IT consulting for data, AI & SAP
Data and AI that work in production.
With a focus on Databricks and SAP: we connect your SAP world with a modern data platform and take AI agents, Copilot and automation into operations on top — on Azure, as code, with governance.
Certified in
- Databricks
- SAP
- Microsoft Azure
The situation
The pilot works. Now the platform has to.
Many companies already have their first AI prototypes, a handful of pipelines and some cloud resources. What’s missing is the foundation for running them in production. That’s exactly where we come in.
Sprawl instead of a platform
Environments, permissions and costs without shared rules.
Our answerOne shared platform with clear roles, governance and cost control — as code, e.g. with Databricks Unity Catalog.
SAP data as a black box
Master and transactional data arrive late, incomplete or duplicated.
Our answerClean master data and reliable SAP integration all the way into your data platform.
GenAI with no path to production
Prototypes, but no serving, no monitoring, no sign-off.
Our answerAn agent platform with tracing, evaluation, monitoring and approvals — all the way into operations.
Services
Thirteen areas, one team — from infrastructure to use case.
Live example
How an AI agent works in your processes.
Pick a case and start the flow. The agent reads, checks and prepares — people make the decisions.
- Invoice arrivesEmailPDF invoice from Sample Ltd. received
- Read the documentDocument IntelligenceInvoice no., date, 12 line items and total extracted
- Match the purchase orderSAP S/4HANAPurchase order found — line 7: quantity 120 instead of 100
- Assess the deviationAI agentDeviation above tolerance → approval required
- ApprovalHuman in the loopProcurement confirms the over-delivery in Teams
- PostSAP S/4HANAInvoice posted, document created, case logged
Agent / systemHuman in the loop
- Order arrivesScan / emailScanned customer order (PDF) with a handwritten note received
- Read the documentDocument IntelligenceCustomer, 8 line items, quantities and delivery date recognized — incl. handwriting
- Map the dataAI agentCustomer item numbers mapped to SAP material numbers, prices checked
- Resolve the exceptionHuman in the loopInside sales confirms one unclear line with a click
- Hand over to SAPSAP BTP · Integration SuiteOrder data validated in an integration flow and passed to S/4HANA
- Create the orderSAP S/4HANASales order created, availability checked
- Confirm & send backE-signatureOrder confirmation generated, digitally signed and sent to the customer
Agent / systemHuman in the loop
- Recording availableTeams / videoProject meeting (48 min, German and English) recorded
- TranscribeAzure AI SpeechTranscript with speakers and timestamps created
- SummarizeAI agentMinutes with decisions, open items and 7 tasks generated
- Add contextKnowledge baseLinks to project, requirements and earlier decisions added
- ReviewHuman in the loopProject lead corrects one task and approves the minutes
- Create tasksPlanner / JiraTasks with owners and due dates created, minutes distributed
Agent / systemHuman in the loop
- Load dataSAP S/4HANANightly extract: orders and customer master into Databricks (bronze)
- TransformDatabricksSilver/gold tables in Unity Catalog refreshed
- Check rulesData qualityRule violated: 312 orders without a valid customer number
- Analyze the causeDatabricks agentCause: new sales organization missing from mapping — fix proposed
- Approve the fixHuman in the loopData owner reviews and approves the mapping
- Refresh the reportPower BIGold tables recalculated, dashboard current and correct
Agent / systemHuman in the loop
- Ask a questionMicrosoft Teams“What is the open order backlog for customer Miller?”
