My Framer Site

AI Ecosystem

Creating visibility into an organization's rapidly expanding AI landscape.

Creating visibility into an organization’s rapidly expanding AI landscape.

AI Security

Visibility

Graph Visualization

Entity Mapping

Contents

01 The Problem

02 User Personas

03 Competitive Research

04 Dashboard Interaction

05 Interaction Design

06 Design Decision

07 Agent Investigation

08 Scalability

09 Taxonomy

10 User Flow

11 Impact

12 What I Learned

The Problem- AI adoption was growing faster than security visibility

The Problem

AI adoption was growing faster than security visibility

Unknown AI Platforms

Which AI platforms and tools are being used across the organization?

Invisible Agents

Which agents exist, where do they belong, and who is using them?

Unclear Access

Which applications, identities, and sensitive resources can AI tools access?

Hidden Relationships

Which relationships create exposure or expand the potential blast radius?

Existing product views could reveal applications and connections, but AI introduced new entity types and relationships that did not fit neatly into the existing model.

User Personas- Three audiences, one interface

SOC Analyst

Primary user. Investigates the AI ecosystem daily - scans for new entities, traces connections, and identifies exposure across platforms, tools, and agents.

Key features

Explorable ecosystem map

Agent drill-down & relationships

Popover + side panel investigation

Job to be done

"See which AI entities exist, how they connect, and where exposure requires my attention."

CISO

Secondary user. Needs a quick high-level overview of organizational AI adoption and exposure without deep investigation.

Key features

Aggregated mini-dashboard

Platform & tool summary

Active users & exposed apps count

Job to be done

"Quickly understand our AI footprint and where the biggest risks are."

App SOC

Secondary user. Focused on specific applications - needs to see which AI tools and agents access their apps and what permissions they hold.

Key features

Filter by specific application

AI-to-app connection mapping

Permission and access details

Job to be done

"Show me which AI entities touch my app, what they can access, and whether that creates risk."

Competitive Research

At the time of our research, most competitors focused on AI inventories or identity-centric agent maps. No single product offered a unified ecosystem view connecting platforms, tools, agents, and users in one explorable model.

Key Insight

Competitors addressed pieces of the AI visibility problem- agent identity, adoption tracking, or inventory management- but none combined all three layers: organizational overview, ecosystem mapping, and agent-level investigation.

Competitor

What worked

What didn't work

Our advantage

Competitor X

Agent-centric graphs showing identity, credentials, permissions, and blast radius

Focused on a single agent's context- no organizational overview or cross-entity relationships

A complete ecosystem map connecting all AI entity types- not just one agent at a time

Competitor Y

Usage trends over time, sanctioned vs. unsanctioned classification, KPI-to-list navigation

Relied on dashboards and lists without a visual map showing how entities connect

Visual ecosystem map with entity relationships- not just metrics and lists

Competitor Z

Centralized agent lists with risk scoring, detail pages, and remediation actions

Optimized for comparability and scale, but no progressive drill-down from overview to investigation

Three-level navigation: organizational overview, ecosystem map, agent investigation

Planning

Design Decisions

The result: A dedicated AI Ecosystem page combining what competitors split across products- organizational overview, visual ecosystem mapping, and deep agent investigation in one unified experience.

The result:

A dedicated AI Ecosystem page combining what competitors split across products- organizational overview, visual ecosystem mapping, and deep agent investigation in one unified experience.

Dedicated page over Discovery layer

AI entities needed their own information hierarchy, filters, and product identity-adding to the existing map would increase cognitive load.

AI entities needed their own information

hierarchy, filters, and product identity-adding

to the existing map would increase

cognitive load.

Overview dashboard above the map

CISOs get quick understanding; analysts get entry points into investigation. Each metric highlights relevant entities in the map.

CISOs get quick understanding; analysts get

entry points into investigation. Each metric

highlights relevant entities in the map.

Horizontal tool layout for sparse states

The same system needed to feel valuable with few entities and remain understandable as AI adoption expanded.

The same system needed to feel valuable with

few entities and remain understandable

as AI adoption expanded.

Entity-type grouping over vendor ownership

Entity-type grouping over

vendor ownership

Users should find tools by what they are, not who owns them-proximity, similarity, and recognition over recall.

Users should find tools by what they are,

not who owns them-proximity, similarity,

and recognition over recall.

Grouped agents with drill-down

Aggregated overview preserves the organizational picture; drill-down reveals individual agents and their application relationships only when needed.

Aggregated overview preserves the

organizational picture; drill-down reveals

individual agents and their application

relationships only when needed.

Dashboard Interaction- Hover to highlight, dim the noise, reveal the signal

Mini Dashboard

Hovering on a mini-dashboard metric highlights the relevant entities in the map and dims everything else. This interaction helps CISOs quickly isolate what matters- instead of scanning the entire ecosystem, they see only the platforms, tools, or agents that relate to the selected signal. The result is a cleaner, more focused picture of organizational AI exposure.

Interaction Design- Two levels of disclosure

A lightweight popover shows a concise summary-alerts, actions, and applications-while keeping the entity and surrounding relationships visible. Data-rich entities open a side panel with deeper information.

Popover

A lightweight popover shows a concise summary-alerts, actions, and applications seen in traffic-while keeping the entity and surrounding relationships visible.

