Upsell & Ancillary-Offer AI Agent

What this agent does

Offers proposed, nothing added

Where an offer starts, ingest the booking, the rate or fare rules and the published offer set.

When source fields differ, normalise them so each offer carries the record it was drawn from.

Reason

Where a guest is eligible, draft an offer from the published set, with its price and its terms.

When the offer is shown, the amount, what it covers and whether it refunds are shown with it.

Where the journey includes air, the offer is marked as one that may not appear pre-selected.

Decide

When a charge cannot reasonably be avoided, flag it as mandatory, whatever the label on it says.

Where a saving or a was-price is implied, refuse it unless a real prior price substantiates it.

Out

When a guest accepts, record that acceptance per item, with the words and the price they saw.

Where a write is needed, execute only inside the approval boundaries agreed during implementation.

→ Product statement

The agent proposes an offer; the guest's own yes is what adds anything, and the operator stays the one charging.

Example workflow

One offer, booking to acceptance

AgentHuman

1Booking receivedReservation, PNR, folio or a commerce session in progress

2Eligibility assembledRate or fare rules, inventory, the published offer set and what the guest already holds

3Offer draftedOffer, price shown, terms, confidence

4Controls appliedOpt-in checks, avoidability checks, reference-price substantiation and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing is shown or charged at any of them — the agent is drafting, and the owner's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold

High confidence

Goes to the offer owner to release.

Low confidence

Adds a pricing-lead read first.

Owner releases

The offer is held with its price, its terms, its flagged lines and the confidence.

Release · Adjust · Send to pricing review

Released — the guest chooses▼

6Commerce systems updatedOnly where write access and approval policy allow it

7Outcome evaluatedAdjustment distance, flagged-offer outcomes, opt-in quality and post-sale refunds

Adjustments

Every owner adjustment is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required8 items

Adding a charge to a booking the guest did not choose.

Pre-selecting or defaulting any optional item.

Deciding whether a fee is mandatory or optional.

Publishing a was-price, a saving or a discount claim.

Automation boundaryAgent acts unaided

✓Draft an offer from the published set for a single booking into the review queue.

✓Show the amount, what it covers and its terms, at the offer for the named owner.

✓Record an acceptance per item, in the words the guest saw.

✓Flag what a person must weigh, and hold the offer for them.

Any write happens inside the boundaries agreed at implementation, never ahead of approval.

Setting which items may be sold, and at what price.

Bundling an air service into a booking on its own.

Explaining what a fee covers or whether it refunds.

Changes to offer, disclosure or acceptance rules.

Example output

One offer, annotated

Everything the agent offers is attached to the published item it was drawn from.

Ancillary offer output · single bookingIllustrative example

Booking

Offer shown

Price at offer

Source of record

Confidence

Attribution

Two-night stay

Late checkout, refundable until the night before arrival

$45.00

Published offer set

90%

Property and offer owner

As receivedTaken from the published offer set and the rate rules on file — nothing on this side is written by the agent.

Records used Published offer set Rate rules on file Prior sold price

Why this wordingIt carries the price and the terms at the offer, not at the payment screen.

ActionReleaseAdjustSend to pricing review

What the score decidesBelow the configured threshold the offer picks up a pricing read before the owner 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 bookingFrom the published set

03Offering

Offer from the published set

Draw on the rate and fare rules on file and the offer set your commercial team already publishes.

01Approved path

Offer it, never pre-select it

Routine bookings arrive with an eligible offer already drafted and already priced.

02Human review

Report opt-in quality, not attach rate

The page reports how many acceptances were affirmative and taken item by item, because an attach rate says nothing about how the yes was obtained.

04Build an evidence trail

The offer, the price shown with it and the guest's own acceptance stay on the record.

