Multimodal Q2 2026: Three New Reports, AgentFlow Updates, and More

June had no conference booths, and it didn't need them. This was a heads-down month: three new reports for credit union and bank decision-makers, two press releases picked up by CUInsight, new Main Street AI episodes every week, and steady product work that credit unions are already testing and shaping. Here's what happened.

AgentFlow Product Update: June and July Ships

The last two months of product work focused on one theme above all: trust. Every answer AgentFlow produces is now easier to trace back to the source, easier to review, and easier to govern, plus there's a new way to get documents in.

Answers you can trace back to the source

AgentFlow's agents read documents and produce structured results: extracted fields, decisions, and written reports. Those results are now traceable. Reports come with citations attached to each claim, and every citation opens to show the exact supporting evidence from the source document, along with a trust indicator showing whether the support was verified deterministically or assessed by the model. Operators no longer have to take an AI-written answer on faith. They can click any claim and see precisely where it came from.

Ask your documents, by typing or by voice

Reviewers can ask questions about a document set in plain language and get sourced answers back. Answers now stream in progressively; every answer is backed by citations to the specific files it drew from (including decision results, so you can ask why a document was flagged or approved and see the reasoning), and voice mode grew up: background-noise filtering, Space-bar push-to-talk, and instant interruption of a spoken answer.

Email intake

Documents no longer have to be uploaded through the app. Each workflow can now have its own email address. Forward a document to it, and AgentFlow turns that email into a processing run automatically, body text and attachments included, with built-in sender verification so a spoofed address can't slip documents into a workflow.

Enterprise-grade governance and collaboration

Only a run's owner (or an admin) can edit outputs or approve and reject; everyone else sees an accurate read-only view. Simultaneous edits are caught and flagged instead of silently overwriting each other. Users can manage their own multi-factor authentication, and admins can reset it when needed.

A sharper review workspace

Reviewers now get clear, persistent confirmation of exactly what was saved when they correct an extracted field, guided diagnostics that point straight to low-confidence or failed extractions, and first-class handling for more file types, including text, Markdown, HTML, SVG, and ZIP archives.

Reliability and polish

Expired document links refresh automatically, workflows gain proper lifecycle controls (publish, unpublish, archive, restore), workflow setups can be exported and duplicated, and Playbooks are now available to everyone on the platform.

Coming soon: a refreshed, faster home experience that puts the work needing attention front and center, plus a friendlier in-app support experience.

Three New Reports for Credit Union and Bank Leaders

June was the biggest research month in Multimodal's history, with three reports built for the people who have to make the case for AI inside their institutions.

2026 Agentic AI Field Report: 10 Mid-Market Insights

What 445 sales conversations told us about agentic AI adoption in mid-market finance, including the number nobody publishes: at one $4 billion credit union, 97% of incoming indirect auto loan packets arrive incomplete. Read the report here.

The Board-Ready AI Business Case for Credit Union Lending Automation

A board framework for credit unions: an ROI model built on five financial levers, a fair-lending and risk governance checklist aligned to NCUA, NIST, and SR 11-7, a 90-day roadmap with board gates, and a board-resolution template. Built on results from FORUM, Centris, and Del-One. Check the report here.

The True Cost of Manual Document Processing in Credit Unions

Data from 70 credit union deployments. A $500M credit union spends $4M to $6M per year on manual document processing. Deployments showed an 82% median processing-time reduction, 69% lower cost per document package, and a 150 to 300 basis point efficiency-ratio lift in 12 to 18 months. Find the report here.

In the Press: CUInsight and NCUA's 2026 Supervisory Priority

CUInsight published two press releases on our June research, tied to a milestone worth pausing on: the NCUA named AI a supervisory priority for 2026, the first time it has done so. Both feature commentary from our Founder and CEO, Ankur Patel.

• New report gives credit union boards a decision framework for AI in lending. Read the press release here.
• New report quantifies the hidden cost of manual document processing at credit unions, most of it invisible to the board. Read the press release here.

Main Street AI: New Episodes Every Week

The Main Street AI podcast published new episodes on June 2, 9, 16, 23, and 30, with guests including Chris Ortega, Courtney Rowan, Michael Heller, Randy Salser, and Phil Caputo. The second edition of the Main Street AI newsletter also went out on June 4.

Alongside the podcast, the blog kept pace, with posts covering AI loan approval, agentic AI platforms, private equity AI tools, and why copilots fall short in credit union lending, among others.

Advisory Boards Are Launching

Our Credit Union Marketing Advisory Board is off to a strong start, with the first sessions kicking off in early July. Members who have confirmed participation include Courtney Rowan from Citadel Credit Union, Chris Palumbo also from Citadel Credit Union, Hashim Forrester from Wescom Credit Union, Bhavani Palukuri from Patelco Credit Union, Shawn Dunn from WSECU, Jon Douglas from Credit Union of America, Craig Boivin from UMassFive College Federal Credit Union, Autumn Rainwater from Kelly Community Federal Credit Union, and Aaron Cain from Credit Union 1.

Additional advisory boards are in motion, and we'll be highlighting members in the coming months.

Pipeline: The Filene Research Institute Trial

Through the Filene Research Institute's trial, participating credit unions tested AgentFlow for over six weeks and provided structured feedback. We're actively working on improvements based on what they told us.

New customer stories are also in the works, and we'll share them here as they're ready.

See You in the Next Quarter

June proved the cadence doesn't depend on conference season: three reports, press recognition tied to a first-ever NCUA supervisory priority, weekly episodes, and a product that credit unions are actively testing and shaping. July is already moving. See you next month.