Sub-function
Discovery Research
Owns research thread work end to end inside Research & Development. Runs 3 lines across 17 stations, drains the repeatable share into no shared tower, and holds 2 human gates.
Estate rating
BB
67 / 100
Literature synthesis and hypothesis generation are exactly where agents outperform — they read everything. A scientist validates before anything becomes a program.
Agents
7
Lines
3
Units / wk
2,471
Timetable vs actual
22.6h → 25.3h
Cost / unit
$17
was $50
Run-rate saving
$4.2M
annualized at this volume
The 7 agents running this sub-function
Grouped by what they are for, not by where they sit. The agent org is process-shaped.
Discovery Orchestrator
Owns the discovery research lane end to end. Sequences the other agents, holds the timetable, and decides what surfaces to a human.
ceiling A3 · escalates to Escalates to Director, Discovery Research when the unit falls outside the decision envelope or confidence drops below the floor.
Literature Synthesis Agent
Executes the discovery research step it owns, posts its evidence, and hands the unit to the next stage.
ceiling A3 · escalates to Escalates to Director, Discovery Research when the unit falls outside the decision envelope or confidence drops below the floor.
Prior Work Agent
Executes the discovery research step it owns, posts its evidence, and hands the unit to the next stage.
ceiling A3 · escalates to Escalates to Director, Discovery Research when the unit falls outside the decision envelope or confidence drops below the floor.
Feasibility Screening Agent
Executes the discovery research step it owns, posts its evidence, and hands the unit to the next stage.
ceiling A3 · escalates to Escalates to Director, Discovery Research when the unit falls outside the decision envelope or confidence drops below the floor.
Hypothesis Generation Agent
Authors positions on discovery research. Frames the question, models the options, and writes a recommendation a human can ratify or reject.
ceiling A3 · escalates to Escalates when two options sit inside the confidence band and the evidence cannot separate them.
Research Direction Agent
Authors positions on discovery research. Frames the question, models the options, and writes a recommendation a human can ratify or reject.
ceiling A3 · escalates to Escalates when two options sit inside the confidence band and the evidence cannot separate them.
Discovery Research Challenger Agent
Adversarial reviewer for discovery research. Argues the opposite case on every position and flags where the evidence does not carry the claim.
ceiling A3 · escalates to Escalates when a ratified position is still running against a break condition it flagged.
Estate rating breakdown
Seven control dimensions. Judgment readiness runs against autonomy on purpose.
People
What the humans stopped doing, and what they do now.
Redeployed into exception judgment and supplier relationship work.
Lines running in this sub-function
Each line is a workflow. Click a line to walk it station by station.
Strategy positions this sub-function holds
An agent that only executes is a robot. These are the calls it made, the argument against each one, and what would prove it wrong.
Research Direction Position
Supersededv2What operating shape carries twice the current research thread volume without adding headcount, and what breaks first if we are wrong?
Hold the ceiling at A3 for discovery research. Literature synthesis and hypothesis generation are exactly where agents outperform — they read everything. A scientist validates before anything becomes a program. Raise it only after two consecutive quarters where the override rate stays under five percent and every override has a written cause.
confidence
79%
The challenge — Discovery Research Challenger Agent
Discovery Research Challenger Agent argues the recommendation leans on 3 quarters of data from a period with no volume shock. The confidence band overlaps the alternative, and the position does not say what it would take to be wrong. Recorded as a dissent, not a block.
Break conditions
- ·Override rate rises above 6 percent for two consecutive months
- ·Cost per research thread stops falling while volume keeps rising
- ·A control failure in this lane reaches a customer or a regulator
Evidence gaps
- ·No external comparator on a like-for-like unit definition
- ·Exception cases under $46k are sampled, not fully measured
| Alternative considered | Cost | Risk | Verdict |
|---|---|---|---|
| Hold the current position | $231k run rate | Known and priced. Cedes ground if comparators move faster. | live |
| Raise the ceiling one level now | $1435k to build controls | Override rate is 8 percent; raising the ceiling before that settles imports the error into production. | rejected on evidence |
| Move the exception tail to the spine | $885k transition | Loses local context. Rework risk on the cases that are hardest to recover. | under review |
Discovery Research Operating Position
Rejectedv1Is the current cost per research thread defensible against external comparators once inference spend is counted in full?
Keep the repeatable research thread volume in the spine and pull the exception tail back into the function. The spine is cheaper per unit; the tail is where the context that prevents rework actually sits.
confidence
66%
The challenge — Discovery Research Challenger Agent
Discovery Research Challenger Agent argues the recommendation leans on 4 quarters of data from a period with no volume shock. The confidence band overlaps the alternative, and the position does not say what it would take to be wrong. Recorded as a dissent, not a block.
Break conditions
- ·Override rate rises above 6 percent for two consecutive months
- ·Cost per research thread stops falling while volume keeps rising
- ·A comparator publishes a materially lower unit cost on a like-for-like basis
Evidence gaps
- ·No external comparator on a like-for-like unit definition
- ·Exception cases under $50k are sampled, not fully measured
| Alternative considered | Cost | Risk | Verdict |
|---|---|---|---|
| Hold the current position | $640k run rate | Known and priced. Cedes ground if comparators move faster. | live |
| Raise the ceiling one level now | $357k to build controls | Override rate is 13 percent; raising the ceiling before that settles imports the error into production. | rejected on evidence |
| Move the exception tail to the spine | $684k transition | Loses local context. Rework risk on the cases that are hardest to recover. | under review |
Policy register
The rules the agents above are bound by. Version, owner, approver, and the agents each rule constrains.
| Ref | Rule | Scope | Owner agent | Approved by | Version | Status |
|---|---|---|---|---|---|---|
| RD1-POL-200 | Discovery Research Decision Envelope Agents in this lane may act without a human when the research thread sits inside the stated value, risk and confidence envelope. Outside it, the unit holds at a gate with a named approver and a running clock. binds 3 agents · Global · EU | envelope | Discovery Orchestrator | Director, Discovery Research | v2.7 | under review |
| RD1-POL-201 | Discovery Research Evidence Standard Every autonomous decision writes inputs, the rule version applied, the model and prompt version, the output and a reversal path. An action with no evidence record is treated as a control failure, not a fast decision. binds 3 agents · LATAM · US | evidence | Discovery Orchestrator | Director, Discovery Research | v2.3 | Active |
| RD1-POL-202 | Discovery Research Escalation Rule 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. Director, Discovery Research owns the response clock. binds 3 agents · US · Canada | escalation | Discovery Orchestrator | Director, Discovery Research | v1.4 | Active |