AI Governance

A new Scytale module that discovers every AI agent and tool in an organization, shows what each one can reach, and gives security teams one place to govern them.

Company: Scytale
Role: Sole product designer
Platform: B2B SaaS, web
Status: In development

The problem

Most organizations have no idea how many AI agents they run.
Agents are created by teams, employees and inside SaaS tools, with no intake process.
Little visibility into what each agent does and what data it can reach.

88.4%

of organizations had at least one agent related security breach in the past 12 months.

The research

We distilled the goal and the requirements together with product.
Interviews with compliance and security leaders.
The key finding: users don't need another list of AI tools. They need to know which AI puts them at risk.
That's why we built the whole solution around risk.

Remote interview session with four participants

The user journey

From the first scan to a decision on every agent.

User journey from discovery to approving or unsanctioning an agent

Solution 1

One clear starting point

Problem: Most companies have no idea how much AI is running across their teams.
Solution: The module opens with one clear call to action: start discovery, and Scytale finds the AI for you.

Solution 2

Connect only what you use

Problem: Every AI provider holds a different piece of the picture.
Solution: Users connect only the sources they actually use, and can add the rest later, so getting started takes minutes.

Solution 3

Proof that the scan worked

Problem: After connecting, users need to trust the results.
Solution: The results screen shows how many agents were found, and exactly which source each one came from.

Solution 4

Know where to look first

Problem: With hundreds of agents, users don't know where to focus.
Solution: Summary cards surface high risk, need review and no owner, and the concentration view shows which models the company relies on.

Solution 5

An inventory built around risk

Problem: Customers have hundreds of AI assets.
Solution: The inventory puts risk first: risk, owner and review status in one table, with filters for source, risk, status and owner.

Solution 6

A full page for every agent

Problem: A drawer was too tight to explain what an agent decides, reaches and runs as.
Solution: Each agent opens as a full page: risk, ownership and activity in one view, with a clear approve or unsanction decision.

Solution 7

Risk comes from connections

Problem: Looking at one tool at a time hides how risk travels between systems.
Solution: The relationship map shows how every agent, model and data source connect, colored by risk level, with write and read only access marked.