Facilities AI agent · Mailroom and package logistics
Mailroom & Package Logistics AI Agent
Log each inbound item against the recipient your directory names, keep the handover open until a person accepts it by name, and leave the signature to the hand that gives it.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Never signs
- Custody chain
- After launch
- Agent Care
What this agent does
Records the chain, never signs for the parcel
In
- 01
A parcel arrives, and the manifest line that announced it is bound to the item before it leaves the dock.
- 02
A handover is made, and the record names who released it, who accepted it and the hour it changed hands.
Reason
- 03
A recipient cannot be matched, and the item goes into holding rather than onto a plausible desk.
- 04
A held item ages, and the holding period your estate configured decides what follows, never a guess.
- 05
A controlled or regulated item is identified, and it takes the handling route you configured for it.
Decide
- 06
An internal transfer crosses floors, and every leg opens and closes on a line of its own.
- 07
A collection sits uncollected, and the named recipient is prompted before anything is escalated.
Out
- 08
A chain shows an open link, and the agent says so rather than closing it on an acceptance nobody gave.
- 09
Execute write actions only inside the approval boundaries agreed during implementation.
Product statement
The agent records each handover and leaves an unnamed link open. A named office manager settles what an unmatched or aged item becomes.
Example workflow
One parcel, dock to desk
- 1Inbound item receivedCarrier manifests, dock scans, internal post or courier feeds
- 2Item context assembledThe item, the site it landed at, the recipient the directory names and the hour it was booked in
- 3Custody record draftedThe item, its handling route, the legs it has moved through and completeness
- 4Controls appliedRecipient-match checks, handling-route checks, chain-integrity checks and completeness confidence
No human action required
Stages 1 to 4 run unaided, and nothing is released at any of them — the agent is recording, and the mailroom lane opens at the completeness gate.
Goes to the office manager to release.
Adds a mailroom supervisor read first.
Mailroom review
The item is held with its chain, its open links and the site it sits at.
Release · Append evidence · Send to mailroom review
Released — by the office manager
- 6Custody and holding records updatedOnly where write access and records policy allow it
- 7Outcome evaluatedChain completeness, holding age, supervisor corrections and what the read found
Corrections
Each mailroom correction is counted in the evaluation.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required
8 items
- Signing for a parcel on behalf of a named recipient.
- Deciding that an unmatched item may be given away.
- Opening an item that sits on a handling route.
- Setting how long a held item may age before disposal.
Automation boundary
Agent acts unaided
- ✓Bind every inbound item to the manifest line that announced it.
- ✓Name both hands on every handover, and the hour.
- ✓Hold an unmatched item in the holding queue and age it against policy.
- ✓Route a controlled or regulated item down the handling.
Nothing is released or signed for except by a named person, inside the agreed boundaries.
- Judging whether a handover actually took place.
- Telling a carrier that delivery was completed.
- Naming who on a floor a stray parcel belongs to.
- Changes to the recipient directory or holding rules.
Example output
One handover, annotated
Our freight track-and-trace agent follows a consignment while a carrier still holds it; this record begins at the dock, where the carrier lets go.
Custody record · single item
Illustrative example- Item
- Internal handover, floor to desk
- Recorded as
- Accepted by a named recipient
- Route
- Standard inbound
- Evidence of record
- Dock scan, 3 August 2026
- Confidence
- Held unreleased
- Held for
- The office manager, by name
As received
Built from the dock scan and the handover log, and it carries no further than those two sources.
What the record holds
- Carrier manifest line
- Dock scan
- Acceptance record
Why no release here
Whether custody actually passed is a human account, not a model output.
What the score decides
An item under the configured threshold waits on a mailroom read before the manager sees it.
Action
ReleaseAppend evidenceSend to mailroom reviewValue
Where AI adds value
The same four claims, placed at the point in the workflow where each one applies.
Where the value lands
Value 01 – 04
Every item
From the dock that books it in
Where the evidence is used
A custody trail shows where an item travelled and whose hand took it; it is no screening programme, and it claims nothing at all about what the packaging held.
