LensReading which lens this session carries.

Function cockpit

Research & Development

Agents run research operations. Scientists own hypotheses and conclusions.

Autonomy0% / 74%

Phase: Phase 2 — Supervised execution

Connect scientific programs to resources, evidence and decision gates: retrieval, scheduling, data integrity, simulation and reproducibility are agent work; hypothesis selection, safety and scientific conclusions remain human.

Trust score

0

Touchless

0%

Human review

0%

Override rate

0%

Capacity released

20-35% research-operating capacity (no presumption that success rates improve)

Accountable human

Chief Scientific Officer

CoS agent: R&D Chief of Staff

Every agent, its ceiling, and its escalation path

Agent roster

Every agent working Research & Development

Name, tier, autonomy level, the ceiling it operates under, and the human it escalates to. If an agent is not on this list, it is not running in production.

10 of 10

R&D Chief of Staff

A3Active

Orchestrator desk · rnd-cos

Connects programs to resources, evidence and gates, and presents portfolio decisions with explicit uncertainty to the CSO.

Confidence89%

Tasks / 24h

214

Success

93%

Override

7%

CeilingCoordination only — no scientific or funding authority

EscalatesAny program crossing a stage gate requires the named scientific owner before progression.

Evidence Scout

A3Active

Origination desk · evidence-scout

Retrieves scientific, technical and patent material with full citation provenance and licensing status.

Confidence91%

Tasks / 24h

1620

Success

95%

Override

5%

CeilingLicensed sources only; no paywall circumvention

EscalatesAny citation that cannot be resolved to a verifiable source is withheld and reported.

Hypothesis Mapper

A2Active

Origination desk · hypothesis-mapper

Organizes competing explanations, evidence strength and contradictions without selecting a preferred truth.

Confidence78%

Tasks / 24h

142

Success

86%

Override

18%

CeilingStructuring only; never asserts a scientific conclusion

EscalatesContradictory high-quality evidence is surfaced to the principal investigator unresolved.

Experiment Designer

A2In review

Execution desk · experiment-designer

Proposes protocols, power calculations and resource requirements where an approved method already exists.

Confidence82%

Tasks / 24h

88

Success

87%

Override

17%

CeilingApproved method library only; no novel protocol authorization

EscalatesPhysical experiments, hazardous materials and protected subjects require named human authorization.

Lab Scheduler

A3Active

Execution desk · lab-scheduler

Coordinates equipment, samples, reagents and personnel against the authorized experiment queue.

Confidence93%

Tasks / 24h

412

Success

96%

Override

4%

CeilingScheduling within authorized capacity only

EscalatesResource conflicts affecting a critical-path program escalate to the program lead.

Data Steward

A4Active

Execution desk · data-steward

Verifies dataset lineage, permissions, licensing and integrity before any analytical use.

Confidence95%

Tasks / 24h

686

Success

97%

Override

3%

CeilingBlocking authority on data quality; no analytical interpretation

EscalatesUnlicensed or unattributed data halts the study and notifies research governance.

Simulation Agent

A3Active

Execution desk · simulation

Runs authorized models and parameter sweeps within an approved compute and cost envelope.

Confidence88%

Tasks / 24h

264

Success

91%

Override

9%

CeilingApproved models only; compute budget capped per program

EscalatesResults outside the validated parameter domain are labeled as extrapolation and withheld from decisions.

Reproducibility Agent

A4Active

Oversight desk · reproducibility

Checks methods, code, environment and traceability so that any reported result can be re-derived.

Confidence94%

Tasks / 24h

176

Success

94%

Override

6%

CeilingVerification only; cannot approve a scientific claim

EscalatesA failed reproduction blocks internal publication and routes to the principal investigator.

Invention Agent

A2Active

Realization desk · invention

Captures novelty evidence, contributor records and disclosure timelines for candidate inventions.

Confidence86%

Tasks / 24h

34

Success

92%

Override

8%

CeilingRecord keeping only; no filing or disclosure authority

EscalatesAll patent strategy and external disclosure decisions route to Legal and the CSO.

Portfolio Agent

A3Active

Realization desk · portfolio

Tracks milestones, dependencies, option value and stage-gate readiness across the research portfolio.

Confidence87%

Tasks / 24h

118

Success

93%

Override

7%

CeilingReporting only; no funding or termination authority

EscalatesMilestone slippage beyond the published tolerance escalates to the portfolio board.