- Route to the agentCopilotCopilot detects a data question and calls the A2H data agent (Copilot Studio)
- Query the dataDatabricks agentAgent queries gold tables in Unity Catalog — only with the user’s permissions
- Compute the answerDatabricksOpen backlog: 14 orders, €286,400, 3 of them overdue
- Answer with sourceMicrosoft TeamsAnswer as a card in Teams, with a table and a link to the Power BI report
- Next stepHuman in the loopSales creates a task for the overdue orders with one click
Agent / systemHuman in the loop
- Load historySAP S/4HANASales of the last 36 months per material loaded into Databricks
- Create featuresDatabricksSeasonality, holidays and promotions computed as features
- Train the modelMLflowForecast model trained, evaluated against last year and versioned
- Explain anomaliesDatabricks agent3 materials with strong deviation — reasoning generated per material
- Approve the forecastHuman in the loopPlanners review the outliers and approve the forecast
- Update planningSAP BTP · Integration SuiteApproved values passed to SAP demand planning
Agent / systemHuman in the loop
- Request arrivesSAP BTP · Fiori appNew supplier “Example Inc.” requested in a Fiori app on SAP BTP
- Search for duplicatesAI agentSimilar record found: “Example Incorporated” (93%)
- Validate dataValidation serviceAddress normalized, VAT ID valid
- DecisionHuman in the loopMaster data team confirms: new site, not a duplicate
- Request creationSAP MDGChange request with all mandatory fields created
- DistributeSAP S/4HANASupplier created and distributed to downstream systems
Agent / systemHuman in the loop
- Enter the needSAP BTP · Fiori app“I need 20 laptops for the new project team by the end of May.”
- Understand the requestAI agentMaterial, quantity 20, requested date and cost center recognized
- Check catalog & budgetSAP S/4HANAContract item found, cost center budget sufficient
- Start approvalSAP Build Process AutomationApproval workflow sent to the cost center owner
- ApproveHuman in the loopCost center owner approves in the Fiori inbox
- Create requisitionSAP S/4HANAPurchase requisition created and handed to procurement
Agent / systemHuman in the loop
Why A2H-Consulting
Why companies work with us.
Data, AI and SAP in one team
Master data, integration, Databricks and AI from one team — no hand-off losses.
Enterprise experience, delivered lean
Experience from projects at companies with 10,000+ employees — delivered with short paths.
All the way to production — and beyond
Not a demo but an accepted solution as code, with documentation and handover.
A direct network of experienced specialists
Vetted senior specialists from our network when needed — with one point of contact.
Projects
Experience from regulated, data-intensive industries.
- Pharma & Medical Technology · 5,000–10,000 employees
Databricks platform with GenAI as a shared service
One shared platform on which teams deliver new GenAI use cases from existing building blocks.
- Transportation & Logistics · 1,000–5,000 employees
Scalable data pipelines with Databricks and Data Factory
Automated, monitored data pipelines instead of manual loads.
- Automotive · 10,000+ employees
Central business partner data in SAP
One system of record for business partners with automated quality checks.
- Lakehouse platformsbuilt from architecture to acceptance
- Data productsdesigned and taken into production
- 10+ yearsdata, AI and SAP projects
- CertifiedDatabricks, Microsoft and SAP
Process
End to end — from project start to operations.
- Phase 1
Initiate & understand
- 01Project start
- 02Analysis & requirements
OutcomeGoals, scope and requirements are clear and agreed.
- Phase 2
Design
- 03Functional design
- 04Technical design
OutcomeFunctional and technical design are in place — delivery is plannable.
- Phase 3
Build & assure
- 05Development
- 06Testing & QA
- 07Documentation
OutcomeA tested, documented solution as code.
- Phase 4
Roll out & operate
- 08Acceptance
- 09Go-live & rollout
- 10Operations & evolution
OutcomeAccepted, live and supported in operations.
Frequently asked questions
How does working with you look in practice?
In the initial call we clarify your goals and current situation. You then receive a short proposal with a clear scope and defined deliverables. We deliver together with your team, document as we go and hand over so you can carry on independently.
Databricks, Microsoft Fabric or SAP Business Data Cloud — what fits us?
We don’t decide by preference but by your data, the share of SAP, your AI plans, existing skills and cost. You get a reasoned recommendation — often it is a combination.
Can we start with a small piece?
Yes, we even recommend it. For example with a platform review, a single AI use case, Terraform and CI/CD, introducing Unity Catalog or one master data process.

