GENAI AREDEZ
GENAI AREDEZ · AI engineering applied to business · MMXXVI · German industrial business ·
← All projects

German industrial business / AI and leadership development

A business leader who builds and directs his own AI systems.

A leader at a German industrial group wanted to solve reporting and learn to build his own solutions. We worked on his real process: architecture, data, security, automation and evaluation. He now creates skills, directs agents and develops new projects.

DiscoveryArchitecture and securitySkills and agentsInternal capability

Business collaboration · anonymised presentation
Supported development and validation · status documented as of October 2026

01 / The problem and its context

German industrial business

Reporting required consolidating information across companies, checking its provenance and detecting anomalies before delivering figures. The process owner wanted to understand and extend the system. We designed the programme so he could learn to build and make technical decisions about his own work.

Who did what

Ignacio led discovery, environment design, evaluation methods and technical guidance. The leader built and adapted the tools in his own repository, using his data and business criteria. Each session combined explanation, decisions and work on the actual system.

A view of what was built

The knowledge stays inside the business.

The documentation records an observable change: from getting started with the tool to creating procedures, directing reviewers and developing another project. The deliverables help retain that knowledge and use it in everyday work.

Process / Guide / Checks / Continuity

  1. 01

    Internal knowledge

    Process, criteria and sources

    The system starts with how the business works.

  2. 02

    Working architecture

    Rules, skills, permissions and handover

    Knowledge is stored and can be reviewed.

  3. 03

    Evaluation capability

    Agent audits and control tests

    The owner decides what to accept and correct.

  4. 04

    Building capability

    Another project, built by him

    Learning applies beyond the first process.

Anonymous reconstruction of documented capabilities and deliverables. No client documents, names or financial data.
A view of the work

Understand. Operate. Create. Direct.

The programme was designed to change what the owner can do with AI. Each step is practised on his own process.

  1. 01

    Understand

    • Map the process and its decisions
    • Know which data the system uses
  2. 02

    Operate

    • Request, review and verify
    • Resume work with context
  3. 03

    Create

    • Write his own skills
    • Build automation and another project
  4. 04

    Direct

    • Assign reviews to multiple agents
    • Challenge results and decide corrections
A view of the work

An owned foundation for continued development.

Knowledge, controls and tools have different roles. Separating them supports security review and system evolution.

  1. 01

    Business knowledge

    • Process, criteria and documentation in his repository
    • Guides and lessons that can be updated
  2. 02

    Security and control

    • Permissions for reviewing sensitive actions
    • Protected originals, tests and human validation
  3. 03

    Tools and providers

    • Transferable methods and architecture
    • Adapt integrations and retest when switching

02 / Engineering decisions

Business knowledge becomes an internal capability

Business knowledge becomes an internal capability Explanatory diagram

A real problem + an owner who learns by building

01Understand and protect

  • Process and criteria

    Sources, decisions and human validation

  • Working foundation

    Repository, permissions and data rules

02Build and test

  • Skills and automation

    Procedures built around his work

  • Agents and controls

    Independent reviews and adversarial tests

03Direct and extend

  • Own judgement

    Review, correct and decide what to automate

  • Next project

    Apply the learning to another need

↺ Tests improve the system; learning enables the next project.

System decisions and boundaries

The programme develops transferable architecture, methods and judgement. The working environment is Claude Code; switching providers requires adapting and testing integrations. Demonstrated learning does not mean final reporting acceptance.

01

Learn on the real business problem

Reporting provides the constraints, data and usefulness criteria. Each technical concept is applied to the work the leader needs to solve.

02

Enable the owner to build

The client creates and executes in his repository. Ignacio guides methods, design and corrections, explaining decisions so the owner can repeat them.

03

Test before extending

Test data, instructions and controls. The owner retains decisions affecting figures and learns to review what agents propose.

04

Retain knowledge and judgement

Rules, guides and procedures support continuity and extension. The second project demonstrates application beyond the original process.

Operations, automation and controls

How it works and is maintained

Security that is also learned

The initial configuration restricts destructive actions and requests review for sensitive operations. Sessions teach the owner to read what the agent proposes, understand the permission and decide. Security needs technical mechanisms and the owner's judgement; a written rule alone is insufficient.

Test a control by trying to make it fail

Invalid inputs are paired with valid cases. The work exposed checks that accepted internally consistent figures without comparing them to sources. Learning this distinction changes how the owner evaluates the next automation.

Simplification is an architecture decision

Diagnosis identified instructions that turned recommendations into tasks and accumulated mechanisms. Causes were reviewed, session closure simplified and current rules separated from history. The owner learns to question agent proposals before adding complexity.

Transferable knowledge, specific tools

The client retains the process, decisions, guides and procedures. The programme teaches architecture, context, delegation and evaluation that apply with other tools. Platform-specific skills and integrations need adaptation when migrating.

03 / How it is evaluated

How it is evaluated

Process instructions and controls were reviewed, with findings and corrections addressed. Tests distinguish invalid cases that must stop from valid cases that must pass. Acceptance of figures remains with the owner. End-to-end reporting completion remains subject to validation.

What was built

Demonstrated internal capability: the owner creates skills, coordinates agents, runs audits, leaves processes running and has started a second project. He retains the repository, instructions, procedures and guides to maintain and extend what he built. The comprehensive review and final reporting acceptance remain pending.

Supported development and validation · status documented as of October 2026

What we can bring to your project

Developing leaders who can direct work with AI