New Product Roundup: What Launched Across Fund AI Tools This Month
The three weeks to 13 August brought a steady run of launches and updates across the fund AI stack, from deal sourcing through to diligence, client-facing assistants, and the enterprise infrastructure that sits underneath them. Here is what changed, grouped by function.
Deal sourcing and origination
Metal added a “similar deals” feature that surfaces past transactions from a firm’s own history that resemble a new target, matched on the firm’s own definition of similarity rather than public market comparables alone. The pitch is specific: instead of a generic AI answer built from public information, the feature pulls in a firm’s own thesis notes, outcomes, and flagged risks for each comparable, and copes with the fact that deal records are rarely complete or consistent. (source ↗)
Research and diligence
Marvin Labs shipped two updates. Private document upload lets users add their own files (PDF, plain text, or markdown, up to 500MB) so broker research, internal memos, and third-party documents get parsed the same way as a filing and made available to chat and research agents. Files stay private by default on Standard and Pro plans, with sharing available to Enterprise customers; the company says it does not train models on uploaded content. (source ↗) Separately, Marvin Labs made its Deep Research Agents schedulable, so they run on a set time (weekdays at 7am, for example) or trigger off a company publishing a new filing or earnings transcript, skipping any company with nothing new since the last run. (source ↗)
Rogo launched Deal Room, positioned as infrastructure for dealmaking. (source ↗) Separately, it added Claude Opus 5 to its model selector. Rogo’s applied AI lead said the model’s biggest gains showed up on longer-running work such as building and revising a full deck, with fewer formatting and slide issues. (source ↗)
Client-facing AI assistants
Unique introduced Context Memory, which builds a durable profile of a user’s role, communication preferences, and current projects across conversations rather than storing a full transcript, so financial professionals working the same accounts don’t have to re-establish context each session. (source ↗) The same update round (platform versions 2026.28 and 2026.30) also made the Code Execution tool generally available, letting the assistant run Python for precise calculations rather than estimates, added governance controls to its Agentic Table spreadsheet feature (locked columns, per-space formatting, comment threads), and expanded model choice to include Claude Sonnet 5, GLM 5.1/5.2, and the GPT-5.6 family. (source ↗)
Hebbia introduced Max, described as an AI team member built around firm-specific workflows rather than a general chat interface. It draws on a firm’s own data and financial-workflow “skills” to produce slides, reports, and models in house style, and supports email-based interaction for senior staff between meetings. It is rolling out to a small set of firms first. (source ↗)
Enterprise AI infrastructure
Snowflake launched Cortex AI Gateway, a centralised control layer for governing both first-party agents built on Snowflake and third-party agents from platforms such as Claude Code and Cursor. It centralises policy over which models, data, and tools an agent can reach (supporting over 100 MCP servers), gives a single record of agent activity, and enforces spending limits across teams and workloads. Snowflake paired the launch with integrations from 1Password, Aembit, Linx Security, Okta, SailPoint, and Saviynt for task-scoped agent access. (source ↗)
Data and market intelligence
OpenBB added Carbon Arc as a data integration, bringing over 210 alternative data assets, including card spend, web traffic, app usage, and foot traffic, into OpenBB with a consistent schema across datasets so different signals can be compared for the same company without manual matching. Carbon Arc is a data provider rather than a Qaike-tracked AI vendor, so there is no vendor page to link here. (source ↗)
A pattern worth flagging: model choice and agent governance showed up together this month. Rogo and Unique both widened the range of underlying models users can pick from, while Snowflake built the control layer to govern what those models and agents are allowed to touch. As firms run more agents against more models, the two are becoming a matched pair rather than separate concerns.
For the full list of tracked vendors and what each one does, see Qaike’s vendor directory.
This article is based on publicly available information and reflects Qaike’s own analysis and opinions. It is not intended to provide professional advice, endorsement, or a definitive assessment of any vendor or product. While every effort has been made to ensure accuracy, completeness and correctness cannot be guaranteed. Created by Qaike – Powered by AI.