Standard operating procedures · Research & Development
The Research & Development procedure manual
33 written procedures across 11 teams, made of 184 numbered steps. 40 of those steps are a person deciding and 144 are agent work … more
Procedures
33
Median 6 steps long.
Numbered steps
184
Every one carries exactly one actor.
Decision points
40
22% of this function's steps stop for a person.
Agent steps
144
54 of them on the shared spine.
Named agents
86
Registered against the teams inside this function.
Units a week
88,371
Across every procedure in this manual.
Who this manual belongs to
Accountability above the procedures
Accountable
Chief Scientific Officer
The end of the escalation path for every procedure in this manual. more
Chief of staff
R&D Chief of Staff
Holds the operating rhythm of the function: the cadence these procedures are reviewed on and the queue of changes waiting to be made to them.
Autonomy across this function
62% now · 74% target
This is the function’s declared autonomy, not a reading taken from the procedures. Measured off the procedures themselves, 78% of the steps in this manual run without a person. The two are different questions and they do not have to agree.
The service level
Experiment record completeness
accuracy · target 95 % · Calendar month
Nothing measures this yet. The target is declared; the attainment is not.
The electronic notebook is one of the two systems with no feed at all. more
Governance gates on this function
4 declared
These are rules about the function as a whole, not about one procedure. more
| Gate | What trips it | Who releases it | Clock | Waivable |
|---|---|---|---|---|
Physical experiment authorization approval | Any laboratory experiment involving materials, equipment or subjects | Principal investigator and safety officer | 2 business days | No |
External data acquisition approval | Any new external dataset, license or third-party evidence source | Research governance and Legal | 3 business days | No |
Publication or external disclosure approval | Any material leaving the research boundary, including conference abstracts | CSO and General Counsel | 5 business days | No |
Reproduction sampling oversight | Independent re-derivation of 15% of reported internal results | Reproducibility board | Monthly | No |
Run any procedure in this manual
33 to choose from
This is a run, not a forecast. One unit enters the procedure and is walked station by station in the recorded order, each station taking exactly the dwell the estate records for it, held by whoever the estate says holds it. The clock you watch is worked time: the sum of the time the work is actually being done. Pick a procedure on the left and press run. Happy means nothing goes wrong: no station raises an exception, no gate sends the unit back, no rework loop fires and nothing sits in a queue between stations. Viable means the path is the one the estate already permits: the lanes are the recorded lanes, every decision that stops for a person still stops for that person, and no step is moved unless the owning team has autonomy ceiling left to cover it. It is a floor for how fast this procedure can go while staying inside its own rules, not a promise about any particular Tuesday.
Nothing has happened yet. more
- The lanes are the recorded lanes. 2 of the 8 stations stop for a person and every one of them still does on this run.
- Every one of the 2 decisions is still taken by the people who hold them. The happy path is that each one is released first time, not that it is skipped.
- The owning team is held to a 83% autonomy ceiling and this run sits at 72% unattended, inside it.
- 2 systems are attributed across the stations, and 0 of them are connected. Nothing in this run has been watched executing against a real system of record.
1 of the 2 decisions can be handed over inside the 83% ceiling. more
- 06 Reviewer approve — held by Director, Data Science, 1.4 hrs a unit
- 08 Deploy — The last decision on the line. Something on every procedure ends with a person, so this one is never offered.
