Nestack Agent Care
Industries / Operations / Order fulfilment agent

Operations AI agent · Order fulfilment

Order Fulfilment & Inventory AI Agent

Weigh the promise on an order against what stock and the pick line can actually do, then draft the notice the rule requires and hold it for the fulfilment lead, who sends it.

4–6 weeksTypical delivery
Your stackDeployment
Shipment clockNamed lead
Agent CareAfter launch

What this agent does

Times the promise, never moves it

In
01

An order completes when payment and every element needed to ship are there, and the clock starts then.

02

A shipping time stated in the solicitation displaces the thirty-day fallback, a shorter one included.

Reason
03

An order ships when the goods are placed in the possession of the carrier, which is not delivery.

04

A promise moves, and 435.2(b) wants the option offered clearly and without waiting to be asked.

05

A revised date thirty days or less past the applicable time is one silence is treated as consent to.

Decide
06

A longer or open-ended delay cancels the order instead, unless consent arrives inside thirty days.

07

A card already on file is payment, not an application for credit, so no fifty-day clock opens.

Out
08

An order shipped late with no notice offered is already cancelled, and the shipment does not cure that.

09

Execute write actions only inside the approval boundaries agreed during implementation.

Product statement

Timing, reading stock and drafting the notice belong to the agent. Sending it, cancelling an order and refunding belong to a named fulfilment lead.

Example workflow

One order, promise to notice

AgentHuman
1Order receivedWeb checkout, marketplace feed, telephone order or an order-management record
2Clock set from the promiseThe time the channel stated, the day the order was completed and the date shipment is owed
3Stock and pick line readOn-hand, allocated, inbound and what the pick line can clear before that date
4Controls appliedPromise checks, completeness checks, carrier-tender checks and shipment confidence
No human action required

Stages 1 to 4 run unaided, and no notice leaves at any of them — the agent is timing, and the lead lane opens at the shipment gate.

5DecisionSplits at the shipment gate
Shipment on time

Goes to the named fulfilment lead.

Anything late

Adds a customer operations read first.

Lead review

The order is held with its stated time, its revised date and the notice drafted for it.

Release notice · Amend notice · Send to operations review
Released — by the fulfilment lead
6Order and notice records updatedOnly where write access and records policy allow it
7Outcome evaluatedPromise accuracy, notice timing, lead corrections and what review found
Corrections

Each fulfilment-lead correction is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required8 items
Deciding that a promise on an order may move.
Sending the delay notice a buyer receives.
Releasing a prompt refund inside seven working days.
Setting the shipping time a channel advertises.
Automation boundaryAgent acts unaided
Check what each order still needs before shipment.
Draft the delay notice and the definite revised date it must carry.
Time the applicable shipping clock from the order that started it.
Flag each order whose promised time is at risk.
No notice leaves except by a named fulfilment lead, inside the agreed boundaries.
Judging whether an order was properly completed.
Committing stock that no count has confirmed.
Telling a regulator what an order record means.
Changes to promise rules, channels or thresholds.

Example output

One order promise, annotated

This serves an operations team whose records have to show systems and procedures assuring compliance in the ordinary course; below is one order exactly as the agent leaves it.

Order record · single promiseIllustrative example
Order
Recorded as
Applicable time
Evidence of record
Confidence
Held for
Marketplace order, promised date
Late against the stated time
Two days, as advertised
Carrier tender, 6 May 2026
Held unsent
The named fulfilment lead
As receivedTaken from the order record and the carrier tender on file, and it asserts nothing beyond them.
What the record holds Order completion Carrier tender Stock allocation
Why no notice hereMoving a promise on an order is a judgement the fulfilment lead makes.
ActionRelease noticeAmend noticeSend to operations review
What the score decidesBelow the configured threshold the order gets an operations read before the lead sees it.

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value landsValue 01 – 04
Every orderFrom the channel that promised it
03Evidence

Where the promise is tested

Warehouse orchestration releases and sequences work inside the four walls; this is the promise made to the customer, the clock it started and the notice owed when that promise moves.

