Event dossier · Paris · 3 June 2026
The production divide came into focus.
Across four industry talks, seven live demos, and two panels, the strongest through-line was not a model release. It was the movement of AI into the systems where financial work is already done: research, advice, portfolio construction, client intelligence, and risk.
This dossier is an editorial synthesis of the published programme and the verified event record. It separates what the source material supports from what would require a transcript or speaker-approved session recap.
The programme kept returning to the point where a model meets a real decision: inflation positioning, intraday signals, advisor preparation, portfolio construction, and investment research. The useful distinction was not AI versus no AI, but whether inference sat inside the operating loop or beside it as another screen.
A semantic client record, structured financial news, and production-grade web access all addressed the same constraint from different directions. Models are only as current, governed, and machine-readable as the information layer they can reach.
Neuralk and Databricks showed Seldon, a tabular foundation model designed to start from a strong pre-trained base rather than a bespoke model for every dataset. That changes the cost and speed calculation for financial teams with abundant tables and limited ML capacity.
The two panels compared shipped systems with proof-of-concept work across wealth management and agents. Regulation, confidentiality, controls, and human judgment were treated as architecture inputs, not as a compliance paragraph added after the demo.
The sessions were deliberately mixed rather than split into separate quant, wealth, data, and agent tracks. Grouped after the fact, they reveal the stack the room was actually discussing.
Who was in the room
The verified report describes 115 places held across 91 firms. Twenty-two percent of classified attendees came from the buy side, alongside banks, fintechs, data providers, researchers, and technology teams.
The public report shows the arithmetic, coverage limits, programme mix, partners, and curated gallery without exposing attendee names.
Open the verified reportThe practical question is no longer whether finance will use AI. It is which decisions deserve automation, which data can support it, and where a human must still own the judgment.
Paris put those questions on the same programme rather than isolating them by product category. That made the dependencies visible: an agent needs current context; a model needs governed data; a workflow needs an accountable decision owner.
Follow the next AI in Finance edition