This is a run, not a forecast. more
Happy means nothing goes wrong: no station raises an exception, no gate sends the unit back, no rework loop fires and nothing sits in a queue between stations. more
The second clock is the allowance. more
A decision may be handed to the agent lane only if two things hold … more
The run reads the station rows only. more
What this run does not model: arrival patterns, contention between procedures for the same person, shift patterns, weekends, learning curves, and the cost of anything. more
Discovery Research
3 procedures · 17 steps · 2 decisions
Runs the line
Discovery Orchestrator
Accountable
Director, Discovery Research
Layers in play
strategy · observe · workflow
That team’s remit
Owns research thread work end to end inside Research & Development. more
Why the ceiling sits there
Literature synthesis and hypothesis generation are exactly where agents outperform — they read everything. more
The 3 rules every procedure in this team runs inside
Discovery Research Decision Envelope
under reviewRD1-POL-200 · Envelope · v2.7 · approved by Director, Discovery Research
Agents in this lane may act without a human when the research thread sits inside the stated value, risk and confidence envelope. more
Discovery Research Evidence Standard
activeRD1-POL-201 · Evidence · v2.3 · approved by Director, Discovery Research
Every autonomous decision writes inputs, the rule version applied, the model and prompt version, the output and a reversal path. more
Discovery Research Escalation Rule
activeRD1-POL-202 · Escalation · v1.4 · approved by Director, Discovery Research
Escalation is mandatory when confidence falls below the floor, when two options sit inside the confidence band, or when a break condition on a ratified position fires. more
| Code | Procedure | Steps | Decisions | Who does the work | Agent % | A week | |
|---|---|---|---|---|---|---|---|
| RD1A | Discovery SprintA | 7 | 1 | 86% | 40 | ||
| RD1B | Research Direction PositionJ | 6 | 1 | 83% | 5 | ||
| RD1C | Prior Work SweepT | 4 | 0 | 100% | 2,426 |
Innovation Portfolio
3 procedures · 16 steps · 4 decisions
Runs the line
Innovation Portfolio Orchestrator
Accountable
VP Innovation
Layers in play
strategy · observe · workflow
That team’s remit
Owns program work end to end inside Research & Development. more
Why the ceiling sits there
Agents assemble evidence and even recommend killing a program. more
The 2 rules every procedure in this team runs inside
Innovation Portfolio Decision Envelope
supersededRD2-POL-200 · Envelope · v5.1 · approved by VP Innovation
Agents in this lane may act without a human when the program sits inside the stated value, risk and confidence envelope. more
Innovation Portfolio Evidence Standard
activeRD2-POL-201 · Evidence · v2.2 · approved by VP Innovation
Every autonomous decision writes inputs, the rule version applied, the model and prompt version, the output and a reversal path. more
| Code | Procedure | Steps | Decisions | Who does the work | Agent % | A week | |
|---|---|---|---|---|---|---|---|
| RD2A | Stage Gate ReviewT | 6 | 2 | 67% | 5,106 | ||
| RD2B | Kill Signal EscalationT | 5 | 1 | 80% | 6,881 | ||
| RD2C | Portfolio Bets PositionJ | 5 | 1 | 80% | 5 |
Experiment Design
3 procedures · 16 steps · 4 decisions
Runs the line
Experiment Design Orchestrator
Accountable
Director, Research Methods
Layers in play
strategy · policy · observe · workflow · gbs
That team’s remit
Owns protocol work end to end inside Research & Development. more
Why the ceiling sits there
Protocol drafting and power analysis are agentic. more
The 2 rules every procedure in this team runs inside
Experiment Design Decision Envelope
activeRD3-POL-200 · Envelope · v1.6 · approved by Director, Research Methods
Agents in this lane may act without a human when the protocol sits inside the stated value, risk and confidence envelope. more
Experiment Design Evidence Standard
activeRD3-POL-201 · Evidence · v1.3 · approved by Director, Research Methods
Every autonomous decision writes inputs, the rule version applied, the model and prompt version, the output and a reversal path. more
| Code | Procedure | Steps | Decisions | Who does the work | Agent % | A week | |
|---|---|---|---|---|---|---|---|
| RD3A | Protocol DevelopmentT | 7 | 2 | 71% | 430 | ||
| RD3B | Protocol DeviationT | 5 | 1 | 80% | 7,292 | ||
| RD3C | Design Standard RefreshT | 4 | 1 | 75% | 2,913 |
Lab Operations
3 procedures · 15 steps · 1 decisions
Runs the line
Lab Operations Orchestrator
Accountable
Director, Laboratory Operations
Layers in play
strategy · policy · observe · workflow · gbs
