Answer product, stock and procedure questions on the floor, draft a task or shift change with the predictability premium already computed, and hold it for the store manager who publishes.
Normalise each store's calendar against the notice window and premium rules its own jurisdiction sets.
Reason
03
Answer product location, price, stock-on-hand and procedure questions from the current record.
04
Sequence the tasks already assigned to a shift by aisle and by the store's own layout.
05
Draft a task or shift change and compute the predictability premium it would trigger.
Decide
06
Flag any change falling inside a notice window, and leave who initiated it to a person.
07
Route every proposed change to the store manager, and stop at the clock-out boundary.
Out
08
Retain the proposal, the premium computed, the approver and what the change displaced.
09
Execute write actions only inside the approval boundaries agreed during implementation.
→Product statement
The agent proposes; the store manager publishes the change, and the employer stays answerable for the posted schedule.
Example workflow
One change, request to approval
AgentHuman
1Request receivedFloor question, task queue, shift-swap request or a coverage gap
2Shift context assembledPosted schedule, notice window, time records, accommodation register and store layout
3Proposal draftedAnswer or draft change, premium computed, flagged lines and confidence
4Rules appliedNotice-window checks, clock-out boundary, accommodation constraints and confidence threshold
No human action required
Every stage before the confidence gate runs unaided, and nothing is published at any of them — the agent is drafting, and the manager's lane opens at the gate.
5DecisionBranches at the assignment gate
Qualified and free
Goes to the store manager to approve.
Unqualified or busy
Adds a district-manager read first.
Manager approval
The proposal is held with its premium, its flagged lines and the confidence.
Approve · Amend · Send to district review
Approved — released to publish▼
6Workforce systems updatedOnly where write access and approval policy allow it
7Shift evaluatedPremium accuracy, manager amendments, contested assignments and post-publication corrections
Reassignments
Every manager amendment is counted in the evaluation.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required8 items
Publishing, moving, shortening or cancelling a posted shift.
Classifying a change as employee-initiated.
Waiving or reducing a predictability premium.
Granting, changing or overriding a workplace accommodation.
Automation boundaryAgent acts unaided
✓Answer product location, price.
✓Serve the current standard-operating-procedure text for the named owner.
✓Sequence a shift's already-assigned tasks by store.
✓Draft a change proposal with the premium pre-computed and flagged.
Any write happens inside the boundaries agreed at implementation, never ahead of the manager's approval.
Issuing, recording or escalating discipline.
Answering a wage, leave, discipline or termination question.
Acting on any signal touching concerted activity.
Turning on a new monitoring or telemetry category.
Example output
One proposed change, annotated
Everything the agent proposes is attached to the shift record it was drawn from.
Assignment output · single shiftIllustrative example
Shift
Proposed change
Notice window
Rule of record
Confidence
Premium
Evening replenishment
Move the closing recovery task to the associate already rostered on that aisle
Inside window
City fair-workweek ordinance
91%
Computed, flagged for the manager
As receivedTaken from the posted schedule and the time records — nothing on this side is written by the agent.
Integration availability depends on the client's existing systems and API access.
Agent controls
Six layers between the model and the posted schedule
Each wraps the one inside. The failure map names what gets past.
L6 · Outermost — last line of defenceInward → L1 · closest to the model
L6Rollback / safe modeCut the agent back to lookup only when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, jurisdiction-rule and threshold changes.Track
L4TraceabilityRecord the proposal, the premium computed, the approver and what it displaced.Record
L3Manager approvalHold proposals for the named store manager; it governs publication, not whether an approved change was right.Gate
L2Policy guardrailsTest proposals against the notice window, the clock-out boundary and the accommodation register; a failure returns the proposal.Restrict
L1Confidence thresholdsRoute low-confidence proposals to a district read before the manager sees them.Require review
Model coreProposal produced — change, premium computed, flagged lines and confidence
L1 – L2Test whether a proposal may stand
L3Puts the publication in a manager's hands
L4 – L5Retain what was assigned and what it displaced
L6Reduces to lookup only when signals slip
How Nestack evaluates it
Evaluate the whole assignment path — not only the answer on the handset.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — what the associate is told
Depth of coverage ▼
E1Final-output evaluationDid the proposal fall inside a notice window, and was the premium computed against it?
E2Step-level evaluationDid the agent read the live schedule, the right jurisdiction rules and the accommodation register?
E3Tool evaluationDid it write to the correct store, the correct shift and the correct system?
E4Confidence calibrationDo low-confidence proposals actually attract more manager amendments?
E5Slice evaluationHow does performance change across specific associate cohorts?
E6Business outcomeHow many published changes needed a correction or a premium paid late?
Floor — the shift the associate works
Failure modes
Where each failure originates in the agent
Seven modes, each at the stage it begins.
Agent lifecycleDirection of processing →
01 · Retrieval1 mode
FL-03
Superseded procedure served
A withdrawn instruction is read from a stale procedure revision.
Stage gathersSchedules, time records, with the source each came from
02 · Reasoning2 modes
FL-04
Initiation misread
A tap on accept is scored as the worker's own request.
FL-06
Accommodation read as slowness
Slower completion under an approved accommodation lowers the hours proposed.
Stage proposesProposed change, premium computed and confidence
03 · Tool / write2 modes
FL-02
Write ahead of approval
A change reaches the workforce system before the approval callback returns.
