Public field lab · AI operations research

Human judgment and automation in the right place.

Systems Detective investigates the hidden architecture of work in the age of AI agents.

I document live-coded projects, workflow failures, governance questions, and the boundary between human judgment and automation, in public, inspectable form.

What this is

A public field lab, not an agency website.

Systems Detective is an independent project where I document live-coded experiments, operational diagnostics, and the practical boundary between human judgment and automation.

Current status

Systems Detective is not operated as a registered company and does not publish fixed commercial packages or prices. Serious collaboration, research conversations, interviews, and advisory work are welcome when there is clear fit.

The Proof

Built, not pretended.

Most of the work stays private. What can be shown is public, inspectable, and kept current. Four entry points into the full index.

Proof of Work · live trace

8.15B tokens, 148 days operating

A live, locally read trace of where AI spend creates value, waste, or risk. Not a trophy number, a working record.

Open the live trace →
Case study · Curriculum

AI Evaluation Workshop

A complete, teachable evaluation method covering evals, datasets, guardrails, red teaming, and quality gates.

Read the case study →
Case study · Production playbook

Voice Agent Deployment Kit

What it actually takes to ship a voice agent under EU regulation, mapped across 39 modules before launch.

Read the case study →
Open skill · Prompt system

Call Readiness Coach

An evidence first preparation drill, public and installable as a skill, that refuses the comfortable hero story.

Read the case study →

See the full proof index →

The Method

The Iceberg Method™

Most operational problems are not visible at first. The visible symptom is usually only 10 percent: slow work, inconsistent decisions, tool confusion, missed handoffs, or client friction. The hidden 90 percent is ownership, data structure, decision rules, review points, exceptions, and accountability. The method maps the hidden structure, then produces a clear operating recommendation, before a tool, agent, or automation is expanded.

Iceberg diagram. Above the waterline, 10 percent is the visible symptom. Below it, 90 percent is the hidden structure where the work is.
The Decision Roadmap

From leak detection to execution.

A thinking tool, not a product: it separates high judgment work, agent assisted work, and repeatable operations.

I

Human Judgment

For decisions that need trust, discretion, responsibility, or senior review.

Use when
  • The cost of being wrong is high
  • The client relationship matters
  • The context is sensitive
  • The decision needs ownership
II

Agent Support

For work that benefits from speed, memory, research, structure, and synthesis.

Use when
  • The team needs better inputs
  • The data is messy
  • Patterns are hidden
  • A human still owns the decision
III

Safe Automation

For stable workflows where the rules are clear and the risk is controlled.

Use when
  • The task is repeatable
  • The inputs are structured
  • The output can be checked
  • Failure is reversible
What I work on

Domains I investigate and build in public.

These are not fixed packages. They are working capabilities I develop, document, and test across live projects.

Deeper work on regulatory readiness, decision dashboards, and continuing counsel is discussed privately when a first conversation confirms fit. It is not offered as a public menu.

Good fit, bad fit

Good fit, bad fit.

Good fit

You lead a workflow that affects clients, revenue, reputation, compliance, or team performance.

You are considering tools, agents, automation, or workflow redesign, but the real issue may be ownership, data structure, decision rules, or review.

You want a clear operating decision before spending more money, adding more tools, or creating more risk.

Bad fit

You want a prompt pack, chatbot setup, or quick tool install.

You want someone to automate a broken workflow without questioning it.

You are not willing to name owners, review points, exceptions, or decision rules.

Field Notes

Recent writing.

Essays on automation, operations, human judgment, and decision quality, written as the work happens.

23 June 2026 · Essay

You’re Not Preparing. You’re Keeping the Lie Alive.

Read the essay →
09 June 2026 · Essay

A major AI lab shipped the gap. A Japanese builder shipped the fix.

Read the essay →
03 June 2026 · Essay

I Read the Study About AI Breaking the Law. Then My AI Deleted My Files.

Read the essay →

See all field notes →

The Practitioner

Otman

Otman, El Capitano

Systems Detective is led by Otman, a business engineer working at the intersection of workflow design, human judgment, and accountable automation.

The project documents what I build, test, and learn in public. Training and advisory work across 700+ professionals inform the thinking. The site shows the inspectable record.

Read the full bio →

Situation Estimator

Situation estimator.

Four questions. A reading of where the friction likely sits, not a quote, not a package.

1. What kind of workflow do you want to improve?
2. What is the current pain?
3. How urgent is this?
4. If this goes wrong, what is exposed?

Reading of the situation

Start a Conversation
Start a Conversation

Start with the situation.

Use this for serious notes, research conversations, collaboration ideas, interviews, project related questions, or thoughtful disagreement. Systems Detective is currently an independent public project. I read serious messages personally. Or write to otman@systemsdetective.com.

By sending this form, you allow Systems Detective to use the information you provide to read and reply to your message. Your data is not sold. Do not include confidential, sensitive, or client data through this form. You can request access, correction, or deletion at any time. See the Privacy Policy.

Replies when possible. Messages are treated as confidential.

Field Notes

Notes worth sending.

Occasional writing on automation, operations, human judgment, and decision quality. Sent only when there is something worth saying.