That team’s remit
Owns lab request work end to end inside Research & Development. more
Why the ceiling sits there
Scheduling, replenishment and instrument monitoring are pure logistics and run touchless. more
The 4 rules every procedure in this team runs inside
Lab Operations Decision Envelope
activeRD4-POL-200 · Envelope · v5.4 · approved by Director, Laboratory Operations
Agents in this lane may act without a human when the lab request sits inside the stated value, risk and confidence envelope. more
Lab Operations Evidence Standard
activeRD4-POL-201 · Evidence · v2.1 · approved by Director, Laboratory Operations
Every autonomous decision writes inputs, the rule version applied, the model and prompt version, the output and a reversal path. more
Lab Operations Escalation Rule
activeRD4-POL-202 · Escalation · v1.4 · approved by Director, Laboratory Operations
Escalation is mandatory when confidence falls below the floor, when two options sit inside the confidence band, or when a break condition on a ratified position fires. more
Lab Operations Autonomy Ceiling
activeRD4-POL-203 · Ceiling · v2.0 · approved by Director, Laboratory Operations
The ceiling for this lane is 91 on the platform scale. more
| Code | Procedure | Steps | Decisions | Who does the work | Agent % | A week | |
|---|---|---|---|---|---|---|---|
| RD4A | Instrument BookingT | 5 | 0 | 100% | 5,522 | ||
| RD4B | Consumables ReplenishmentT | 4 | 0 | 100% | 1,317 | ||
| RD4C | Calibration CycleT | 6 | 1 | 83% | 2,882 |
Data Science & Modeling
3 procedures · 19 steps · 4 decisions
Runs the line
Modeling Orchestrator
Accountable
Director, Data Science
Layers in play
strategy · policy · observe · workflow · gbs
That team’s remit
Owns model work end to end inside Research & Development. more
Why the ceiling sits there
Training, validation and drift detection are fully agentic. more
The 4 rules every procedure in this team runs inside
Data Science & Modeling Decision Envelope
activeRD5-POL-200 · Envelope · v2.7 · approved by Director, Data Science
Agents in this lane may act without a human when the model sits inside the stated value, risk and confidence envelope. more
Data Science & Modeling Evidence Standard
supersededRD5-POL-201 · Evidence · v2.3 · approved by Director, Data Science
Every autonomous decision writes inputs, the rule version applied, the model and prompt version, the output and a reversal path. more
Data Science & Modeling Escalation Rule
activeRD5-POL-202 · Escalation · v5.0 · approved by Director, Data Science
Escalation is mandatory when confidence falls below the floor, when two options sit inside the confidence band, or when a break condition on a ratified position fires. more
Data Science & Modeling Autonomy Ceiling
activeRD5-POL-203 · Ceiling · v1.7 · approved by Director, Data Science
The ceiling for this lane is 83 on the platform scale. more
| Code | Procedure | Steps | Decisions | Who does the work | Agent % | A week | |
|---|---|---|---|---|---|---|---|
| RD5A | Model DevelopmentT | 8 | 2 | 75% | 2,326 | ||
| RD5B | Drift ResponseT | 6 | 1 | 83% | 7,707 | ||
| RD5C | Model RetirementT | 5 | 1 | 80% | 4,999 |
Prototype & Pilot
3 procedures · 17 steps · 4 decisions
Runs the line
Prototype Orchestrator
Accountable
Director, Product Development
Layers in play
strategy · observe · workflow · gbs
That team’s remit
Owns prototype work end to end inside Research & Development. more
Why the ceiling sits there
Test execution, iteration tracking and feedback synthesis are agentic. more
The 4 rules every procedure in this team runs inside
Prototype & Pilot Decision Envelope
activeRD6-POL-200 · Envelope · v1.5 · approved by Director, Product Development
Agents in this lane may act without a human when the prototype sits inside the stated value, risk and confidence envelope. more
Prototype & Pilot Evidence Standard
activeRD6-POL-201 · Evidence · v3.2 · approved by Director, Product Development
Every autonomous decision writes inputs, the rule version applied, the model and prompt version, the output and a reversal path. more
Prototype & Pilot Escalation Rule
activeRD6-POL-202 · Escalation · v4.9 · approved by Director, Product Development
Escalation is mandatory when confidence falls below the floor, when two options sit inside the confidence band, or when a break condition on a ratified position fires. more
Prototype & Pilot Autonomy Ceiling
activeRD6-POL-203 · Ceiling · v4.3 · approved by Director, Product Development
The ceiling for this lane is 76 on the platform scale. more
| Code | Procedure | Steps | Decisions | Who does the work | Agent % | A week | |
