AI Transparency and Governance

01 Our approach to responsible AI

We build AI that does real work for businesses, and we think the responsible way to do that is to keep a person in charge, be clear about what the technology is, and design for the rules that govern AI in Europe rather than around them. We aim to align with the EU AI Act and with applicable data-protection law, and we treat that alignment as part of building a product worth trusting.

02 What Digital Employees are

A Digital Employee is an AI system that performs a defined role and the tasks within it, configured by the customer for its business. A Digital Employee is not an autonomous decision-maker that operates without supervision: it is designed to work with a person on the loop, who configures it, reviews its work and stays accountable for the decisions taken. The customer decides what a Digital Employee does and what data it works with.

03 Knowing when you interact with AI

People should know when they are dealing with AI. Article 50 of the EU AI Act sets two transparency duties that become applicable on 2 August 2026: telling a person when they are interacting with an AI system (Article 50(1)), and marking AI-generated content in a machine-readable form (Article 50(2)). We are committed to meeting both, and we treat them as product features to build into the Digital Employee channels (web chat, voice, email and Teams) ahead of that date, working with our customers so that, where a Digital Employee interacts with a person or generates content the Act requires to be marked, the appropriate disclosure and marking are in place.

04 Human oversight

Human oversight is built into how a Digital Employee works, and it is also a responsibility we ask our customers to keep. Outputs can be reviewed, corrected and overridden, and significant actions are designed to keep a person in control. Customers are responsible for maintaining appropriate oversight of the Digital Employees they configure, and for reviewing outputs before relying on them, as set out in our Terms of Service.

05 How the AI Act applies to our systems

The EU AI Act takes a risk-based approach, with different obligations for different levels of risk, applied in stages. Its prohibitions and AI-literacy duties have applied since 2 February 2025, and its rules for general-purpose AI models since 2 August 2025. The remaining obligations, including the Article 50 transparency duties described above, become applicable on 2 August 2026. An amending regulation provisionally agreed at EU level in May 2026 is expected to postpone the high-risk regime for Annex III use cases to 2 December 2027; until that amendment is formally adopted and published, the current dates remain the law, and we plan against them.

Our current services, focused on finance and data workflows, are designed to fall within the Act's transparency expectations rather than the high-risk category. Whether any particular use is high-risk, however, can depend on the context in which a customer deploys a Digital Employee. Where a customer uses a Digital Employee in a context that the Act treats as high-risk, such as certain uses in employment or recruitment, additional obligations may apply, and we work with customers to support compliance in those settings. We keep our assessment under review as our catalogue, the Act's guidance and the pending amendment develop.

06 Practices we do not enable

We do not build Digital Employees for any purpose prohibited under the EU AI Act, such as manipulative or exploitative techniques, social scoring, or prohibited uses of biometric data. Our Acceptable Use Policy prohibits customers from using the Service for those purposes, and we will act on breaches. The Service is not designed to make decisions that produce legal or similarly significant effects on individuals without appropriate safeguards and meaningful human review.

07 Data, training and accuracy

We are careful with the data that flows through a Digital Employee. As set out in our Data Processing Agreement, we do not use customer data to train foundation models, and we do not pool or share it across customers. The customer's data, and the contextual memory a Digital Employee learns within the customer's instance, belong to the customer. We use reputable foundation-model providers and work to reduce errors and bias, but AI outputs can still be inaccurate or incomplete and are not a substitute for professional advice. This is why human review matters, as described in Section 4.

08 AI literacy

Article 4 of the EU AI Act asks providers and deployers of AI to make sure the people working with their systems have a sufficient level of AI literacy. We take measures to maintain AI literacy among our own people, and we aim to give customers the documentation and context they need to understand what a Digital Employee does, how to oversee it, and where its limits lie.

09 Accountability and contact

Responsibility for our approach to responsible AI sits with our leadership, and we keep this statement and our practices under review as the EU AI Act, its guidance and the relevant codes of practice continue to develop. If you have a question or concern about how we build or operate our AI, please get in touch: