGPT can't do it, but because the compliance envelope, the audit trail, the firm-specific tuning and the IP ownership all matter.
Related questions
Can we just wrap ChatGPT?
For internal productivity, yes. For regulated client-facing workflows, no — the envelope, the tuning and the ownership are not vendor-provided.
What about ChatGPT Enterprise?
Excellent for internal use. Still not appropriate as the runtime for regulated broker workflows.
Cost comparison?
ChatGPT Enterprise per-seat is cheap; a firm-owned system for a regulated workflow is a six-figure build. The comparison is only fair inside a specific workflow.
Where does bespoke start earning?
The moment a workflow becomes competitive differentiator or falls under regulator scrutiny.
Price of bespoke?
£15k Diagnostic → £75k+ Build → from £8k/mo Operator.
The right answer is not ChatGPT versus owned AI. It is ChatGPT for internal productivity and owned AI for the regulated, competitive layer.
The £15k AI Diagnostic maps which workflows should be which for your firm.