AIDC Research
Research
We study how AI is built, deployed and used inside real organisations: the infrastructure that lets agents persist and scale, how organisations of agents govern and improve themselves, and which kinds of work, in which industries, AI actually takes on. Our papers come with their data and the code that produces every number.
Research areas: Economics of AI Agent systems Organizations Domain intelligence
Economics of AI
Where AI is used across the economy, what the work it touches is worth, and which kind of AI each industry needs.
Agent systems
Runtimes, orchestration and evaluation for agents that persist, scale and work together.
Organizations
How organizations of agents share tools, improve themselves and stay governed.
Domain intelligence
Bringing agents to specialist domains without retraining, starting with life cycle assessment.
Precision or Automation? The Value–Population Ratio and the Demand for AI across Industries
Industries that create more value per worker use more AI per worker, through pay rather than capital and largely because their work is exposed. Task length, not pay, drives compute and failure; the cheap automation that labour-dense industries need lies in short tasks that usage data barely record. A study of 65 US private industries and the Anthropic Economic Index.
AIDH: An Agent-Independent Distributed Harness for Persistent Agents
Agents usually live and die with one process on one machine. AIDH keeps no instance state between requests, so a persistent agent costs in proportion to its work rather than its existence, recovers from failed machines without losing requests and runs each execution inside its own access closure.
Neural Orchestration: Spatio-Temporal Communication for Large-Scale Multi-Agent Systems
Instead of messaging each other, agents act one at a time in a shared, partitioned task space, and the order of their commits is the channel. Orchestration reduces to typed decisions answered by a small distilled encoder at no token cost.
From Language Models to Agent Systems: A Survey of Training, Architectures, Orchestration, and Open Problems
A review of 434 works across nine layers, from pre-training to self-improvement and evaluation, with a taxonomy, a 2017–2026 roadmap and three open problems stated as testable hypotheses.
Publications
| Date | Area | Title |
|---|---|---|
| Sep 30, 2026 | Economics of AI | Precision or Automation? The Value–Population Ratio and the Demand for AI across IndustriesPaper |
| Sep 26, 2026 | Agent systems | AIDH: An Agent-Independent Distributed Harness for Persistent AgentsWorking paper |
| Sep 23, 2026 | Agent systems | Neural Orchestration: Spatio-Temporal Communication for Large-Scale Multi-Agent SystemsWorking paper |
| Sep 23, 2026 | Agent systems | From Language Models to Agent Systems: A Survey of Training, Architectures, Orchestration, and Open ProblemsSurvey |
| Sep 25, 2026 | Domain intelligence | Closure Instead of Training: Agentic Domain Modelling for Life Cycle AssessmentWorking paper |
| Sep 25, 2026 | Domain intelligence | From Neural Surrogates to Language Agents: A Survey of Artificial Intelligence for Life Cycle AssessmentSurvey |
| Sep 26, 2026 | Organizations | Recursive Self-Improvement in Self-Organizing Multi-Agent SystemsWorking paper |
No publications match.