ABOUT AURANOM
Can management consulting be carried out autonomously by agents?
AURANOM comes out of an independent research programme on multi-agent systems in management consulting — run by a practising management consultant and built by AI agents under one human lead.
Christian Mikolasch, MBA, CMCFounder · Human lead
01 · THE RESEARCH
One question drives the work
The independent research programme on multi-agent systems in management consulting asks whether management consulting can be carried out autonomously by agents. AURANOM is the framework built to explore that question in practice.
Out of this research came a 13-level model for structuring autonomous, AI-supported consulting processes, with the human at the top as the final escalation instance. This conceptual framework is now being tested in a field study; a prototype comes next.
Practice-oriented research
Practice-oriented research that develops, tests and improves solutions for real business problems.
Design science research
The framework is the research artefact: designed, built and improved in iterations.
Field experiments
Multi-agent systems tested in real work with VS Code, GitHub Copilot, Greptile and Claude.
How findings are kept
Every finding goes into a research log with its source. Observation is kept apart from interpretation.
HOW THE RESEARCH IS DONE
Field trials, not lab demos
The research field is multi-agent systems in management consulting: how teams of AI agents can take on real consulting and development work, and where the human has to stay in charge.
The work runs as field trials and experiments inside real development and consulting workflows, not in a lab setting.
02 · HOW AURANOM IS BUILT
Built by AI agents, under a human lead
AURANOM is developed by a team of AI agents, separate from the agents that work inside the products. There is one human participant and decision-maker in this project. Every other role — research, development, review, consulting, coaching — is carried out by AI personas. The agents write specifications, code and documentation; the human sets the goals, reviews the work and makes all design decisions.
- 1
Human lead
Sets goals, reviews and makes every design decision and every merge.
- 8Powered by AI
Development persona agents
Strategist, Discoverer, Researcher, Designer, Storyteller, Developer, Coordinator and Auditor — fictional personas, clearly labelled.
- +2Powered by AI
Review agents
A review agent and a multi-review agent check the work before it reaches the human.
Up to ten roles · sprints of up to twelve hours · several experiments in parallel · meet the agents
- Green · human
- Solid border · autonomous AI agents
- Dashed border · automatic
How a sprint works
Ten steps from plan to deploy. Agents do the work; the human approves the plan and decides every merge. The human decides how far the agents work on their own — up to the next approval gate.
Step 01: Plan
Audits the codebase against the docs and the five standards and proposes one sprint: an epic with at most five sub-issues.
Human starts and approves the plan
Owner Auditor agent · AI
Step 02: Prepare
Creates the issues, runs an AI impact and security check, opens one sprint branch.
Owner PM agent · AI
Step 03: Build
Implement the issues in parallel; every change comes with a test.
Owner Developer agents · AI
Step 04: Sprint audit
Checks every change against the acceptance criteria, the Definition of Done and the standards.
Owner Separate auditor agent · AI
Step 05: Hotfix loop
Every gap becomes a hotfix issue on the same branch.
Owner PM + developer agents · AI
Step 06: One pull request
One pull request per sprint, with a sprint report.
Owner PM agent · AI
Step 07: Staged review pipeline
Several review-and-fix stages (correctness, quality, security, architecture), a shared severity rubric and the repository’s own checks.
Owner Multiple independent AI reviewers
Step 08: Merge gate
Green only when all stages ran, all checks pass and no serious finding is open; otherwise flagged for the human.
Owner Automatic
Step 09: Human decision
Reviews the report and merges — or not. Every merge is a human decision.
Owner Christian Mikolasch · human
Step 10: Deploy
Automated deployment after the merge.
Owner Automatic
How content is made
The content pipeline behind AURANOM Busina: no article is published without a human approval.
Step 01: Topic
Chosen from the ongoing research.
Owner Christian Mikolasch · human
Step 02: Research
Gathers sources.
Owner Research agent · AI
Step 03: Write & critique
Three write-and-critique rounds.
Owner Writer agent + critic agent · AI
Step 04: Audit
Checks the draft against the critiques and the rules.
Owner Auditor agent · AI
Step 05: Human approval
Approve, or send back for revision. Revisions go through the audit again.
Owner Christian Mikolasch · human
Step 06: Quality checks
Grammar, style and image checks.
Owner Automatic
Step 07: Publish & distribute
Website first, then adapted versions for other channels.
Owner Automatic
03 · EVIDENCE
Evidence so far
- 12months of AI-led development
- 8experiment repositories, ~3,770 commits
- 241analysed review-pipeline runs
- 22logged findings from the review-pipeline field experiment
- HUMANEvery merge is a human decision
- up to 10agent roles: coding, writing and review agents
- 45+sprints across four product repositories, each planned as an epic
- ~790pull requests, each reviewed before a human merge decision
- 44logged design decisions in the decision log
- ~70versions of the architecture map
Figures as of October 2026, from the AURANOM research log and the project repositories.
04 · THE HUMAN LEAD
The pilot behind the crew
Christian Mikolasch, MBA, CMC
Founder · Human lead
Management consultant. For 14 years he led his own consultancy in St. Pölten — digitalisation, sales and change in B2B — from founding to a successful handover: 500+ mandates, a team grown from 3 to 8, a partner network of 20+ consultants. Lecturer at FH St. Pölten (2008–2011). Writing code since 1990.
His research on the architecture of consulting processes began with his MBA thesis (2023). It produced the 13-level model and the AURANOM architecture.
QUALIFICATIONS AND RESEARCH
“At the same point, knowledge always got lost: between closing the deal and starting the project.”
05 · PRINCIPLES
What we hold ourselves to
Standards first
Technically implementable requirements of ISO 9001, ISO 20700, ISO 21500/21502, ISO/IEC 27001 and ISO/IEC 42001, mapped as machine-checkable rules. No ISMS or certificate is claimed.
Human decides
One human makes every design decision and every merge. The human is the final escalation instance.
Transparent AI
Agents are labelled as AI, personas as fictional, portraits as AI-generated. The tools are named.
Evidence over claims
Findings are logged with sources; observation is kept apart from interpretation.
Meet the crew, or talk to the pilot
Get to know the AI agents behind AURANOM, or send a request about the research or the products.