This schedule (S4) applies when we provide AI services to you, such as Microsoft Copilot readiness and roll-out, AI products we provide and AI agents and automations we build for you. It forms part of your Contract with our Customer Agreement and adds to clause 16 of it.
Key points
- AI outputs can be wrong. People in your organisation must review them before relying on them.
- Copilot shows users what they already have access to, so your permissions and data labels matter.
- You decide what AI is used for. Significant decisions about people need meaningful human involvement or the safeguards the UK GDPR requires.
- Usage-based AI costs from Microsoft and other providers are charged as Third-Party Products and can vary.
- We do not use your data to train AI models, other than models built only for you.
1. The AI services
1.1 Our AI services include:
- readiness assessments, reviewing your data, permissions and governance before AI is switched on,
- Copilot deployment, configuring and rolling out Microsoft Copilot and training your users,
- agents and automations, designing and building AI agents and automated workflows for you, for example with Microsoft Copilot Studio or Azure AI services and
- AI products, AI tools we have built that we provide to you on a subscription basis.
1.2 Projects are carried out under a Statement of Work and our Professional Services Schedule, as well as this schedule. AI products are provided on the terms in your Order and this schedule. If this schedule conflicts with the Professional Services Schedule, this schedule applies.
1.3 In this schedule:
- Agent means an AI agent or automation we build or configure for you that can carry out tasks, such as answering questions, drafting content or taking actions in your systems,
- AI Model means an AI model provided by a third party, such as Microsoft or OpenAI, that an AI service uses and
- Output means content produced by an AI service, such as text, summaries, images, code or recommendations.
2. Your data and permissions
2.1 Microsoft Copilot and many Agents can find and use any content the user running them is allowed to access. If your permissions are too broad, AI can surface information to people who should not see it.
2.2 You are responsible for your permissions, sharing settings and data labels, unless the work to review or fix them is in the scope of your Statement of Work. We recommend a readiness assessment before any roll-out.
2.3 Clause 7 of our Data Processing Agreement applies to Sensitive Data. We will not knowingly connect an Agent to Sensitive Data that you have identified to us under clause 7.2 of the Data Processing Agreement without your written approval.
3. Outputs
3.1 Outputs are produced by AI Models and can be inaccurate, incomplete, out of date, biased or inappropriate, even when an AI service is working correctly. The same request can produce different Outputs at different times.
3.2 You are responsible for:
- reviewing Outputs before relying on them or sharing them,
- deciding which tasks AI is used for in your organisation and
- telling people when they are dealing with AI where the law or good practice requires it.
3.3 Outputs are not legal, financial, medical or other professional advice.
4. Decisions about people
4.1 You must not use an AI service to make decisions that have legal or similarly significant effects on individuals, such as decisions about employment, eligibility for services or safeguarding, unless:
- there is meaningful human involvement in each decision or
- you have met the requirements of Articles 22A to 22D of the UK GDPR, including the restrictions on special category data in Article 22B and the safeguards in Article 22C.
4.2 You are responsible for any data protection impact assessment needed for your use of an AI service. We will help as described in clause 4.7 of the Data Processing Agreement.
5. Agents that take actions
5.1 Some Agents can take actions, such as sending emails, updating records or starting workflows. Before an Agent goes live, we agree with you in writing the actions it may take, the permissions it uses and how it will be monitored.
5.2 We build each Agent so it can be paused or switched off by your administrators.
5.3 We are responsible for building Agents in line with the Statement of Work and with reasonable skill and care. You are responsible for the actions an Agent takes within the scope and permissions you approved, and for monitoring its use once accepted.
6. AI Models and usage costs
6.1 AI services rely on AI Models provided by third parties. Their terms apply as Third-Party Terms under clause 7 of the Customer Agreement. This includes the Microsoft Product Terms for Microsoft Copilot and Azure AI services.
6.2 AI Model providers change, update and withdraw their models. A change can affect how an AI service or Agent behaves. Work to adapt an Agent to such a change is outside the original scope, unless your Order or Statement of Work includes ongoing support.
6.3 Many AI services are charged by usage, for example per message, per credit or per amount of data processed. These usage costs are Third-Party Products and are charged to you at the rates in your Order, or passed through at cost if no rate is stated. Usage costs can vary with how much the service is used. Where the platform allows, we set spending alerts or limits that you agree.
7. Testing and acceptance
7.1 Because Outputs vary, acceptance of a Deliverable that uses an AI service, such as an Agent, is measured against a test set and pass rate agreed in the Statement of Work, not on every Output being correct. If none is agreed, clause 9.1 of the Customer Agreement applies, taking account of clause 3.1 of this schedule.
7.2 Clause 9 of the Customer Agreement applies to acceptance.
8. Intellectual property
8.1 Clause 13 of the Customer Agreement applies. Prompts, configurations and Agents built specifically for you are Deliverables. Our agent frameworks, reusable components and methods are Saueglobal Materials.
8.2 As between you and us, you own the Outputs produced from your use of an AI service. Copyright may not exist in some Outputs, and Outputs may sometimes resemble third-party material. Some AI Model providers, including Microsoft, offer their own protections against third-party copyright claims under their terms, subject to their conditions. Clause 18 of the Customer Agreement does not apply to Outputs.
8.3 We do not use Customer Data, prompts or Outputs to train AI models, other than models built only for you under an Order.
9. AI products
9.1 Where we provide an AI product on a subscription, we will maintain it and use reasonable efforts to keep it available during your subscription, subject to the usage limits in your Order. Clause 17.3 of the Customer Agreement applies.
9.2 We may improve or change an AI product, including the AI Models it uses. We will not make a change that materially reduces its core functions during your current term unless an AI Model provider or the law requires it. If we have to make such a change, we will tell you as soon as we can. You may then end the subscription by written notice within 30 days of being told, and we will refund the Charges you have paid for the rest of the term. Clause 11.7 of the Customer Agreement does not apply.
9.3 Where an AI product processes Customer Personal Data, we do so as your processor under the Data Processing Agreement. The AI Model providers it uses are our Sub-processors. We will add them to Annex 3 of the Data Processing Agreement, giving notice under clause 6.2 of it, before the AI product processes Customer Personal Data.
10. Use outside the UK
10.1 If you use an AI service in the European Union, you may have obligations as a deployer under the EU Artificial Intelligence Act. You are responsible for meeting them. We will give you information about the services we provide that you reasonably need to do so.