01Approved path

Shipping is not delivering

The clock in 435.1(e) turns on the act by which merchandise is physically placed in the possession of the carrier, so the EU and UK delivery rules are answering a different question.

02Human review

What was checked, and not found

Checked in the current text: Part 435 carries no signature, no certification, no filing and no prescribed retention period, and nothing in it regulates inventory accuracy for an ordinary company — no count cadence, no stock-record format, no tolerance.

04Build an evidence trail

The order, the notice sent about it and the date the promise moved stay together.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

Order managementNetSuite · Manhattan Active
Order records and the dates promised
Commerce channelsShopify · BigCommerce · marketplaces
Product-page and checkout shipping claims
Warehouse and inventoryBlue Yonder · Körber · SAP EWM
On-hand, allocated and pick-line status

Agent

Order fulfilment and inventory

Reads the promise
Times the clock
Holds for the lead

Carriers and dispatchTender records · tracking events
When the carrier took possession
Observability & evaluationOpenTelemetry · Langfuse
Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six scales between the model and the buyer

Six scales in a line, the last the truest. What weighs through is set out in the map below.

L6 · Outermost — last line of defenceInward → L1 · closest to the model
L6Rollback / safe modeNarrow the agent to listing orders at risk when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt and promise rules, and note the version each order was timed under.Track
L4TraceabilityRecord each order, the time it was promised at, the clock it ran on and every notice drafted.Record
L3Lead releaseHold the notice for a named lead; the hold governs release, not whether the revised date is right.Gate
L2Promise guardrailsTest each order against the time its channel stated, and refuse a clock with no completed order under it.Restrict
L1Confidence thresholdsRoute a thin shipment forecast to an operations read before the notice reaches the lead.Require review
Model coreOrder timed — the stated promise, the shipment forecast, the revised date and what is missing
L1 – L2Test whether a promise may stand
L3Leaves the notice to a named lead
L4 – L5Keep the promise and the notice behind it
L6Sends the notice unshortened when signals degrade

How Nestack evaluates it

Evaluate the whole promise — not only the notice that goes out.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the shipping time a buyer reads at checkout
Depth of coverage ▼
E1Final-output evaluationDid the order record the time its channel actually stated?
E2Step-level evaluationDid the agent read the live stock position, the right channel and a completed order?
E3Tool evaluationDid it read and write the correct order and the correct notice?
E4Confidence calibrationDo low-confidence shipment forecasts actually attract more lead corrections?
E5Slice evaluationHow does performance change across specific order channels?
E6Business outcomeHow many orders needed a correction before the lead released the notice?
Floor — the stock a promise rests on

Failure modes

Where each failure originates in the agent

Seven failure modes, each shown at the stage where it surfaces.

Agent lifecycleDirection of processing →
01 · Retrieval1 mode
ND-03

Stale stock position read

The on-hand figure read is not the live one.

Stage gathersThe orders, the promises, the stock and the dates
02 · Reasoning2 modes
ND-04

Delivery read as shipment

A date is timed to arrival, not to tender.

ND-06

Notice drafted without a date

A delay is written with no revised date on it.

Stage proposesThe promise, the forecast, the date and the gaps
03 · Tool / write2 modes
ND-02

Thin forecast passed forward

An order moves on without the operations read.

ND-05

Notice bound to wrong order

The notice is filed against another order.

Stage writesOnly where write access and approval policy allow it
04 · Output1 mode
ND-01

Shipped late, notice unrecorded

The record shows a shipment but not the notice.

Stage returnsThe promise a buyer reads and a lead answers for
05 · Change / Version1 mode
ND-07

Silent promise drift

A channel badge changes and the clock keeps the old time.

Stage tracksModel, prompt, promise rules and channel badges
Sev-1 · a promise moved with no notice sent Sev-2 · a late shipment reaches the buyer Sev-3 · stock degrades, order held for review

Affected slices

Marketplace orders absorb the corrections

A channel-level promise-accuracy figure can read clean while marketplace promised-date orders carry most of the corrections. Nestack reports the correction rate by channel, not only in total.