Popover

Side Panel

Agent Drill-Down

Progressive Disclosure

Popover

A lightweight popover shows a concise summary-alerts, actions, and applications seen in traffic-while keeping the entity and surrounding relationships visible.

Popover

Side Panel

Agent Drill-Down

Progressive Disclosure

The interaction model supported investigation without requiring page navigation. Users could move from scanning the ecosystem to deep investigation within the same view.

Design Decision- Discovery Layer vs. Dedicated Page

Add to Discovery

✕ Rejected

Keeps all maps in one location and reuses an established product area.

+ Keeps all maps in one location

+ Reuses an established product area

- Adds another entity model to an already complex map

- Increases cognitive load significantly

Dedicated AI Ecosystem Page

Dedicated Page

✓ Chosen - Phase 1

+ Lower cognitive load - AI gets its own hierarchy

+ AI-specific exploration with focused filters

+ Tailored UI for AI entity relationships

+ Clearer product positioning for Sales & Marketing

+ Clearer product positioning for

Sales & Marketing

+ Room for the experience to evolve independently

+ Room for the experience to

evolve independently

The decision was based on four considerations: lower cognitive load, AI-specific exploration needs, a tailored UI for AI entity relationships, and clearer product positioning. The result was not simply a separate screen- it gave the emerging AI domain a clear place in Vorlon's product architecture.

Agent Investigation- From ecosystem overview to individual agent context

Agent Investigation

Selecting an agent opens a dedicated side panel with its own relationship map, related alerts, actions, access details, and timeline- all without losing the ecosystem context behind it.

Scalability- Designing for an ecosystem still emerging

Scalability

We built AI Ecosystem at an early stage of enterprise AI adoption. Some customers had very few AI platforms, tools, or agents. The same system needed to feel valuable with sparse data and remain understandable as AI adoption expanded.

Sparse state- Early adoption balance

Horizontal tool layout and balanced composition made the map meaningful even with few discovered entities, missing agents, or no platforms.

Dense state- Grouping and progressive disclosure

As the ecosystem grows, agents are grouped, relationships are revealed on demand, and filters narrow the view.

Roll-up- Information scent over completeness

Rather than treating missing data as empty states, available signals were designed as invitations to investigate.

Taxonomy- Group by how users understand entities

Group by how users understand entities

The AI Ecosystem follows strict rules to keep the experience predictable and trustworthy:

Entity Type Over Vendor

Platforms contain agents. Standalone tools appear together regardless of vendor. Users find tools by what they are, not who owns them.

Adaptive Layout

The map composition adjusts based on entity count and type. Fewer than 5 tools display vertically; larger ecosystems spread horizontally. Each edge carries semantic meaning.

Proximity & Similarity

Technically equivalent tools placed together are perceived as one category. Gestalt principles guided the spatial organization.

User Flow- Overview to Highlight to Hover to Drill-down to Side panel to Act

User Flow

Understand the situation, locate relevant entities, scan relationships, validate relevance, focus investigation, explore details and take action. This layered architecture balanced immediate clarity, visual context, and access to deep security data-without forcing all information onto the screen at once.

Impact- Product differentiation and sales enablement

Impact-

Users- Clear Security Model

Transformed fragmented AI activity into a model teams could understand and investigate-from overall footprint to single agent exposure.

Product- Competitive Advantage

No competitor offered one centralized view of the entire organizational AI ecosystem at the time of launch.

Sales- Visual Demo Story

Gave Sales a visual, immediate way to communicate that Vorlon extended beyond SaaS monitoring to the broader AI ecosystem.

Business- AI Security Positioning

Strengthened Vorlon's position as a platform that reveals and monitors the organization's complete AI environment.

Strengthened Vorlon’s position as a platform that reveals and monitors the organization’s complete AI environment.

What I Learned

Data products must work with both abundance and scarcity

Data products must work with

both abundance and scarcity

A new product category may initially have too little data to communicate its future value. The experience must remain useful and credible at both extremes.

A new product category may initially have too little data to communicate its future value. The experience must remain useful and credible at

both extremes.

How you group information shapes how users think

How you group information

shapes how users think

How you organize information determines how users think about the product. Grouping by vendor ownership vs. entity type completely changes whether users can find what they need-and that's a product decision, not a technical one.

How you organize information determines

how users think about the product.

Grouping by vendor ownership vs. entity

type completely changes whether users

can find what they need-and that’s

a product decision, not a technical one.

Show less when there's less to say

Not every selection deserves a full panel. Progressive levels of disclosure made exploration feel lighter while preserving access to detailed data.

Cognitive principles can drive both UX and business outcomes

Cognitive principles can drive

both UX and business outcomes

Separating AI Ecosystem into its own page was grounded in cognitive load theory-one complex map per view, not two. But the same decision that protected users from overload also created a distinct product capability that Sales could name, demo, and differentiate against competitors.

Next Project

Traffic Inspector- AI Conversation

Helping security teams investigate application traffic and AI conversations.

Next Project

Traffic Inspector- AI Conversation

Helping security teams investigate application traffic and AI conversations.

Next Project

Traffic Inspector- AI Conversation

Helping security teams investigate application traffic and AI conversations.

Next Project

Traffic Inspector- AI Conversation

Helping security teams investigate application traffic and AI conversations.