Skip to main content

AI FAQs

V 1.0 This document contains the most Frequently Asked Questions (FAQs) for the AI features available to users within the StructureFlow Application

Updated over 2 weeks ago

📚 General

What AI features does StructureFlow offer?

StructureFlow offers two AI-powered features: an AI assistant chatbot, and an import tool that automatically converts spreadsheets, documents, and images into diagrams. All features are optional and controlled by your organisation's administrator.

Do we have to use the AI features?

No. AI features are disabled by default and will not be visible to your users unless your System Administrator actively enables them. You remain in full control of whether AI is used within your organisation.

Who controls the AI settings for our organisation?

Your designated System Administrator manages AI settings through the admin settings panel. They can choose to disable AI entirely or enable it under either the "Your Region" or "Global" mode.

Can users see AI features when AI is disabled?

No. Users are not able to see the disabled AI features within the StructureFlow Application.

Does StructureFlow provide additional AI documentation or AI training to support user adoption?

Yes. Should you or your team require training on AI usage, please reach out to the Support team at [email protected]. There is also a variety of supporting documentation stored within the Application’s Help Centre which users can navigate to using the link located in the top-right corner of the canvas:


🤖 AI Providers and Modes

Which AI providers does StructureFlow use?

  • Microsoft Azure OpenAI — GPT models (e.g. GPT-5)

  • Anthropic (via Microsoft Azure) — Claude models (e.g. Claude Sonnet 4.6)

  • Google Vertex AI — Gemini models (e.g. Gemini 3)

Which AI mode isrequiredfor each provider?

Provider

Required mode

Azure OpenAI (GPT)

Your Region or Global

Anthropic (Claude)

Global only

Google Vertex AI (Gemini)

Global only


⚡️Performance

How do you ensure output accuracy and reduce AI bias?

To ensure accuracy and mitigate bias, we conduct manual and automated regression testing in our staging environments after updating model versions. This process helps prevent regressions in accuracy and bias.

Are some AI providers better than others for certain processing?

We will always automatically use the best model for each feature in the application. Users can override the default model with a model of their choice if they choose to.


🔐Data, Privacy and Security

Is our data used to train AI models?

No. Your data is never used to train or improve any AI model. AI inputs are processed solely to generate the relevant output and are not retained beyond that purpose.

Is our data stored after an AI request is made?

StructureFlow stores AI input data only for as long as it takes to produce a response. Once an output is generated, it is saved as part of your project data within StructureFlow in the usual way. Input data is cached with a maximum TTL (Time to Live) of 24 hours by model providers.

Where is AI data processed?

This depends on the AI mode your administrator has selected:

  • Your Region

    • Data is processed in the Azure data centre closest to your application region (e.g. for UK customers, this is the UK or EU).

  • Global

    • Data may be processed across global infrastructure, including US, EU, Asia, and Australia, depending on the AI provider used.

Do Anthropic and Google models process data within the Azure environment?

No. Although Anthropic's Claude models are accessed via Microsoft's Azure platform, data is processed on Anthropic's own infrastructure. Similarly, Google Vertex AI processes data entirely within Google's infrastructure. Both providers are bound by formal data processing agreements.

What happens to our data when it leaves our region?

When data is transferred outside the European Economic Area (EEA), those transfers are protected by EU Standard Contractual Clauses (SCCs). Each AI provider – Microsoft, Anthropic, and Google – has a Data Processing Addendum (DPA) in place that governs how your data is handled and sets out the legal basis for any international transfers.

Are Anthropic and Google Cloud sub-processors under GDPR?

Yes. StructureFlow recognises Anthropic and Google Cloud as sub-processors under GDPR. Both entities are not expected to process high volumes of personal data, undertake systematic profiling/automated decision-making, or handle special category personal data.

Can we restrict processing to a specific country?

No. Within the "Your Region" mode, processing is directed to the closest available data centre for all customers on the same application instance. It is not possible to select a single country exclusively – for example, UK-only processing is not available on the UK instance as it also uses EU data centres for AI processing.

Are the AI features within scope of the penetration tests?

Yes.


⚖️ Compliance

Is StructureFlow required to comply with the EU AI Act?

No. After an internal evaluation, StructureFlow has identified that the AI system used within the Application is classified as minimal risk under the EU AI Act as it is used solely to support diagramming tasks and does not affect individuals' rights, health, or safety. Microsoft, Anthropic, and Google are identified as the AI system Providers and hold primary responsibility for model-level compliance. StructureFlow operates as a Deployer under the Act.

Is StructureFlow GDPR compliant in its use of AI?

Yes. International data transfers are governed by Standard Contractual Clauses and each provider's Data Processing Addendum, which serve as the lawful transfer mechanism under Article 46 of the GDPR.

What voluntary commitments has StructureFlow made under the EU AI Act?

Beyond mandatory transparency obligations, StructureFlow has voluntarily committed to:

  • Ongoing monitoring of AI system performance and availability

  • Ensuring AI usage remains within recommended guidelines from each provider

  • Encouraging users to review all AI outputs, as responses are not guaranteed to be fully accurate


🌐 Self-hosting Options

Can we still use AI if we host StructureFlow on-prem?

Clients who are hosting StructureFlow on their own infrastructure will need to provision several models depending on which features they want enabled. The section above on Processing Locations will not apply to self-hosting customers.


Please reach out to the StructureFlow support team at [email protected] for any further enquiries

Copyright © 2025 by StructureFlow Ltd.

Did this answer your question?