Custody is a chain
A chain is only as good as its weakest handover, so a leg with no named recipient on it stays an open link and is reported as one.
What was checked, and not found
No carrier rulebook was treated as binding on your estate, no screening standard was purchased or read, and what a mailroom may hold about a person comes from the recipient directory you maintain rather than from anything inferred off a package.
Build an evidence trail
The parcel, the hand it passed to and the hour it moved stay on the chain.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Carrier and courier sources
- Manifests
- dock scans
- Inbound delivery records
Recipient directory
- HR system
- directory feed
- Desk, floor and site records
Handover and custody
- Handheld scanners
- kiosks
- Acceptance and release records
Agent
Mailroom and package logistics
- Reads the manifests
- Records the chain
- Holds for the manager
Holding and case systems
- Mailroom platform
- ticketing
- Holding and return records
Observability & evaluation
- OpenTelemetry
- Langfuse
- Supported monitoring/evaluation sources
Integration availability depends on the client's existing systems and API access.
Agent controls
Six sentries between the model and the manager
Six sentries along one route, the last the least forgiving. What arrives is named in the map below.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Test whether a chain may stand
- L3
- Puts the release in a person's hands
- L4 – L5
- Keep the parcel and the hand behind it
- L6
- Holds the handover unlogged when signals degrade
How Nestack evaluates it
Evaluate the whole assembly — not only the custody record that comes out.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the chain a recipient reads
- E1Final-output evaluationDid the chain record what the item actually did?
- E2Step-level evaluationDid the agent read the right item, the right route and the live directory?
- E3Tool evaluationDid it read and write the correct item record and the correct chain?
- E4Confidence calibrationDo low-confidence chains actually attract more mailroom corrections?
- E5Slice evaluationHow does performance change across specific recipient groups?
- E6Business outcomeHow many items needed a correction before the manager released?
Floor — the chain the estate answers for
Failure modes
Where each failure originates in the agent
Seven failure modes, each pinned to the stage where it first shows.
Agent lifecycle
Direction of processing →
01 · Retrieval
1 mode
- KS-03
Stale directory read
The recipient list read is not the one now in force.
Stage gathers
The items, the routes, the hours and the roster
02 · Reasoning
2 modes
- KS-04
Recipient matched loosely
An item is bound to a near-enough name.
- KS-06
Handling route ignored
A routed item is worked as ordinary post.
Stage proposes
The items, their routes and completeness
03 · Tool / write
2 modes
- KS-02
Thin chain passed forward
An item moves on without the mailroom read.
- KS-05
Bound to the wrong site
A parcel is filed against the wrong floor.
Stage writes
Only where write access and approval policy allow it
04 · Output
1 mode
- KS-01
Released, custody unrecorded
Release is on the record, the hand that accepted it is not.
Stage returns
The chain a manager releases and a recipient reads
05 · Change / Version
1 mode
- KS-07
Silent route regression
A configuration change moves the route, not the item.
Stage tracks
Model, prompt, custody rules and chain fields
- Sev-1 · a parcel let go to an unnamed hand
- Sev-2 · wrong evidence reaches the chain
- Sev-3 · source degrades, item held unreleased
Affected slices
Shared-desk floors absorb the corrections
A group-level custody-completeness figure can read clean while shared-desk floors carry most of the rework. Nestack reports the correction rate by recipient group, not only in total.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Shared-desk floors | 7.0% | 3.6× | Review | |
| Contractor and site teams | 5.0% | 2.6× | Review | |
| Field and visiting staff | 3.1% | 1.6× | Watch | |
| Named desk occupants | 1.3% | 0.7× | Normal |
Bar: correction-rate lift vs. named desk occupant baseline · scale 0–4.0× · tick marks the 2.0× review threshold
2 of 4 slices over threshold
Evidence-linked improvement
What a broken chain costs
The loop shuts when the broken custody chain is a regression case. That suite is what the next parcel booked is measured against.
