AI applications and agents
Assess whether Microsoft Foundry or Azure OpenAI fits your use case. Define evaluation examples, expected answer quality, access to data and where a person should review or approve results.
AURUM helps you design and build AI applications, business software and Microsoft 365 integrations on Azure. We start with the workflow and your data, then assess where Azure adds value and what your team will need to run the solution.
Microsoft Solutions Partner for Digital & App Innovation (Azure) · Based in Nuremberg since 2014 · About AURUM
We can start with an architecture review, validate a workflow with a pilot, or implement and hand over an agreed solution.
Assess whether Microsoft Foundry or Azure OpenAI fits your use case. Define evaluation examples, expected answer quality, access to data and where a person should review or approve results.
Use Azure Document Intelligence where it suits the document types and fields you need. Test representative files, define when someone needs to check an extraction, and plan how to pass approved data downstream.
Connect SharePoint, Teams or a Microsoft 365 workflow with an ERP, CRM or internal application where supported interfaces allow. Plan permissions, data mapping, retries and monitoring.
Design a custom web application, API or background process where standard tools leave a gap. Depending on the workload, Azure Functions can run code and Azure Logic Apps can orchestrate integrations. Explore custom software development.
Map where data is stored and processed. Review Microsoft Entra identities, permissions, logs and retention, and check which deployment options meet your requirements.
Estimate running costs from expected use and identify what changes with volume. Agree who owns deployment, alerts, access reviews and support before going live.
Choose the platform after checking the data, users, workflow and support requirements. A project can use more than one of these approaches.
Use built-in features when they cover the task. If you need an internal assistant, assess the Microsoft 365 Copilot or Copilot Studio options that fit your licences and governance requirements.
Power Platform can suit forms, approvals and integrations where the connectors and access model fit. Check connector licences, exceptions and who will maintain the flow.
Consider Azure when you need your own application or API, specialised document processing, or more control over model deployment and integration behaviour. Plan testing and operating responsibilities alongside development.
Deployment types in Microsoft Foundry differ in where model inference can run. Check the selected model, service and deployment type against your requirements. Microsoft’s Foundry deployment guidance.
A project recommendation should make the design decisions and assumptions visible to the people who will approve and operate it.
Identify the files, records and systems the application needs. Record who can access each source and what data leaves each system.
Pick representative test cases and agree what counts as a usable result. Define what happens when a source is unavailable, an extraction is incomplete or someone needs to review a result.
Estimate the main cost drivers and agree who handles deployment, alerts and updates. For a related integration example, read our guide to connecting Copilot Studio to a CRM.
By the end of the pilot, you should have enough evidence to decide whether to proceed and what it will take.
Discuss your requirementsIdentify who will use the application, which systems it touches and what a useful outcome looks like.
Compare Microsoft 365, low-code and Azure approaches. Map data flows and permissions, estimate costs and check deployment constraints.
Test the workflow, integration or AI output against agreed examples. Review failure cases and decide what to refine before rollout.
Agree how to deploy, monitor and maintain the solution. Document responsibilities, access and any ongoing support needs.
Answers to common questions about platform choice, costs, and where data is processed.
It depends on the task. Microsoft 365 features or Power Platform can cover many internal workflows. Azure can help when you need an application, integration or AI feature that requires capabilities beyond the current setup.
Yes, where supported APIs or connectors are available and approved. We assess authentication, permissions, what data should move and how to handle failed or duplicate requests.
We review implementation effort and recurring costs separately. Depending on the design, recurring costs can include model use, document pages, workflow runs, monitoring and support. A pilot helps validate the assumptions.
Some services and deployment types can support an EU processing requirement. We check what is available for the selected model and service, and review inference, storage, logging and connected systems individually.
Yes. We can scope a review or pilot around one use case, with agreed test cases and a decision at the end about whether to proceed.
We agree ownership during planning. We can document the deployment and train your team, and ongoing support can be included in the agreed scope.
Tell us what you want to build, which systems it needs to connect to and any constraints on data handling. We’ll help you identify a starting point in a free initial consultation.

Mustafa Yaylak
Managing Director, AURUM GmbH