Build TBD Innovation Labs / AI Research
We investigate how people and AI can do useful work together. That means asking difficult questions, examining the assumptions, and making the reasoning visible. Our research connects context, decisions, work design, and agent operations to questions organizations can test in practice.
Discuss a research questionAn agent can produce a convincing output while the surrounding work remains slow, expensive, or difficult to trust. Our research asks what the whole system needs: relevant context, a clear decision, explicit responsibilities, and a way to judge the result.
The purpose of the lab is to develop knowledge that can become useful practice: a method a team can apply, an experiment it can repeat, or a capability it can evaluate and license.
Four connected research areas
Context and inference
How can an agent find the relevant context without carrying everything?
Explore bounded navigation and inference over permitted records, with attention to relevance, missing information, conflicting sources, and the cost of retrieval.
What to examine: Compare source relevance, important omissions, and retrieval effort on an agreed task and dataset.
Current scope: Available for a scoped capability evaluation against your use case and agreed evidence.
Decisions and evidence
Can a team reconstruct why a decision was made—and what happened next?
Connect the question, supporting evidence, alternatives, accountable owner, delegated work, and returned result. Study how decisions remain understandable as people and agents hand work over.
What to examine: Evaluate reconstruction effort, missing evidence, and the receiving team's ability to continue the work.
Current scope: Method and illustrative examples are available; broader outcome claims require a defined study.
Human and agent work
What must change in a workflow when agents become participants?
Develop a body of knowledge, practical instruments, and learning materials for context, responsibilities, authority, handoffs, measurement, and recovery. Connect engineering and continuous-improvement practice to agent work.
What to examine: Test whether a redesigned workflow improves an agreed outcome while tracking review burden and failure modes.
Current scope: An evolving discipline with public writing and tools in development; partner and curriculum pathways are being developed.
Coordination and delivery
How should agents plan, exchange work, and coordinate toward an accepted result?
Investigate Scrum-inspired planning, transferable work packets, receiver-checked handoffs, and swarm orchestration. Parallax provides an environment for developing and examining these operating practices.
What to examine: Compare time to accepted delivery, coordination cost, rework, and recovery—not just the number of agents running.
Current scope: Applied research agenda and internal operating practice; customer results and portable effectiveness need separate evaluation.
The research collection
Read specific observations with their context and limitations. These are the starting points for discussion and further testing.
Explore the Human-Agent Alignment Map, Decision Thread, and the developing ACD practice. Worked examples explain a method; they are not customer outcome claims.
The Agent Centered Design field guide and learning materials are in development. Follow the research or discuss a learning programme tied to the work of your team.
A proposed ACD knowledge-access pilot would make a curated, versioned collection available through MCP with attribution. Licensed collections and access are scoped separately from public reading.
Public pages introduce the work. A scoped evaluation can include agreed evidence and technical discussion. Protected implementations, client records, and restricted source materials remain outside the public collection.
Fund a question worth answering
For business, innovation, and operations leaders with a specific workflow question. Fund a bounded investigation with a baseline, agreed measures, milestones, and a decision at the end.
You receive: a study brief, agreed experiment outputs, findings and limitations, and a recommendation to continue, change direction, or stop.
Commercial shape: a scoped research contract or milestone-based programme.
For AI developers, consultancies, and platform teams exploring a shared capability. Co-develop an evaluation or research programme around an agreed integration or delivery problem.
You receive: a joint work plan, defined contributions, evaluation criteria, review points, and a proposed route to adoption.
Commercial shape: a funded collaboration with separately agreed licensing and support.
For investigators and institutions developing a study, replication, or shared research question. Discuss complementary expertise, permitted data, methodology, and publication aims.
You receive: a proposed protocol and clear responsibilities for the research and its outputs.
Funding shape: a project agreement defining contributions, funding, and publication responsibilities.
For organizations interested in a question that spans more than one workflow. Discuss support for a defined sequence of studies, methods, or educational outputs.
You receive: an agreed research agenda, milestone reviews, and specified reports or briefings.
Commercial shape: a programme agreement with a defined budget, duration, and outputs.
Funding supports agreed research effort and deliverables. It does not buy a predetermined conclusion or automatically confer exclusive rights. Data use, confidentiality, ownership, publication, and any license are agreed before work starts.
| Your organization | Explore | Next step |
|---|---|---|
| Corporate teams | Internal method use, maintained instruments, learning, and continued practice. | Corporate practice |
| AI consultancies | Scoped client-delivery rights and supported partner pilots. | Delivery partners |
| Agent developers and platform teams | A defined technology evaluation, integration, or OEM use. | Discuss research licensing |
| Practitioners and learning teams | Public research, practical training, and curated reference access. | Knowledge access |
Each proposal identifies the actual assets, permitted use, evidence available, updates, support, and commercial terms. Research funding, technology licensing, and implementation services are separate scopes that can be combined where useful.
Bring the question, the team it affects, and what an answer would let you decide. Meet Cory Nation, or explore implementation and leadership services if the next step is already clear.
Tell us what needs to work better.
Discuss a research question