A deployment-value review
Discount the ROI to what transfers.
Gate each worker on its data.
Sequence so each deployment compounds the next.
Paste a logistics account brief. The tool discounts HappyRobot's published ROI multiples to what actually transfers for that account, with a named haircut on each one, gates every AI worker on the data foundation it needs, and sequences the rollout so each deployment lands on the data the last one produced.
Live engine
Pick a backtest, edit the brief, run the review.
Pick a backtest or paste your own brief, then Run review for a live extraction by gpt-5.4.
Method
What the model does, what the look-up adds, what the engine does
Four stages, left to right. The model only reads the published figure; the engine owns every haircut and every sequencing call.
01
Model · extract
Reads the messy brief and pulls out findings — data foundations and AI-worker deployments, each with the figure HappyRobot published and how to read it. It never discounts.
02
Look-up · prior rollouts
Checks each deployment against prior HappyRobot rollouts. Confirms a declared sequencing dependency, or surfaces one the brief never mentioned.
03
Engine · discount + gate
Pressure-tests each published figure against the account's call mix and stated volume, then gates every worker on the data it needs and sequences the rollout.
04
Output · plan
The discounted figures up top, then a three-phase rollout. Foundations first, deployable workers in parallel, gated ones after. Deterministic from the same findings.
Extraction, ROI transfer, and readiness stay separate. The model extracts; the look-up grounds it in prior rollouts; the engine discounts the figures and sequences the plan. That logic lives in engine.py, the records in corpus.py, both covered by evals.
Framework
Five ROI rules. One readiness gate.
HappyRobot publishes rosy multiples. The engine reads each one against this account and either holds it, haircuts it, or restates it — with a named, auditable reason. Then a separate gate sequences the rollout.
Then every worker is gated
- Deploy now. Its data foundation is in place. Ship.
- Fix foundation first. A sound deployment, sequenced behind the data it lands on.
- Re-scope. A useful play exists nearby; the named one isn't it.
- Don't pursue. Wrong-shaped, or value too ambiguous to defend.
Merit, ROI transfer, and readiness are three separate calls — so a figure can hold while its deployment still waits.
Future State
How this runs in production
The live demo implements the three highlighted steps; the rest is the automated pipeline around them — same boxed language, LIVE badges, a diamond for the human gate.
Future State
Account calls are transcribed and the ops stack and volumes are pulled in over connectors; everything lands in one account record, PII is redacted before any model call. Then the three live steps run — the model extracts foundations and published figures, the look-up grounds them against prior rollouts, and the deterministic engine discounts each figure and sequences the plan. That fans out into the phased rollout and a business case built only from the figures that held, a person approves before anything ships, and the shipped rollout feeds back into the library — so the look-up gets smarter each time.