Lending operators joined a live agentic AI webinar, signaling a shift from evaluation to execution.
The Filene FiLab Agentic AI Discovery Test is underway with cohort sessions and committees forming.
AgentFlow now automates document evaluation against predefined compliance and eligibility rules.
New execution management features transform AgentFlow into a collaborative team workspace.
Multi-model support enables per-step model selection from OpenAI, Anthropic, or Google.
Get 1% smarter about AI in financial services every week.
Receive weekly micro lessons on agentic AI, our company updates, and tips from our team right in your inbox. Unsubscribe anytime.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
February was about operator engagement and product velocity. We brought lending leaders together for a live conversation about agentic AI in lending, continued our research collaboration with the Filene Research Institute, and shipped the biggest AgentFlow product update since launch. Here is what moved the business forward and why it matters for teams building in financial services.
AI Agents and Lending: A Live Conversation With Operators
On February 18, we hosted a 60-minute live webinar titled "AI Agents & Lending: Unlock Your Competitive Edge." The session brought together lending practitioners from across the industry to discuss where agentic AI is delivering the highest ROI in lending operations today.
Panelists:
Ankur Patel, Founder and CEO, Multimodal
Erin Schmidt, VP of Mortgage Operations, FORUM Credit Union
Thomas Shaw, Chief Marketing and Technology Officer, OCMBC, INC.
The framing was straightforward. With rates easing and application volume climbing, as panelists discussed, lenders who cannot process faster lose deals. The conversation centered on three themes: identifying the highest-ROI agentic AI use cases in lending, understanding what AI systems need to verify for compliance, and building implementation roadmaps that get first workflows live in weeks rather than months.
What made the session different was the composition of the panel. These are practitioners running lending operations at scale. The fact that credit union and mortgage leaders chose to participate signals that the conversation around agentic AI in lending has moved from evaluation to execution planning. The question is no longer whether to adopt; it is how quickly teams can get into production.
FORUM Credit Union, one of the panelists' organizations, has already put this into practice. Their team used AgentFlow to automate document classification, data extraction, and underwriting acceleration, boosting loan processing volume by 70%.
Watch the full webinar recording here. That operator momentum extended into our research work as well.
Filene FiLab Agentic AI Discovery Test: Now Underway
The FiLab Agentic AI Discovery Test, a research collaboration between Multimodal and the Filene Research Institute, is now actively underway. Multimodal serves as both an education partner and a build partner for the program.
The program brings together credit unions to explore how agentic AI can improve member service and operational efficiency. Cohort education sessions have launched, and topic-based committees focused on lending, onboarding, and servicing are forming. Three generalized proof-of-concept workflows will be co-designed with these committees using AgentFlow sandbox environments.
The program runs from January through July 2026, with research findings to follow.
This is a research collaboration by design. Filene's mission is to strengthen the credit union system through rigorous, independent research. The Discovery Test is vendor-agnostic and structured to support that mission. The goal is to produce actionable findings that any credit union can use, regardless of platform or provider.
For more on the partnership, read our detailed breakdown here.
AgentFlow Product Update: Biggest Release Yet
The latest production release marks the largest single AgentFlow update we have shipped. Here is what changed.
Decision AI
AgentFlow now includes a new agent capability that automatically evaluates documents against predefined rules. This covers eligibility checks, compliance criteria, and credit assessments. Where teams previously relied on manual checklist reviews to validate documents against institutional guidelines, this capability now handles that evaluation programmatically.
The backend is fully operational, with a dedicated UI tab coming in the next release. In the meantime, results surface through report outputs within AgentFlow.
A dedicated UI tab is coming in the next release. In the meantime, results surface through report outputs within AgentFlow.
Execution Management
This release introduces a full execution management suite that turns AgentFlow into a proper team workspace:
Lock and unlock: Protect work in progress during review so nothing gets overwritten.
Archive and restore: Move completed executions to the archive while keeping them accessible for reference or audit.
Assign and request access: Route executions to specific teammates. Admins can bulk-assign work across the team.
Notifications: Get alerts when you are assigned work, when someone requests access, or when actions are taken on your executions.
Refreshed executions page: Cleaner layout with real-time status updates.
This matters because AgentFlow is increasingly used by teams, not individual operators. Execution management gives those teams the controls they need to collaborate on AI-driven workflows without losing visibility or accountability.
Multi-Model Support
Workflows can now run models from OpenAI, Anthropic, or Google. Each agent step within a workflow can use a different model, giving teams flexibility to match model strengths to specific tasks. As of the February release, the default model was updated to GPT-5.2.
Workflow Updates
Several quality-of-life improvements shipped alongside the major features:
Workflow search: Find workflows by name instantly.
Execution jump: Navigate directly from a workflow row to its executions.
File upload configuration: Choose between single-file or multi-file upload modes per workflow.
Import and export: Move workflow configurations between environments.
Stability and Reliability
Faster executions page load times.
Report downloads now export all rows, not just the visible set.
Failed cases display a clear error view instead of a blank page.
Smoother login and logout transitions.
CSV and Excel Support
AgentFlow can now process CSV and Excel files alongside PDFs and images. This expands the range of document types teams can run through automated workflows without manual conversion.
Looking Ahead: Where to Find Us Next
NACUSO Reimagine 2026 (Apr 27-30, Lake Buena Vista, FL)
Nicholas Bianchi, our VP of Sales, will be representing Multimodal at NACUSO Reimagine. The event brings together credit union and CUSO leaders to challenge convention and define what comes next for the credit union movement. If you are attending, come say hello.
Milipol TechX (MTX) 2026 (Apr 28-30, Singapore)
Ankur Patel, our Founder and CEO, and Ishita Jaiswal, our Head of Growth, will be at MTX 2026. We will be demoing the AgentFlow platform and a solution we built for a government agency. If you are at MTX, come find us. We would love to connect.
Build AI Your Industry Can Trust
Deploy custom multimodal agents that automate decisions, interpret documents, and reduce operational waste.
February built on the momentum from January's update: traction, feedback, execution. The lending webinar showed that operators are ready to move on to agentic AI in lending. Filene is building the evidence base for credit unions evaluating this technology. And AgentFlow keeps shipping. More to come.