Research · 8Hats Lab
How work changes when AI agents join the team.
We study teams of people and AI agents: how they share context, verify work, and make decisions. Explore the questions, public records, and limits of each research stream.
For the chief executive:who owns work produced by AI agents, how much effort review actually takes, and which operating rules the company writes down.
For the technical leader:task boundaries, the context an AI agent works from, verification, permissions, and failure handling.
Research streams
Three research streams
How teams of people and AI agents manage work, collaborate, and learn. Each stream links to its models, examples, sources, and limits.
Frameworks under AI agents
- The research question
- What happens to a management framework when part of the team is an AI agent?
- Aspect of work examined
- How work is managed: goals, roles, cadence, evidence, and decision rights.
- Current status
- 8Hats working proposal. Not validated against a controlled comparison.
Working with AI agents
- The research question
- Which parts of managing work do not transfer to an AI agent, however capable it becomes?
- Aspect of work examined
- Review, coordination, throughput, and release decisions.
- Current status
- Evidence review and an untested lifecycle proposal.
How AI agents learn
- The research question
- How does a capability get into a person and an AI agent at the same time, and how do we know it arrived?
- Aspect of work examined
- Learning, assessment, deployment, and observation in real work.
- Current status
- HALA specification. Controlled outcome validation and peer review remain open.
Evidence register
Public records, with their limits attached.
Every record keeps its publication status, permanent identifier, supported use, and limit in the same row. A dataset does not lend its standing to a preprint, and neither of them stands in for a result in any one company.
- Published · peer-reviewed journal
From Retrieval to Evidence: A Task-Adaptive Two-Layer Knowledge Architecture for LLM Agents
Machine Learning and Knowledge Extraction (MDPI), vol. 8, no. 10, article 322, October 2026. Authors: Aleksandr Volkov, Taras Pustovoy, Dmitriy Istomin, Tatiana Otbetkina.
MAKE (MDPI) · doi.org/10.3390/make8100322
- Supports
- A tested method that gives retrieval, a reviewed knowledge layer, and evidence admission separate jobs inside one grounded AI agent architecture.
- Limit
- Evidence from course-grounding studies and one observational production case; it does not show that an 8Hats product changes an organizational outcome.
- Public dataset
K-Forge Evaluation Dataset
72 questions across production courses, 7,818 model responses from ten LLMs, and 28,593 evaluative judgments.
Zenodo · CC BY 4.0 · doi.org/10.5281/zenodo.20633464
- Supports
- Comparative evaluation of model responses to production course-design questions.
- Limit
- A bounded course-design dataset, not a universal benchmark for AI agent trust.
- Public dataset
CourseFactory Workflow-Mining Dataset (De-Identified)
Dataset record and licensing information are available on Zenodo.
Zenodo · CC BY 4.0 · doi.org/10.5281/zenodo.20517429
- Supports
- Inspection of de-identified records used in workflow-mining research.
- Limit
- A dataset record; it does not show that an 8Hats product changes an organizational outcome.
- Accepted · IEEE ICDM 2026 workshop
Mining AI-Assisted Course Design Workflows at Production Scale
Accepted at the WMAI workshop held with IEEE ICDM 2026; the preprint is on Research Square.
Research Square · doi.org/10.21203/rs.3.rs-10310358/v1
- Supports
- A method and findings for mining AI-assisted course-design workflows at production scale.
- Limit
- A workshop paper; the IEEE workshop proceedings appear after the conference in November 2026.
These records come from course-design, workflow-mining, and knowledge-architecture research. They are evidence about evaluation, workflow mining, and how AI agents ground their answers. They do not validate the Scrum proposal, HALA, or any commercial outcome.
Research model
A working model of trust
A working model we use to organize measurement, not an established quantitative law. Nothing on this site is scored with it, and no readiness number is derived from it.
Publication discipline
A result keeps its object, status, and uncertainty.
- Open the object.Permanent records and named-author work link to their source.
- Show the status.Dataset, preprint, and under-review work do not borrow credibility from one another.
- Keep the limit.Context evidence does not become product efficacy evidence.
Questions worth testing with others
Measurement of alignment and reliability, human correction in production workflows, and how accepted lessons update organizational practice.
Research collaborationResearch collaboration
Ask about a research stream.
Tell us which research stream or question you want to discuss. The research team reads every inquiry.