Marcelis Digital
Menu

AI Software Factory

Turn AI-assisted engineering into durable delivery performance.

Read more about the AI Software Factory

Marcelis Digital helps CIOs and CTOs redesign software delivery, automate high-value workflows and resolve the architecture underneath. From initial analysis and the business case, we can stay with the work through architecture, execution and handover.

Discuss your situation Explore capabilities

Capabilities

Three ways into the same problem.

AI, architecture and operating model are no longer separate conversations. The work starts where the enterprise is stuck and follows the dependency chain until the answer can be built.

01

AI Software Factory

Redesign the software delivery system around AI — the workflows, context, controls and platform capabilities, not merely the licence count.

 02

Enterprise Agents & Workflow Automation

Redesign high-value work first, then build agents that retain context, use enterprise systems and operate within explicit authority.

 03

Technology Architecture & Transformation

Resolve the architecture, ownership and sequence underneath the strategy — especially when the transformation is already moving.

What we believe

A few positions. Deliberately not a manifesto.

I

AI changes the work, not just the toolset.

The value sits in redesigned decisions, workflows and feedback loops. Seat counts are procurement data.

II

Autonomy needs an authority model.

An agent must know what it may see, decide, change and escalate. Without that, it is either a demo or a future incident.

III

Architecture must leave evidence in delivery.

A principle that cannot change a decision, a boundary or an automated check is mostly wall decoration.

How the work moves

From decision to working system.

Strategy-led, technically executable. The format changes with the problem; the chain of accountability does not.

  1. 01

    Frame the value

    Business objective, economics, workflow, constraints and failure modes.

  2. 02

    Design the system

    Target process, architecture, controls, operating model and sequence.

  3. 03

    Build the critical slice

    Agent, workflow, platform capability or architecture intervention.

  4. 04

    Make it operable

    Security, ownership, evidence, hosting, handover and agreed maintenance.

Experience behind the firm

Work that survived contact with reality.

Selected experience from previous consulting and technology roles. Client names are withheld; the work is described on its merits.

National transport organisation

AI-enabled software delivery

Shaped the sourcing and target capability for Developer Experience and AI across the SDLC, including model access, enterprise context, AI-ready repositories, controls and measurable adoption.

European eMobility scale-up

Platform transformation recovery

Supported the CTO in resetting the target state, domain model, architecture and delivery roadmap for a materially delayed and over-budget platform transformation.

Large Dutch public-sector organisation

Enterprise architecture and operating model

Designed the architecture and operating model for a multi-year information-provision programme, then translated it into a 90-day execution plan.

View selected experience

Why Marcelis Digital

Senior judgment. Working depth. No mystery bench.

Marcelis Digital is led by Thomas Marcelis, a technology strategist and trained software developer with experience at McKinsey, BCG Platinion, Accenture and Eraneos. He works where executive ambiguity meets architecture, software and delivery.

Engagements stay principal-led. When the problem needs more capacity, Marcelis Digital assembles the right engineering, architecture, design and specialist expertise around it — explicitly.

Start a conversation

Point of view

Arguments worth testing.

Short positions on AI, software delivery and architecture. No trend reports assembled by committee.

AI Software Factory

The AI software factory is not a developer-tool rollout

AI changes software delivery when workflows, context, verification and controls change with it. Licence counts are an input, not an outcome.

 Enterprise agents

Enterprise agents need an operating model, not a personality

A useful enterprise agent needs a job, an authority boundary, governed context and an owner when it fails. Personality is optional.

 Enterprise architecture

Architecture as a decision system

Architecture earns its place when principles become decisions, tests and consequences. Static diagrams are useful only when they change what teams do next.
View all points of view

Start with the live question

Bring the technology problem that does not fit neatly into one workstream.

AI strategy, a workflow worth rebuilding, an agent that needs a real operating model, or a transformation whose architecture has stopped holding.

Start a conversation