Integrations

Typical integrations

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

Booking and commerceOpera Cloud · SynXis
Amadeus · Sabre · Shopify

PaymentsAdyen · Stripe
Braintree · Worldpay

MerchandisingPlusgrade · Nuvola
Offer and upsell engines

Agent

Upsell & ancillary offers

Reads the booking
Drafts the offer
Holds for release

Messaging and CRMBraze · Salesforce
Twilio · Iterable

Observability & evaluationOpenTelemetry · Langfuse
Supported monitoring/evaluation sources

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

Agent controls

Six layers between the model and the guest

The layers sit one inside the next. What none of them catches is in the map below.

L6 · Outermost — last line of defenceInward → L1 · closest to the model

L6Rollback / safe modeFall back to the published offer set when evaluation or production signals degrade.Roll back

L5Version monitoringTrack model, prompt, offer-set and disclosure-configuration changes.Track

L4TraceabilityRecord the offer, the price shown, the flags, the adjustments and the acceptance.Record

L3Owner releaseHold offers for the named owner; it governs display, not whether a shown price is right.Gate

L2Policy guardrailsTest offers against configured opt-in, disclosure and reference-price rules; a failure returns the offer.Restrict

L1Confidence thresholdsRoute low-confidence offers to a pricing read before the owner sees them.Require review

Model coreOffer drafted — the item, the price shown, its terms and confidence

L1 – L2Test whether an offer may stand

L3Puts the display in an owner's hands

L4 – L5Keep the offer and the price shown with it

L6Falls back to the published offer set when signals degrade

How Nestack evaluates it

Evaluate the offering workflow — not only the line the guest taps.

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

Surface — the offer the guest sees

Depth of coverage ▼

E1Final-output evaluationDid the offer carry its price, its terms and nothing pre-selected?

E2Step-level evaluationDid the agent read the right booking, rate rules and published offer set?

E3Tool evaluationDid it read and write the correct booking and the correct item?

E4Confidence calibrationDo low-confidence offers actually attract more owner adjustments?

E5Slice evaluationHow does performance change across specific offer types?

E6Business outcomeHow many offers needed an adjustment or a refund after the sale?

Floor — the charge the operator answers for

Failure modes

Where each failure originates in the agent

Seven ways an offer goes wrong, placed at its stage.

Agent lifecycleDirection of processing →

01 · Retrieval1 mode

CO-03

Superseded offer set

An offer is drawn from a set the commercial team has replaced.

Stage gathersBookings, rate rules, inventory and the offer set

02 · Reasoning2 modes

CO-04

Unsubstantiated saving

A comparison is written that no prior price supports.

CO-06

Mandatory sold as optional

A charge the guest cannot avoid is offered as a choice.

Stage proposesThe offer the guest reads and accepts

03 · Tool / write2 modes

CO-02

Item added without a yes

A charge reaches the folio with no acceptance behind it.

CO-05

Terms held back to checkout

Price and terms appear only once payment is reached.

Stage writesOnly where write access and approval policy allow it

04 · Output1 mode

CO-01

Pre-selected optional item

An optional item is presented to a guest already ticked.

Stage returnsThe offer the owner releases and guests see

05 · Change / Version1 mode

CO-07

Silent offer regression

A model or rule change widens what the agent will offer.

Stage tracksModel, prompt, offer set and disclosure config

Sev-1 · a charge lands without a yes Sev-2 · a wrong price reaches a guest Sev-3 · offer set degrades, offer to review

See the complete Travel & Hospitality failure-mode catalogue →

Affected slices

Overall quality can hide one bad cohort

A lift figure means little without the base rate underneath it. The baseline here is a single-item room offer The cohorts that carry it are named, not averaged away Nestack reports it by slice rather than in aggregate..