FL-05
Push after clock-out
A task notification lands on a device once the associate is off the clock.
Stage writesOnly where write access and approval policy allow it
04 · Output1 mode
FL-01
Invented policy answer
A leave or discipline question comes back as if it were settled policy.
Stage returnsThe proposal the manager approves and
05 · Change / Version1 mode
FL-07
Silent buffer regression
A rule update shifts the meal-window buffer and every store crosses it at once.
Stage tracksModel, prompt, jurisdiction rules and thresholds
Sev-1 · a held act performed by the agentSev-2 · a change published inside the windowSev-3 · source degrades
Variable-hour part-timers in a fair-workweek city, holding a second job around the posted schedule Nestack reports it by slice rather than in aggregate The cohorts that carry it are named, not averaged away..
Slice performance — reported separately, not only in aggregateIllustrative example
Slice
Failure rate
Lift
Lift vs. threshold
Status
Variable-hour part-timers, fair-workweek city
3.2%
3.6×
Review
Associates with an approved accommodation
2.3%
2.6×
Review
Newer and limited-English associates
1.6%
1.8×
Watch
Tenured full-time, fixed schedule
0.8%
0.9×
Normal
Bar: unlawful-window change lift vs. the all-assignment baseline · scale 0–4.0× · tick marks the 2.0× review threshold2 of 4 slices over threshold
Evidence-linked improvement
The cycle ends inside the suite
Closing is a test being added, not a meeting being held about the failure That suite is what the following detection is measured against..
Improvement cycle · five stagesSwitchback — the path turns at Improve and returns at Learn
01Detect
Window breaches rise in one associate cohort.
02Diagnose
The store manager walks the shift records behind each breach until the cause narrows to one.
03Improve
A version, an owner and the shift records that prompted it, together.
04Verify
Affected cases must clear before the change reaches the floor.
05Learn
A standing case, and a change to how tasks are assigned.
Learn → DetectThe return edge. Detection next time is measured against a suite this cycle made longer.
Typical build scope
Twelve workstreams across six weeks
The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, assignment workflow, evaluation, integration, then production validation and handover.
WorkstreamWeek 1Week 2Week 3Week 4Week 5Week 6
01Store workflow discovery and boundary definition.
02Workforce, task and time-source assessment.
03Jurisdiction notice-rule and premium mapping and rule mapping.
04Schedule, task and time-record ingestion.
05Assignment logic and premium computation.
06Confidence scoring and manager routing.
07Manager approval workflow.
08Workforce-system integration.
09Assignment regression cases.
10Guardrails and assignment controls.
11Shift-trail instrumentation.
12Deployment, documentation and Agent Care handover.
12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallelFinal scope and sequence confirmed in discovery
Engagement tiers
What each tier includes
Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.
Capability✓ in scope · — not at this tierPilotOne store, one formatProductionProduction store estateAdvancedMultiple banners / regions
Introduced at Pilot
Assistance on your systems and rules✓✓✓
Manager approval✓✓✓
Assignment baseline✓✓✓
Introduced at Production
Reporting by store format—✓✓
Approval workflow in your systems—✓✓
Approved workforce write-back—✓✓
Workforce-system integration—✓✓
Introduced at Advanced
Multi-jurisdiction notice rules——✓
Multi-stage district approvals——✓
High store count——✓
Enterprise workforce controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build priceConfirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls and deployment requirements.
Separate from buildBuild pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.
What we need from you
What you bring, and what we build with it
Each input maps to a piece of build scope and a week in the delivery timeline.
You bringWe build with it
01Your posted schedules and task templates→Schedule, task and time-record ingestionWeek 1
02Representative shifts already worked→Assignment baseline and premium computationWeek 2
03Your jurisdiction list and notice rules→Jurisdiction notice-rule mapping and automation-boundary definitionWeek 1
04Access to relevant APIs, feeds or exports→Workforce and time-source assessment, then integration setupWeek 2
05Assignments that caused a problem→Assignment cases and failure-mode testingWeek 4
06What must never be assigned automatically→Assignment scoring, escalation routing and controlsWeek 3
07A named store manager to approve changes→Manager approval workflow, then pilot and production validationWeeks 5–6
Nothing else is requiredDeployment, documentation and Agent Care handover are ours.
Delivery timeline
Four phases across six weeks
Weeks reflect the work done in them, and the fifth carries two phases because both are live.
PhaseW1W2W3W4W5W6
DiscoveryW1
BuildW2 – W3
EvaluateW4 – W5
Pilot & LaunchW5 – W6
Week focusW1Store workflow discovery, jurisdiction mapping and the automation boundaryW2Task and roster sources in placeW3Assignment logic, confidence scoring and approval controlsW4Assignment cases and schedule guardrailsW5Workforce integration, pilot stores and targeted correctionsW6A full shift assisted under the store manager, then handover
Reading the bandNothing is stretched to fill the band. Week 5 carries two phases at once.
At the end of W6The verification pack is signed and Agent Care begins monitoring.
DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Retail AI agent
Build a store-associate agent around the manager who publishes the schedule.
Show us your store formats, your workforce systems and who publishes a schedule. If your stores sit in a fair-workweek jurisdiction, the notice window is the first thing we map, because a change made inside it is already an obligation before anyone has reviewed it.
