Sascha Denker

AI Transformation & Integration

From an AI idea to a process that actually runs.

AI transformation and integration for companies and organisations.

  • Consulting
  • Integration
  • Process automation
  • Training
Discuss a projectSee services
  • M.Sc. Artificial Intelligence
  • SAP Developer & AI Architect
  • Enterprise Engineering Background

Services

Three areas, one continuous path.

In an enterprise, AI rarely fails because of the model. It fails on integration, permissions, ownership — and on pilots that never make it into day-to-day operation. That is exactly where I work.

  • 01

    AI Consulting

    From an idea to an architecture you can build.

    We establish which use cases will hold up, what they require technically, and in what order they should be tackled — with a candid view of what is production-ready today and what is not.

    • Use case assessment and prioritisation
    • Feasibility, effort, build vs. buy
    • AI and integration architecture
    • Security and governance concept
    • A roadmap rather than a one-off initiative
  • 02

    AI Integration & Automation

    From an architecture to a process in production.

    Implementation happens inside your existing system landscape, not alongside it: connecting to the systems you already run, integrating tools and agents under control, securing them, and moving them into regular operation.

    • Enterprise integration: SAP, ERP, CRM, documents, knowledge systems — on-premise as well as cloud
    • Locally hosted language models, where data cannot leave the company
    • MCP and AI tool integration
    • Agentic workflows with a human in the loop
    • End-to-end process automation
    • Permissions, approvals, auditability
  • 03

    AI Enablement

    From a pilot to something your organisation actually uses.

    A solution only its builder can operate is not a solution. Knowledge transfer, training and a clean handover are part of the project — not a follow-up offer.

    • Workshops for staff, developers and leadership
    • Keynotes and talks
    • Knowledge transfer and documentation
    • Support after handover

Enterprise AI & SAP Integration

SAP + AI

Bringing AI into an SAP landscape does not mean putting a language model in front of the ERP. It means exposing precisely scoped capabilities under control: with identity, permissions, defined boundaries and decisions you can reconstruct afterwards. That explicitly includes on-premise systems: landscapes grown over years that are not in the cloud and, for good reasons, are not going there. That is the difference between an impressive prototype and something that survives an audit.

Integrating MCP servers is the easy part. Establishing a controlled process is the work.

  • MCP & Agentic System Integration

    Making SAP capabilities available to AI agents under control, and connecting existing MCP servers to agents and development tools — as clearly scoped tools with a defined blast radius, rather than open system access.

  • Agentic SAP

    Integrating agents into development and business processes — with a human in the loop exactly where something is written, approved or decided.

  • Secure Enterprise AI

    Identity and SSO, permission models, principal propagation, tool boundaries, controlled write access, review gates and auditability — integrating with the systems you have instead of growing shadow IT.

AI Tool Governance

Connecting an MCP server takes an afternoon. The real work starts afterwards: who approved it, what it is allowed to do, who reviews changes, and how new versions and new risks are handled. That calls for a process, not an integration.

From assessing existing community MCP servers to building and integrating bespoke ones: moving MCP servers and AI tools into your own landscape in a structured way — with defined roles, permissions, security reviews, and clear approval and lifecycle processes.

  1. Discover
  2. Assess
  3. Security review
  4. Architecture review
  5. Approve
  6. Integrate
  7. Monitor
  8. Update

The process is adapted to the organisation you have — not the other way round.

Customer-owned MCP Integration

Bespoke MCP integrations designed and implemented inside the client’s own infrastructure and governance. Source code, deployment and operations stay with the client.

I do not build my product for your company. I implement your solution inside your landscape. The code lives in your repositories from day one and is deployed into your infrastructure.

The client

  • defines purpose and requirements
  • acts as product owner
  • owns repository and infrastructure
  • decides on deployment and operations
  • decides on the lifecycle
  • is responsible for the resulting solution

My role

  • Consulting
  • Architecture
  • Implementation on the client’s behalf
  • Security and governance concept
  • Documentation
  • Handover
  • Enablement

How a project runs

  1. Assess
  2. Architect
  3. Implement
  4. Validate
  5. Hand over
  6. Enable

Discuss SAP + AI

AI Process Automation

Processes that run on their own.

The starting point is never a tool. It is a recurring piece of work that costs time today: a request that has to be sorted, information collected from four systems, a document somebody retypes. Those are analysed, automated, and extended with AI where that genuinely improves the process.

Not another chatbot. Better business processes.