Slice performance — reported separately, not only in aggregateIllustrative example

Slice	Failure rate	Lift	Status
Air-inclusive itineraries	3.5%	3.2×	Review
Recurring and renewing add-ons	2.6%	2.4×	Review
Comparison and saving claims	2.1%	1.9×	Watch
Single-item room offers	0.8%	0.7×	Normal

Bar: owner-adjustment-rate lift vs. single-item-offer baseline · scale 0–4.0× · tick marks the 2.0× review threshold 2 of 4 slices over threshold

Evidence-linked improvement

The cycle closes on a case, not a call

The cycle ends in a regression case, not in a meeting about what happened. That suite is what the next offer shown to a guest is measured against.

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

01Detect

Owner-adjustment rate rises in an offer slice.

02Diagnose

Read the offers that were charged against the ones that were declined, until the cause narrows to one.

03Improve

Every change is versioned against the offers that exposed it.

04Verify

A failing case holds the release back.

05Learn

The case joins the suite for good, and the offer rules are revisited.

Learn → DetectThe return edge. What comes next is tested against a suite that grew.

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, offering workflow, evaluation, integration, then production validation and handover.

Workstream Week 1Week 2Week 3Week 4Week 5Week 6

01Offer and acceptance workflow discovery and boundaries.

02Booking, commerce and payment assessment.

03Offer-set, disclosure and jurisdiction rule mapping.

04Booking ingestion and normalisation.

05Offer logic and published-set binding.

06Confidence scoring and acceptance routing.

07Offer-owner release workflow.

08Commerce and payment integration.

09Opt-in and disclosure cases.

10Guardrails and acceptance controls.

11Offer-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 property, one channel ProductionProduction commerce systems AdvancedMultiple markets / brands

Introduced at Pilot

Offers from your published set✓✓✓

Owner release✓✓✓

Offer-quality baseline✓✓✓

Introduced at Production

Reporting by offer type—✓✓

Approval workflow in your systems—✓✓

Approved write-back—✓✓

Commerce-platform integration—✓✓

Introduced at Advanced

Multi-market disclosure rules——✓

Multi-stage commercial approvals——✓

High offer volume——✓

Multi-market offer controls——✓

Build price From $5,000 From $8,000 Custom 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 published offer set and its prices → Booking ingestion and offer-set mappingWeek 1

02Offers you have already sold → Offer baseline, eligibility logic and published-set bindingWeek 2

03Your disclosure rules and what may be offered → Offer-set, disclosure and jurisdiction rule mappingWeek 1

04Access to relevant APIs, feeds or exports → Booking, commerce and payment assessment, then integration setupWeek 2

05Offers you would not want charged → Opt-in cases and the evaluation suiteWeek 4

06What no offer may add on its own → Confidence scoring, offer routing, guardrails and acceptance controlsWeek 3

07Named offer owners to release offers → Acceptance workflow, 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

The bands follow the real work, which is why evaluation and pilot share the fifth week.

Phase W1W2W3W4W5W6

Discovery W1

Build W2 – W3

Evaluate W4 – W5

Pilot & Launch W5 – W6

Week focus W1Offer workflow discovery, disclosure mapping and the automation boundary W2Source integration and the offer baseline W3Offering workflow, confidence logic and acceptance controls W4Opt-in cases, disclosure guardrails and failure-mode testing W5Commerce-platform integration, pilot offers and targeted corrections W6One selling period run under the commercial team, then handover

Reading the bandA bar covers the weeks its work is named in, and nothing else. The week 5 overlap is real, not padding.

At the end of W6The period closes validation and Agent Care assumes monitoring.

DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Travel & Hospitality AI agent

Build an upsell agent that offers, and lets the guest be the one who adds.

The vacatur everyone cites struck a rule on a procedural ground and left the statute, the older rule and the general deception authority exactly where they were — and in air transport an opt-out ancillary charge is a violation on its face. Send us your offer set, your disclosure rules and the markets you sell into, and we will tell you which of your current defaults has to go before anything ships. Then we build the boundary around what is left.

Discuss this agent →

Nestack Agents · Upsell and ancillaryAGT-TH-08 · Agent Care available after launch