Skip to content

First Steps ​

Set up the foundations for Scout, risk quantification, and audit readiness.

Who is this for?

Organization Admins who set up a new workspace, and anyone who registers the first AI System. If your organization is configured, go to Quickstart.

Why this setup is important ​

Scout, the AI assistant, uses your organization and AI System descriptions as context. If these are empty, the answers of Scout are general. The risk taxonomy and limits that you set here apply across the platform. They set which risks teams can select and how budgets are divided.

This setup takes 20–30 minutes. It prevents hours of corrections after the setup.

Step 1: Configure your organization ​

Organization → Settings

Set the fields of this step on the General tab. Then click Save.

Organization description ​

Modulos puts your organization description into the system prompt of Scout. It is the first information that Scout has about you. If it is empty, Scout does not know what your organization does.

The platform shows this text:

Write a brief description of your organization to help our AI agents understand what context you operate in. The following information would be helpful:

  • What business your organization is in
  • What products or services you offer
  • What countries or regions you operate in
  • Key figures such as how many employees you have, what your revenue is
  • Anything else that would help the agents assist you

Two to three paragraphs are sufficient. Give details. "Series B fintech company offering credit scoring APIs to European banks, 120 employees, headquartered in Zurich" helps Scout much more than "We are a technology company."

Currency ​

The currency sets the unit for all monetary risk quantification, budgets, and portfolio rollups. Select it in this step. If you change it after this step, Modulos does not change historical values to the new currency.

Default language ​

The default language is the first language of each new user and of each new AI System. Each user can change their own UI language. The default language does not translate your content. Control reports, Evidence, and policies stay as you wrote them.

INFO

Learn more: Organization Settings

Step 2: Invite users and give roles ​

Organization → User Management

Invite the persons who will work in the AI Systems before you add the AI Systems. Modulos uses role-based access at two levels. Organization roles control who can configure shared settings. AI System roles control who can do the work, who can examine it, and who can audit it in each AI System.

Organization roles ​

Every user is an Organization Member. Most users have no other organization role. To manage users and settings, speak to your administrator about Organization Admin. Organization Risk Manager owns risk limits. Organization Policy Manager owns Policy Center. Organization Registration Manager decides on new AI Systems.

In each AI System, the role is Owner, Editor, or Viewer. See User Management for what each role can do.

How to invite ​

  1. Go to Organization → User Management and click New User.
  2. Type the name, the last name, and the email address of the user.
  3. Select the organization roles of the user. With no role selected, the user is an Organization Member.
  4. Click Invite User.
  5. The user receives an email with a signup link. The link expires after 7 days.

Invite early, give AI System roles next

You can invite the full team at this step. Give AI System roles when you add each AI System in the next steps.

INFO

Learn more: User Management

Step 3: Make Connectors and Sources available ​

Organization → Source & Connectors

By default, all Connectors and Sources are unavailable. Until you make them available, the platform cannot see your systems. Scout cannot examine your repositories or documents, and Runtime Inspection has nothing to test against. Review this page when you first configure the organization.

The page has two tabs:

  • Sources — service accounts, attached to an AI System, that Metric Tests, Agent Tests, Scout, and the Risk Agent read. See Supported Source types.
  • Connectors — OAuth connections that ground AI-assisted work in your systems: Google Drive, GitHub, Atlassian, Bitbucket.

On each tab, turn on the types that your organization uses. Then click Save Changes. Accept the change in the Confirm Changes dialog. Connectors and Sources that you do not make available are not available in AI Systems, new or existing. A Connector that is unavailable here is unavailable for all teams.

INFO

Learn more: Integrations

Step 4: Register your first AI System ​

AI Registry → New AI System

An AI System is the unit of work in Modulos. Each entry is one deployed AI system or use case, in its deployment context: one model, one use case, one set of users. System-specific compliance, risk quantification, and testing occur here.

Keep AI Systems narrow. One deployed system or use case for each AI System keeps Evidence defensible and risk quantification comparable. If you operate three systems, register three AI Systems. If you use the same tool for two purposes, two entries can be necessary, because the risk classification changes with the use of the system.