How it works

  1. Understand the process
  2. Assess the potential
  3. Automate
  4. Add AI
  5. Secure it
  6. Put it into production

What comes out of it

  • Classifying incoming requests and preparing them for a decision
  • Bringing together information from email, documents and the ERP
  • Automating recurring business processes
  • Removing manual steps
  • Taking routine load off people through AI-assisted support

Recurring patterns turn into reusable building blocks. That shortens every further automation effort in the organisation — and makes the results comparable instead of ad hoc.

Typical applications

  • Document processing
  • Email workflows
  • ERP
  • CRM
  • Knowledge management
  • Reporting
  • Decision support
  • Development workflows
  • Enterprise integration

Discuss automation

AI Enablement & Training

Tools alone do not change how people work.

Rolling out access to an AI tool takes an hour. Getting people to work better than before is a different task. My training is technical and hands-on — not an overview talk about how AI changes everything, but concrete work on the tasks each audience actually has.

Audiences

  • Staff

    Generative AI in everyday work — used safely, productively and with a realistic sense of its limits.

    • Prompting that works in a real work context
    • Research, document and data analysis
    • Putting AI agents to sensible use
    • Safe use and handling of company data
    • Recognising the typical failure modes

  • Developers

    AI-assisted software engineering that holds up in a team, not just in a demo.

    • AI coding with Claude Code, GitHub Copilot, Codex and comparable tools
    • Agentic development
    • Understanding and applying MCP
    • AI-assisted engineering in existing codebases
    • Safe tool use, review and governance processes

  • Leadership

    A basis for decisions instead of buzzwords: what pays off, what it costs, and what it means organisationally.

    • Identifying and assessing viable AI use cases
    • Build vs. buy
    • AI agents and their organisational consequences
    • Governance and the EU AI Act
    • Risks and the transformation of ways of working

Formats

  • Workshops & TrainingBuilt around the systems and tasks of the participants, with hands-on work rather than slides.
  • Keynotes & TalksTalks for conferences, leadership circles and internal formats — technically grounded, accessible, and free of hype.
  • Knowledge TransferStructured handover of architecture, operations and further development to your teams, so the solution keeps running without me.

Enquire about trainingEnquire about a keynote

Approach

How a project runs.

Seven steps that build on each other. Each one produces a result you keep, even if we stop there.

  1. Understand the process

    What actually happens today — not what the process manual says.

  2. Assess the potential

    Where automation pays off, where AI pays off, and where neither does.

  3. Design the architecture

    Systems, interfaces, permissions, boundaries and the operating model.

  4. Build a pilot

    One clearly scoped, real use case instead of a feasibility demo.

  5. Secure it

    Security review, permissions, approval gates and auditability.

  6. Hand over

    Code, documentation and operational ownership move to your team.

  7. Transfer the knowledge

    Your people can extend, operate and judge the solution themselves.

About

Understanding AI, not just using it.

I work where enterprise systems meet applied AI — as a developer, architect and consultant.

My background is software and systems engineering in corporate environments: IT and business consulting, CTO of a mid-sized medical technology company, and today SAP development inside a corporate group. In medical technology, evidence, sign-offs and documentation are part of everyday work; the same questions are now being asked of AI under a different name. Today my focus is integrating AI into existing, complex and regulated system landscapes — the kind where permissions, traceability and operational ownership are not an afterthought.

On top of that sits an academic AI background. It helps less with operating tools and more with the questions that matter: what a model can genuinely do, where it stops, and which architecture will actually carry a requirement. The practical experience with AI architectures in complex and regulated enterprise environments comes from day-to-day work.

What I deliberately do not offer is general software development to order. The engineering background is what the consulting rests on — it is not the product.

Master's thesis

Comparing Temporal Discretizations and Parallel Implementations for Resonate-and-Fire Neurons

Graded 1.0.

Contact

Let’s discuss your project.

An email is enough — no form, no tracking, no cookies.

Each topic has its own address; that is how I sort enquiries without using any analytics. Subject and a draft message are prepared.

  • Discuss a project

    You have something in mind that has not taken shape yet. We work out together what it could become.

    ssd-en-discuss-project-0926@denker.space

  • SAP + AI

    MCP, agentic SAP, tool governance or a customer-owned integration in your SAP landscape.

    ssd-en-sap-ai-0926@denker.space

  • AI Process Automation

    A recurring process that costs too much time or is too error-prone — and should be automated.

    ssd-en-ai-automation-0926@denker.space

  • Training & Workshop

    A workshop for staff, developers or leadership — built around your systems.

    ssd-en-ai-training-0926@denker.space

  • Keynote & Talk

    A talk for a conference, a leadership circle or an internal format.

    ssd-en-keynote-0926@denker.space

Legally binding contact address: ssd-impressum@denker.space

The topic addresses encode nothing but the topic, the page language and the month — no identifier that relates to you as a person.