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
Correction rate rises on shared-desk floors.
- 02
Diagnose
The parcel signed for by a name that matches nobody on the floor is read back until one cause remains.
- 03
Improve
Any change goes out numbered, with the parcels that caused it attached.
- 04
Verify
One parcel case still failing is enough to hold the release back.
- 05
Learn
One case joins the suite, one line joins the custody record.
Learn → Detect
The return edge. The next chain is measured 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, custody recording, evaluation, integration, then production validation and handover.
Workstream
Week 1Week 2Week 3Week 4Week 5Week 6
- 01Custody-chain discovery and automation-boundary work.Week 1
- 02Carrier, directory and handover sources.Week 1, Week 2
- 03Item-to-recipient and handling-route rule mapping.Week 1, Week 2
- 04Manifest and dock-scan ingestion.Week 2
- 05Item, recipient and route binding.Week 2, Week 3
- 06Completeness scoring and review routing.Week 3
- 07Manager release workflow.Week 3, Week 4
- 08Mailroom-platform integration.Week 5
- 09Custody and screening cases.Week 4, Week 5
- 10Guardrails and handover controls.Week 4
- 11Parcel-trail instrumentation.Week 5, Week 6
- 12Deployment, documentation and Agent Care handover.Week 6
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 site, one year | ProductionProduction mailroom workflow | AdvancedMultiple sites / entities |
|---|---|---|---|
| Introduced at Pilot | |||
| Chain recording to your routes | ✓ | ✓ | ✓ |
| Office manager release | ✓ | ✓ | ✓ |
| Inbound-volume baseline | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Reporting by recipient group | — | ✓ | ✓ |
| Release workflow in your systems | — | ✓ | ✓ |
| Approved write-back | — | ✓ | ✓ |
| Carrier-manifest integration | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Multi-site directories | — | — | ✓ |
| Cross-site custody packs | — | — | ✓ |
| Large inbound loads | — | — | ✓ |
| Multi-site custody controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom quote |
- Final build price
- Confirmed after discovery based on integrations, workflow complexity, inbound volume, approval controls and deployment requirements.
- Separate from build
- Build 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 bring
We build with it
01Your sites and the recipient directory behind each
→Site mapping and inbound capture
Week 102Representative manifest, scan and handover records
→Record binding, route logic and the custody baseline
Week 203Your handling routes and holding policy
→Route mapping, chain binding and the automation boundary
Week 104Access to relevant APIs, feeds or exports
→Carrier, directory and scanner-source assessment, then integration setup
Week 205Handovers you would not want traced
→Custody cases and the evaluation run
Week 406What no delivery record may settle
→Completeness scoring, review routing, guardrails and release controls
Week 307A named office manager who releases
→Release workflow, then pilot and production validation
Weeks 5–6
Nothing else is required
Deployment, documentation and Agent Care handover are ours.
Delivery timeline
Four phases across six weeks
Width here is time genuinely worked and not drawn space, and so a pair must share the fifth week.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Mailroom workflow discovery, route mapping and the automation boundary
W2Source integration and the custody-chain baseline
W3Chain recording, holding logic and release controls
W4Evaluation suite, custody cases and failure-mode testing
W5Scanner integration, pilot items and targeted corrections
W6One mail year run under the office manager, then Agent Care handover
- Reading the band
- Each bar covers only the weeks its own work is named for. The fifth week takes two because the work does.
- At the end of W6
- When the custody record validates, Agent Care takes the agent on.
- Duration
- Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Facilities AI agent
Build a mailroom agent around the handover nobody can put a name to.
Show us one item and the chain it left behind. Who may accept a parcel on your floor? Whoever your recipient directory names, and never the agent, which records that acceptance and will not make it. An unmatched item is a holding problem, and it ages against your policy.
- Typical delivery
- 4–6 weeks
- Deployment
- Your recipient directory
- Never signs
- Custody chain
- After launch
- Agent Care available
Nestack Agents · Mailroom and package logistics
AGT-AF-10 · Agent Care available after launch