Intended purpose and description ​

New AI System asks for a name and the owners. It saves the entry as a Draft and opens the wizard. The wizard asks for an Intended Purpose (required) and a Description.

The intended purpose drives the legal classification of the system. Name the task that the system performs and the business procedure it is part of. Name the persons on whom its outputs have an effect (employees, applicants, customers), and the domain. "Answers employee questions about HR policies" and "screens job applications" are two HR chatbots in very different risk categories. Write how you actually use the system.

Modulos puts the description into the system prompt of Scout (up to 5,000 characters). It is the most important input that makes the answers of Scout related to your system. Write what the system is and how it operates. Include the type of technology, the data that goes in, the output, and the users. Write about your deployment, not the marketing text of the vendor.

Write it like a scope statement

Include the system architecture, for example "RAG pipeline using GPT-4o with a Qdrant vector database". Also include the data Sources, the deployment context, and the end users. Scout uses this as context for each answer. Auditors will also use it as the scope statement.

Other fields to set first ​

  • Current State (required): Proposed, Pilot, or Already in use. Write what exists today, not what was approved.
  • The user groups that use the system (required, on step 2). Select all groups that apply.
  • Does the EU AI Act apply to this AI System? (step 2): Yes, No, or Not sure. The classification fields show only on Yes.
  • AI Lifecycle Stage: tells Scout where your system is in its life cycle, for example development or production.
  • AI System Annual Economic Value: the value that the AI System gives in one year. Modulos compares it with the risk value in the Value and Risk view, and can use it to divide risk limits across AI Systems. A rough estimate is sufficient. You can refine it after this step.
  • Geographic Scope: where the system is used (EU, UK, US, APAC, LATAM, MEA, Global). Scout uses it as context.
  • Roles: you are the first Owner. Add a second Owner in the Create New AI System dialog. Add Editors for the working team and Viewers for oversight in AI System → Settings → User Access. Viewers read the AI System and add comments, but cannot change items.

EU AI Act and frameworks ​

If the EU AI Act applies, answer Yes to the EU AI Act question on step 2 and record the classification (Product Category, Role, Use Case). The wizard then pre-selects the EU AI Act framework on step 3. Scout is available from the page header during registration and reads your entries.

Add the other frameworks that apply. OWASP Top 10 for LLM Applications and NIST AI RMF are usual starting points. You can start with one framework and add more in AI System → Settings → Frameworks. Before an audit, an Organization Admin can freeze framework updates there to keep the frameworks stable.

Approval of the registration ​

On the last step, Create AI System sends the AI System for approval, and its status changes to Pending. An Organization Admin or an Organization Registration Manager approves or rejects it on its Overview. Only AI Systems in Pending, Approved, or Rejected count against the AI Systems limit of your plan.

INFO

Learn more: Register an AI System

Step 5: Attach organization-scope frameworks to the organization AI System ​

AI Registry → organization AI System → Settings → Frameworks

Modulos separates two layers of governance. You register AI Systems for individual deployed systems. One organization AI System covers everything that applies across them.

What the organization AI System is ​

The organization AI System covers organization-wide governance programs and shared Controls. Examples are company-wide AI policies, management system documentation (ISO 42001, ISO 27001), and cross-cutting governance processes that apply to all your AI systems. It appears in the AI Registry with Type Organization. An organization needs only one.

Organization-scope frameworks (OFF-x) attach to it. The AI Registry does not make new organization AI Systems. If your organization has none, contact Modulos before you roll out organization-scope frameworks.

What to do ​

  1. Open the organization AI System from the AI Registry (Type Organization).
  2. In AI System → Settings → General Settings, write its description. You can keep it empty when the organization description in Organization → Settings covers it. If it has a defined scope (a certification, a sub-unit of your organization), write that scope.
  3. In AI System → Settings → User Access, give roles. Give a minimum of one user the Owner role. Add Editors for the persons who do the work.
  4. In AI System → Settings → Frameworks, click Add Framework. Select organization-level standards such as ISO 42001, ISO 27001, or NIST AI RMF.