|---|---|---|---|---|---|---|---|
| RD6A | Prototype CycleT | 7 | 1 | 86% | 4,644 | ||
| RD6B | Pilot DeploymentT | 6 | 2 | 67% | 475 | ||
| RD6C | Scale Readiness AssessmentA | 4 | 1 | 75% | 12 |
IP Creation & Capture
3 procedures · 16 steps · 3 decisions
Runs the line
IP Capture Orchestrator
Accountable
Director, IP Capture
Layers in play
strategy · observe · workflow · gbs
That team’s remit
Owns disclosure work end to end inside Research & Development. more
Why the ceiling sits there
Detecting patentable work inside research output is something agents do continuously and people forget to do. more
The 4 rules every procedure in this team runs inside
IP Creation & Capture Decision Envelope
activeRD7-POL-200 · Envelope · v4.8 · approved by Director, IP Capture
Agents in this lane may act without a human when the disclosure sits inside the stated value, risk and confidence envelope. more
IP Creation & Capture Evidence Standard
activeRD7-POL-201 · Evidence · v4.4 · approved by Director, IP Capture
Every autonomous decision writes inputs, the rule version applied, the model and prompt version, the output and a reversal path. more
IP Creation & Capture Escalation Rule
under reviewRD7-POL-202 · Escalation · v5.8 · approved by Director, IP Capture
Escalation is mandatory when confidence falls below the floor, when two options sit inside the confidence band, or when a break condition on a ratified position fires. more
IP Creation & Capture Autonomy Ceiling
activeRD7-POL-203 · Ceiling · v2.8 · approved by Director, IP Capture
The ceiling for this lane is 80 on the platform scale. more
| Code | Procedure | Steps | Decisions | Who does the work | Agent % | A week | |
|---|---|---|---|---|---|---|---|
| RD7A | Invention CaptureT | 6 | 1 | 83% | 634 | ||
| RD7B | Trade Secret RegistrationT | 5 | 1 | 80% | 462 | ||
| RD7C | IP Strategy PositionJ | 5 | 1 | 80% | 4 |
Regulatory Science
3 procedures · 19 steps · 6 decisions
Runs the line
Regulatory Science Orchestrator
Accountable
VP Regulatory Science
Layers in play
strategy · policy · observe · workflow · gbs
That team’s remit
Owns submission work end to end inside Research & Development. more
Why the ceiling sits there
Requirement mapping and evidence assembly are agentic at scale. more
The 2 rules every procedure in this team runs inside
Regulatory Science Decision Envelope
activeRD8-POL-200 · Envelope · v3.3 · approved by VP Regulatory Science
Agents in this lane may act without a human when the submission sits inside the stated value, risk and confidence envelope. more
Regulatory Science Evidence Standard
activeRD8-POL-201 · Evidence · v1.4 · approved by VP Regulatory Science
Every autonomous decision writes inputs, the rule version applied, the model and prompt version, the output and a reversal path. more
| Code | Procedure | Steps | Decisions | Who does the work | Agent % | A week | |
|---|---|---|---|---|---|---|---|
| RD8A | Regulatory SubmissionT | 8 | 3 | 63% | 7,459 | ||
| RD8B | Standards Change ResponseT | 6 | 1 | 83% | 7,276 | ||
| RD8C | Compliance TestingT | 5 | 2 | 60% | 7,479 |
External Collaboration
3 procedures · 19 steps · 6 decisions
Runs the line
Collaboration Orchestrator
Accountable
Director, External Innovation
Layers in play
strategy · observe · workflow
That team’s remit
Owns collaboration work end to end inside Research & Development. more
Why the ceiling sits there
Scouting and milestone tracking are agentic. more
The 3 rules every procedure in this team runs inside
External Collaboration Decision Envelope
activeRD9-POL-200 · Envelope · v5.4 · approved by Director, External Innovation
Agents in this lane may act without a human when the collaboration sits inside the stated value, risk and confidence envelope. more
External Collaboration Evidence Standard
activeRD9-POL-201 · Evidence · v2.3 · approved by Director, External Innovation
Every autonomous decision writes inputs, the rule version applied, the model and prompt version, the output and a reversal path. more
External Collaboration Escalation Rule
activeRD9-POL-202 · Escalation · v2.6 · approved by Director, External Innovation
Escalation is mandatory when confidence falls below the floor, when two options sit inside the confidence band, or when a break condition on a ratified position fires. more
| Code | Procedure | Steps | Decisions | Who does the work | Agent % | A week | |
|---|---|---|---|---|---|---|---|
| RD9A | Research Collaboration SetupT | 7 | 2 | 71% | 3,828 | ||
| RD9B | Data Sharing RequestT | 6 | 2 | 67% | 2,022 | ||
| RD9C | Publication ClearanceT | 6 | 2 | 67% | 928 |
Technology Scouting
