Intellodocs is a next-generation document management ecosystem. While traditional platforms like Dropbox act as "digital filing cabinets," Intellodocs is designed as an autonomous collaborator. My goal was to eliminate "Digital Housekeeping" by utilizing agents that reason, plan, and execute tasks on behalf of the user.
Lead Product Designer (AI Strategy & Systems Design)
Timeline: 2025
Scope: End-to-end UX/UI, Agentic Logic Mapping, Prototyping.
Despite 20 years of cloud storage evolution, the user burden has increased. Users are now "digital librarians" rather than "creators."
Friction: Information is siloed across platforms (WhatsApp, Email, Slack).
Cognitive Load: Manual naming, tagging, and folder-diving consume 20% of the workday.
The Gap: Current AI in storage is reactive (waiting for a search) rather than proactive (anticipating a need).
I moved the design framework from User-Centered Design (UCD) to Agent-Centered Design (ACD).
Software shouldn't just store a file; it should understand the intent behind the file and the workflow it belongs to.
Empathize: Users don't want to "organize files"; they want "projects to be ready."
Define: The core unit of value isn't the file; it's the Information Context.
Ideate: How can an agent "intercept" a workflow before it even hits the app?
I designed the Cross-Platform Intercept Workflow, solving the common friction of mobile-to-cloud filing.
A user receives a critical invoice via WhatsApp while commuting. Traditionally, this file would sit in a gallery or require 10+ clicks to file correctly.
Step 1: Intent Perception (The Intercept)
Through a system-level share-sheet or bot, the user forwards the document. The agent doesn't just see a PDF; it identifies a "Finance Event."
Step 2: Semantic Parsing (Cognitive Reasoning):
The agent "reads" the file. It identifies it as an Invoice, extracts the $1,200 amount, and matches the vendor to an existing entity in the "Project Orion" knowledge graph.
Step 3: Proactive Proposal (Shared Agency):
UI Pattern: The Recommendation Toast. > Instead of a blank screen, Nexus presents a notification: "I’ve identified this as a Project Orion invoice. I’ll rename it to '2026-01_CloudLink_Invoice' and file it in Finance > 2026. Confirm?"
Step 4: Execution & Feedback Loop:
Once confirmed, the agent executes the move. It then asks a follow-up: "This is due on Jan 15. Should I add a payment reminder to your calendar?"
I introduced three specific UI patterns to solve the "Black Box" problem of AI:
To build trust, I designed a "Logic Trace" in the sidebar. Users can see the agent's "thinking": Scanning content... Matching vendor... Validating folder permissions... This makes the agent’s autonomy legible.
Not all tasks require the same level of AI control. I designed a toggle allowing users to set the agent’s "leash":
Assisted: Agent only suggests.
Co-Pilot: Agent acts, user confirms.
Autonomous: Agent acts and provides a "Daily Digest" of actions taken.
When the agent encounters ambiguity (e.g., two projects with similar names), it doesn't guess. It presents a Disambiguation UI, asking the user for a "steering" decision before proceeding.
By applying 14 years of UX leadership to this agentic model, we shifted the user's role from Operator to Supervisor.
Efficiency: Reduced "time-to-file" by 85% for cross-platform documents.
Error Reduction: Proactive folder matching reduced "misplaced" files by 40%.
Key Insight:
In 2025, the best UI is the one that manages itself. My role was to design the trust framework that allows that to happen.