AI in the workplace.
Introduced clearly.
Suitable tools, understandable rules and sound technical integration: DAXS supports the introduction of AI at your workplaces — aligned with tasks, data and existing systems.
Which work should become easier?
Prepare text, find information, structure meetings or review an initial code draft: AI can support a range of workplace tasks. A sound introduction clarifies the benefits, permitted data and responsible people before additional functions access company information.
DAXS helps select suitable services, configure the technology, manage identities and permissions, and integrate the tools into existing workflows. This includes clear usage rules, traceable approvals and an introduction that employees can put into practice.
Approving a product does not approve every feature.
An ordinary chat, a file upload, a search of company data and an agent that takes actions follow different data paths. Approval should cover the specific service, plan, account type and features actually used.
Product marks represent example tools. Available functions and controls depend on the offering selected and its configuration. Other suitable providers and local models remain equally valid options.
Your existing work environment is an important part of the decision.
We compare capabilities, integration, data processing, costs and the operations required. Existing licences and skills also matter. A list of familiar brands alone does not answer these questions.
| Example tool | Possible approach | What to check |
|---|---|---|
| OpenAI / ChatGPT | Suitable business features, technical assistance or custom applications through an API. | Distinguish personal accounts, business offerings and the API; clarify retention, connected services, identities and approved features. |
| Microsoft Copilot | AI within your existing Microsoft 365 work environment and suitable business applications. | Check the specific Copilot product and licence, access to SharePoint, OneDrive and Teams, agents, sharing and retention. |
| Google Gemini | Assistance within a suitable Google Workspace offering or another approved service. | Distinguish the edition, specific feature, data sources, controls and contractual basis; a Workspace rule does not automatically apply to every Gemini offering. |
| Anthropic Claude | Suitable assistance, document processing or API integration according to the task. | Check the business or personal offering, controls, retention, connectors and permitted data for the specific use case. |
| Mistral AI, Meta Llama and other models | Local, hosted or individually integrated processing, depending on the model and offering. | Clarify the model licence, deployment model, quality requirements, operational effort and actual data paths. |
A statement about training usage does not automatically answer questions about storage, logs, region or access by connected applications. We assess these matters separately using current contractual and product information. For your own infrastructure, see local LLM planning.
Faster search can also expose existing oversharing.
A new AI service does not automatically resolve an overly broad sharing setting on a directory or SharePoint site. Introduction therefore starts with data owners, group memberships, sharing links and appropriate retention rules.
In a Microsoft 365 environment, we consider SharePoint, OneDrive, Teams and the Copilot features actually used, among other things. Existing user rights and information protection influence what the service can find and use. Permissions are corrected before broad deployment and reviewed regularly afterwards.
- Identities: managed accounts, appropriate sign-in, roles and groups.
- Lifecycle: handle joining, changes of responsibility and leaving in a traceable way.
- Connectors and agents: approve required sources and actions individually, rather than enabling every integration across the board.
- Data spaces: identify especially protected information and limit technical access appropriately.
Employees need practical examples for their daily work.
Usage rules should explain which tasks and information are permitted in each feature. A practical classification can begin with data classes and then be refined for your organisation.
Public information
For example, prepare already published product text for a draft. Copyright, technical accuracy and approval for publication still need consideration.
Internal working information
Use only within services and features approved for that purpose. The intended audience, storage and possible access by connected applications form part of the decision.
Confidential or especially protected content
Customer data, contract contents, personnel information and credentials require their own assessment. Necessary approvals and technical paths are defined before processing.
An internal policy can also govern permitted accounts, uploads, result checking, sharing, allowed automation and a reporting path for errors or unsuitable answers. Examples from actual tasks are more helpful than a long list of abstract prohibitions. An AI answer does not thereby receive automatic professional or legal approval.
Apply approvals traceably on managed devices.
Depending on the existing platform, measures can include managed accounts, device and browser configuration, app approvals, connector rules and suitable network and data protection controls. Which measures are available and appropriate depends on device type, product, licence and the operating concept.
Technical restrictions support clear guidance. They do not automatically cover every personal device, personal sign-in or possible data transfer. Permitted working methods are therefore explained and implementation is coordinated with the responsible parties.
Operational data and monitoring individuals are separate matters.
For troubleshooting and security, we define data for specific purposes, authorised access and retention periods. This does not imply blanket collection of all employee prompts or general permission for surveillance. Management, those responsible for data protection, legal advisers and, where appropriate, the works council are involved early according to the measure.
Start with a suitable pilot and learn from daily work.
- Identify tasks: specific workflows, desired benefits and suitable outputs.
- Clarify data and responsibilities: sources, accounts, approvals and necessary specialist coordination.
- Configure the technology: implement selected features, identities, access and operating rules.
- Provide practical training: explain typical questions, limits, result checking and reporting paths through real tasks.
- Review and adjust: monitor quality, acceptance, permissions, new features and ongoing costs.
Data protection, employment law and the applicable AI Act requirements are assessed for the specific purpose, role and application. DAXS handles the agreed technical implementation and supports a traceable introduction. Professional and legal decisions remain with the responsible parties.
Assess the intended use of AI, review existing identities and data sources, prepare a limited introduction plan and an understandable proposal for technical measures. Existing effective workflows remain the starting point.
We clarify this in our conversation.
Which AI tool suits our business?
That depends on the tasks, data, existing platforms and required operations. We compare specific products, editions and features. A chat service, API, coding tool and integrated Office feature have different requirements.
Is an existing Microsoft 365 environment automatically ready for Copilot?
Not necessarily. Existing access rights, sharing and document collections need to fit the intended use. Permissions that are already too broad can make information more widely discoverable. We therefore review the data foundation and actual permissions.
Which data may employees enter?
This is defined according to purpose, protection needs, product and approved configuration. Practical data classes and examples support everyday decisions. Especially sensitive or contractually restricted information requires a separate assessment.
Must all workplace tasks use a local AI server?
No. Local, external and hybrid solutions are equally valid options. Some tasks suit an existing business service, others a local model or conventional automation. Benefits, data paths and operations determine the choice.
How do we introduce AI without overwhelming daily work?
With a few specific tasks, a suitable pilot group and understandable rules. Short training sessions, verifiable examples and a clear point of contact support use. After evaluation, features and participants are expanded gradually.
A company account needs clear responsibilities.
Personal accounts and a managed company workspace have different ownership and administration models. Before introduction, we clarify joining and leaving, permitted data classes, connectors, approvals, retention and export. No use for model training does not automatically mean no storage. Actual contractual and product options are checked. Internal rules and training explain permitted tasks, source checking and escalation; they do not replace technical permissions. Organisational and legal requirements are agreed for the specific use case.
Let's discuss your use case.
What should become easier, which data is involved and which systems are already in place? We use this to develop a practical starting point and a manageable next step.