3 procedures · 16 steps · 3 decisions
Runs the line
Technology Scouting Orchestrator
Accountable
Director, Technology Scouting
Layers in play
strategy · observe · workflow
That team’s remit
Owns candidate work end to end inside Research & Development. more
Why the ceiling sits there
Landscape scanning and screening are continuous and touchless. more
The 4 rules every procedure in this team runs inside
Technology Scouting Decision Envelope
activeRD10-POL-200 · Envelope · v4.9 · approved by Director, Technology Scouting
Agents in this lane may act without a human when the candidate sits inside the stated value, risk and confidence envelope. more
Technology Scouting Evidence Standard
under reviewRD10-POL-201 · Evidence · v4.3 · approved by Director, Technology Scouting
Every autonomous decision writes inputs, the rule version applied, the model and prompt version, the output and a reversal path. more
Technology Scouting Escalation Rule
activeRD10-POL-202 · Escalation · v1.5 · approved by Director, Technology Scouting
Escalation is mandatory when confidence falls below the floor, when two options sit inside the confidence band, or when a break condition on a ratified position fires. more
Technology Scouting Autonomy Ceiling
activeRD10-POL-203 · Ceiling · v2.1 · approved by Director, Technology Scouting
The ceiling for this lane is 84 on the platform scale. more
| Code | Procedure | Steps | Decisions | Who does the work | Agent % | A week | |
|---|---|---|---|---|---|---|---|
| RD10A | Scouting CycleA | 6 | 1 | 83% | 25 | ||
| RD10B | Build-Buy-Partner PositionJ | 5 | 1 | 80% | 2 | ||
| RD10C | Technical DiligenceT | 5 | 1 | 80% | 3,097 |
Portfolio Analytics
3 procedures · 14 steps · 3 decisions
Runs the line
R&D Analytics Orchestrator
Accountable
Director, R&D Analytics
Layers in play
strategy · observe · workflow · gbs
That team’s remit
Owns analysis work end to end inside Research & Development. more
Why the ceiling sits there
Health scoring and spend attribution are fully agentic. more
The 2 rules every procedure in this team runs inside
Portfolio Analytics Decision Envelope
activeRD11-POL-200 · Envelope · v2.6 · approved by Director, R&D Analytics
Agents in this lane may act without a human when the analysis sits inside the stated value, risk and confidence envelope. more
Portfolio Analytics Evidence Standard
activeRD11-POL-201 · Evidence · v1.9 · approved by Director, R&D Analytics
Every autonomous decision writes inputs, the rule version applied, the model and prompt version, the output and a reversal path. more
| Code | Procedure | Steps | Decisions | Who does the work | Agent % | A week | |
|---|---|---|---|---|---|---|---|
| RD11A | Quarterly Portfolio ReadA | 6 | 1 | 83% | 64 | ||
| RD11B | Success Rate AnalysisA | 4 | 1 | 75% | 76 | ||
| RD11C | Benchmark StudyA | 4 | 1 | 75% | 33 |
Where this manual comes from
Nobody wrote these procedures. more
What it cannot claim
The twenty-four controls and eighteen exception classes in this platform are held at estate level … more
These procedures describe the modeled estate. more
Actions
What is waiting on a person
What a single function manual is for, and the three things worth checking the moment you open one.
governing rules bind the teams inside these functions
Every team carries between two and four rules covering evidence, envelope, escalation and ceiling. 395 are active. more
governance gates are declared across the function estate
A governance gate is a rule about the function, not about a single procedure: a threshold, a named approver and a clock. more
points separate the most and least agent-run function, Customer Service and Legal
Customer Service runs 90% of its steps without a person and Legal runs 76%. The gap is not a target and not a failing … more
Operations
What this desk is allowed to start
A surface that only reports is not operable. more
Trigger and bound
A function manual gathers every procedure the function owns, grouped by the team that runs it, with the orchestrator, the accountable human, the autonomy ceiling and the governing rules printed against each team. more
Live observability
What the record shows right now
How the manuals differ in size, by number of written procedures. more
Current distribution
211 procedures
Is policy and strategy coming to fruition
Whether the written intent is holding here
Holding on structure, not on assurance. more
Not holding on the record
Every manual is complete: 475 procedures with nothing unassigned, 154 teams each carrying an orchestrator and a named accountable human, and every step attributed to a lane. more