Slice performance — reported separately, not only in aggregateIllustrative example
SliceFailure rateLift Lift vs. thresholdStatus
Marketplace promised-date orders10.9%3.7× Review
Pre-orders and backorders7.8%2.6× Review
Split shipments across sites4.8%1.6× Watch
Standard in-stock orders2.0%0.7× Normal
Bar: correction-rate lift vs. standard-order baseline · scale 0–4.0× · tick marks the 2.0× review threshold 2 of 4 slices over threshold

Evidence-linked improvement

What a moved promise costs

A loop shuts when the promise moved without notice is a standing case. That suite is what the next order released is measured against.

Improvement cycle · five stagesSwitchback — the path turns at Improve and returns at Learn
01Detect

Correction rate rises on marketplace promised-date orders.

02Diagnose

The two-day badge on the product page for an item that had not been picked in a week is worked backwards until one cause is left standing.

03Improve

The change leaves numbered, and the orders that caused it ride with it.

04Verify

One order case still red is enough to hold the release back.

05Learn

It stays a standing test, and the notice rules are amended alongside it.

Learn → DetectThe return edge. The next order released runs against a suite one case 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, promise timing, evaluation, integration, then production validation and handover.

Workstream Week 1Week 2Week 3Week 4Week 5Week 6
01Promise-boundary and automation-boundary work.
02Order, stock and carrier sources.
03Channel-promise and shipment-clock derivation.
04Order and stock ingestion.
05Promise, clock and order binding.
06Shipment scoring and review routing.
07Lead notice-release workflow.
08Order-system integration.
09Promise and notice cases.
10Guardrails and cancellation controls.
11Order-trail instrumentation.
12Deployment, documentation and Agent Care handover.
12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel Final 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 tier PilotOne channel, one site ProductionProduction fulfilment workflow AdvancedMultiple channels / sites
Introduced at Pilot
Promise timing to your channels
Named lead notice release
Order-promise baseline
Introduced at Production
Reporting by order channel
Lead review workflow in your systems
Approved write-back
Order-management integration
Introduced at Advanced
Multi-channel promise sets
Cross-site shipment packs
Large order volumes
Multi-channel promise controls
Build price From $5,000 From $8,000 Custom quote
Final build priceConfirmed after discovery based on integrations, workflow complexity, order 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 live channels and the shipping time each one states Promise capture and clock derivationWeek 1
02Representative orders, stock positions and carrier tenders Stock binding, clock logic and the promise baselineWeek 2
03Your notice templates and the leads they name Promise mapping, clock derivation and the automation boundaryWeek 1
04Access to relevant APIs, feeds or exports Order, stock and carrier source assessment, then integration setupWeek 2
05Promises you would not want quoted back Notice cases and failure-mode testingWeek 4
06What no shipment notice may waive Shipment scoring, review routing, guardrails and release controlsWeek 3
07A named fulfilment lead who releases notices Release to the fulfilment lead, then pilot and production validationWeeks 5–6
Nothing else is required Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

This is a schedule rather than an illustration, which is why two phases genuinely share week five.

Phase W1W2W3W4W5W6
Discovery W1
Build W2 – W3
Evaluate W4 – W5
Pilot & Launch W5 – W6
Week focus W1Promise discovery, clock derivation and the automation boundary W2Source integration and the order-promise baseline W3Stock binding, shipment logic and release controls W4Evaluation suite, notice cases and failure-mode testing W5Order-system integration, pilot channels and targeted corrections W6One peak season run under the fulfilment lead, then Agent Care handover
Reading the bandEach bar covers only the weeks its own work is named for, and week five is shared by design.
At the end of W6Once the shipment record validates, Agent Care picks the agent up.
DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Operations AI agent

Build an order fulfilment agent around the promise your product page makes before the pick line has seen it.

Show us one channel, the shipping time it states and the last order that missed it. What does the record show when a promise moves — a notice offered without prior demand, or nothing, which is the absence 435.2(d) turns into a presumption against the seller?

Nestack Agents · Order fulfilmentAGT-OP-12 · Agent Care available after launch