Applied AI / Automation / Reliable agent systems

Building AI systems that can be inspected, governed and used in practice

I combine applied AI with more than 20 years in enterprise IT, automation, operations and cybersecurity.

Where would you like to start?

What can Fredrik build?

Explore selected work

Four flagship systems with implementation evidence, demonstrations and stated limits.

See projects →Which way of working fits my problem?

Find a method

Start from the problem you have. 20 methods, sorted by what they help with.

Find a method →Can I try something?

Try the labs

Short interactive exercises that run in the browser on synthetic data.

Open the labs →

Or go straight to the 46 articles, the Build log or all pages.

Harnessing / Learn by trying

What turns a model into a useful working system?

Follow an agent at a freight terminal, inspect the methods at the moment they matter and explore how workflows become pipelines.

15 methods / 30 workplace situations / 6 pipelines

Harnessing Library

Explore source orientation, boundaries, verification, coordination and continuity through concrete examples and small interactive labs.

Explore the Harnessing Library →

Selected work

Four current flagships

All projects →

Retrieval and graphs

Intelligence Engine

Hybrid lexical, semantic and graph retrieval with REST, MCP and an interactive web UI.

Explore →Human-in-the-loop agents

MADS

A controlled development environment where agent proposals become reviewable changes.

Explore →AI product engineering

Adaptivearts.ai

Applied AI product engineering, editorial workflows and a server-side provider boundary.

Explore →Runtime governance

Gate Monitor

Policy, evidence and explicit decisions for long-running agent sessions.

Explore →

Now buildingIn development

Blueprint AI Studio

A learning platform for From Blueprint to Application. Try the public reader labs today and follow the development of the companion studio and LMS.

2026-09-28CBE and CBE-IX added to the reader labs

Available nowRules, Authority & Enforcement reader labs

Coming soonBlueprint AI Studio and its learning platform

Follow the progress →Try the reader labs →

Writing

Latest articles

All 46 articles →

The same woman appears five times in one warm office scene, each performing a different part of a control system: pinning up a printed notice, presenting an access keycard at a reader, holding open a glass door, reviewing a printed checklist, and closing a storage box.

Technical Guide2026-09-17

When a Rule Is Not a Control

A rule can be present, understood and structurally valid while the boundary it describes still isn't reliably enforced. A practical look at instruction, authority, enforcement, verification and preservation for AI agents, with hands-on reader labs.

Read on Adaptivearts.ai ↗

Technical Guide2026-08-26

When Names Stop Meaning the Same Thing

AI systems often encounter names, paths, repositories and pages that appear to refer to the same thing. This article shows how to separate observed surfaces from candidate subjects, evaluate identity from evidence and preserve unresolved cases instead of turning labels into identity verdicts.

Read on Adaptivearts.ai ↗

Professionals in an operations room in front of a wall display where a noisy tangle of lines resolves into a clear ordered system.

Essay2026-07-22

Are Companies Really Ready for AI? The Real Problem Is Not the Technology

AI readiness is not a technology problem. The greatest advantage goes to organisations that can describe, observe, govern, automate and verify how they work.

Read on Adaptivearts.ai ↗

Build log Latest published update, 2026-11-07: Harnessing Library: learning by following workplace situations. Open the Build log →

About

Systems and Automation Engineer

Applied AI, operations, security, integration and agent-oriented infrastructure. Independent initiative: Adaptivearts.ai ↗

GitHub ↗

How I work: evidence before confidence

Pin source and implementation receipts.
Separate tests, runtime evidence and documentation claims.
Use synthetic data in public demonstrations.
Keep limitations beside strengths.
Route risky actions through policy or human approval.
Preserve earlier evidence when systems evolve.

Earlier showcases remain in the project archive.