How the two layers work together ​

Organization AI SystemAI System
ScopeOrganization-wide policies, management systemsOne deployed AI system or use case
Typical frameworksISO 42001, ISO 27001, NIST AI RMFEU AI Act, OWASP, NIST AI RMF
ControlsShared policies, training, governance proceduresTechnical Controls and procedures of one system
RiskGovernance risks that apply to all AI SystemsTechnical and legal risks of one system, and the risks of its daily use
How many?OneOne for each deployed system or use case

AI Systems that use a Control in the organization AI System can refer to it. An example is "AI Ethics Policy exists and is reviewed annually". This reduces duplicate work.

INFO

Learn more: AI Systems

Step 6: Examine and adapt your risk taxonomy ​

Organization → Risk Management → Category Taxonomy / Risk Taxonomy / Threat Vector Taxonomy

Only an Organization Risk Manager can change the risk taxonomy. An Organization Admin without this role can only read it.

Modulos includes a default taxonomy with five categories:

CategoryCovers
Technical RisksModel quality, robustness, cybersecurity
Operational RisksHuman factors, monitoring, operations
Legal & Compliance RisksPrivacy, liability, regulatory exposure
Ethical & Reputational RisksFairness, offensive or dangerous outputs, trust
Governance RisksStructure of the governance work, documentation, oversight

The default taxonomy is a starting point. A taxonomy that matches your actual systems makes risk quantification meaningful and comparable across AI Systems.

What to do ​

  1. Review the default categories and risks. Examine if each one is related to your domain.
  2. Add domain-specific risks. For LLM applications, examples are prompt injection in customer-facing contexts, hallucination in regulated advice, and training data contamination.
  3. Curate threat vector associations. When a team adds a risk, it can select only the threat vectors that the taxonomy links to that risk. Unlinked threat vectors cannot be selected.

Start with what you know

A perfect taxonomy on day one is not necessary. Add the risks that your team discusses. Change the taxonomy as incidents and audits show more.

INFO

Learn more: Organization Taxonomy

Step 7: Set risk limits for quantification ​

Organization → Risk Management → Risk Limits and Organization → Risk Management → AI System Risk Limits

Only an Organization Risk Manager can change the risk limits. An Organization Admin without this role can only read them.

The limits have three levels:

  1. Total Organization Risk Appetite — the monetary maximum of the risk that your organization accepts.
  2. Category allocations — the percentage of the appetite for each risk category. When the percentages do not add up to 100%, Modulos shows a warning.
  3. AI System risk limits — the parts of the total appetite for each AI System. When they do not add up to the total, Modulos shows a warning.

These warnings do not stop risk quantification. Only Pending and Approved AI Systems get a part of the risk appetite. A Draft or a Rejected AI System has no risk limit.

What to do ​

  1. On the Risk Limits tab, set the Total Organization Risk Appetite in your chosen currency.
  2. Give percentages across categories. An example is Technical 30%, Operational 25%, Legal 20%, Ethical 15%, Governance 10%.
  3. Click Save Changes.
  4. On the AI System Risk Limits tab, select a method: Equal Distribution (default) or Economic Value-Based Allocation.
  5. Click Save Changes.
  6. Review the resulting AI System risk limits. Override them manually if needed.

Start with rough estimates

The goal is a first set of limits. Change the numbers when you quantify threats and learn your exposure.

INFO

Learn more: Risk Operating Model

Result of the seven steps ​

After these seven steps:

  • Your team has the right roles and permissions.
  • Connectors and Sources that your organization uses are available. Scout and Runtime Inspection can then connect to your systems.
  • Scout has the context to give answers that are related to your organization and AI Systems, not general answers.
  • The two layers are set up: system-specific compliance in your registered AI Systems, and shared governance in the organization AI System.
  • Risk limits are set. Your team can quantify threats in monetary terms and compare them with the limits.
  • Your risk taxonomy reflects your actual domain, not a generic template.

Next steps ​