The commercial opportunity for better service

GuestMemoryOS is designed to create commercial value when useful customer knowledge changes what a team does: a room is prepared correctly, today's dining suggestion reflects what the guest enjoyed yesterday, a concern reaches the team while it can still help, or a returning customer receives dependable service from a different employee. These eight studies explore proposed sector applications and quantify what selected improvements would be worth.

These are anonymised commercial opportunity studies based on public business disclosures. They are not accounts of deployments at the benchmark operators. All GuestMemoryOS revenue increases, service improvements, repeat-business effects and savings below are explicitly modelled, not achieved results. No customer relationship or endorsement is implied.

The studies use recognisable business profiles without naming the benchmark companies in the narrative. Each has its own public evidence, proposed service workflow, financial model and test. Public disclosures establish the business scale; they do not establish a deficiency in the operator's current systems or the effectiveness of AFG's product in that environment.

The dated operating snapshot

AFG's research readout dated 21 September 2026 covers activity since go-live on 14 July 2026. It reports 12,716 total activity entries, 5,375 staff observations or notes, 2,512 facts filed or updated, 1,453 personalisation opportunities identified, 114 service actions recorded and 703 guest confirmations or reactions recorded. Categories may overlap and must not be added together. An opportunity, a recorded action and a recorded response are different measures; a response is not necessarily a verified outcome. This company-reported snapshot is separate from the residence's record of 40 five-star Airbnb reviews. See the dated operating snapshot

GuestMemoryOS was built and used in live hospitality service. The dated operating snapshot, current residence reviews and internal engineering simulations answer different questions; they do not by themselves establish higher retention, lower costs, additional revenue or enterprise performance across eight industries. Those require measured customer outcomes, a credible comparison and complete costs.

The product helps teams turn observations into relevant briefs, suggestions and service actions, then carry useful learning forward. It is intended to work alongside the operator's PMS and CRM. Enterprise connections and each proposed sector workflow require their own scope and validation; a specific connector, certification or deployment timetable must be established for the intended operation. Platform description and integration scope

How to read the figures

Reported facts come from the linked sources. Assumptions are analytical inputs selected to make the economics testable. Calculated results follow from those inputs. A percentage point is an absolute change in a rate: conversion rising from 10% to 11% is a one percentage point increase, equivalent to 10% relative growth.

Revenue is additional sales after refunds, discounts and displaced purchases. Contribution is revenue less the incremental cost of delivering those sales. Released staff time becomes a cash saving only when an actual cost falls. Each case shows both. The central scenario is a calculation, not a forecast, and an unsuccessful deployment can produce no benefit or a loss.

Eight modelled opportunities at a glance

The table shows the central annual scenario at the stated deployment scope. Different scopes and currencies make the rows unsuitable for ranking market sizes or adding into a total.

Value before programme costs equals the modelled incremental contribution plus the modelled cash saving. It excludes unmonetised staff capacity and the separate repeat-business illustrations. All amounts are before tax. Deduct software, integration support, training, incremental supervision and other recurring programme costs to obtain net annual operating benefit. Deduct implementation and rollout investment separately in a cash-payback calculation.

The contribution-margin assumptions are 60% for hotel upgrades, 50% for cruise spending, 60% for airline upgrades, 25% for additional private flights, 45% for resort dining, 50% for retail sales, 80% for residential operator fees and 50% for club spending. These are scenario inputs, not the benchmark companies' disclosed incremental margins. They must be replaced with product and contract economics before investment approval.

Why the mechanisms differ

Hotels can earn upgrade revenue and reduce avoidable preparation work, while the brand receives only its contractual share. Cruise lines can connect preferences with purchases and delivery throughout a voyage. Airlines can improve service continuity while preserving fare, seat and safety controls. Private aviation can make repeat travel easier to arrange and deliver.

Integrated resorts can improve dining and hospitality across venues. Luxury retailers can preserve a client relationship when an advisor or location changes. Residential operators can improve recurring services over years. Clubs can strengthen member recognition and engagement, with renewal value assessed against membership capacity and replacement demand.

The commercial test is additional performance against the operator's existing tools and service practices. A stronger database alone is insufficient; the knowledge must lead to an appropriate action that a customer values.

Independent evidence for the mechanisms

The commercial argument has three parts: relevant recommendations can help customers choose; useful knowledge can help employees serve; and remembering an individual across visits could connect those benefits. Independent research supports the first two mechanisms. The third, and its financial value in each AFG sector, requires direct testing.

Relevant recommendations can improve purchasing outcomes

A randomised field experiment reported in Management Science found that product recommendations increased consumers' purchase probability by helping them discover products offering better value through lower prices, a closer match to their tastes, or both. The researchers identified product value as the principal mechanism, rather than easier navigation or simply showing more information. Recommendation experiment E1

The commercial inference is that a recommendation should earn its place by improving the customer's choice. A remembered preference could help a team propose an appropriate room, experience, item or service. However, this research studied an online recommendation system, not GuestMemoryOS or these benchmark businesses. It does not establish a transferable revenue percentage, a premium-price effect or repeat-booking gains. The models therefore do not borrow an uplift from it.

Useful knowledge can improve frontline productivity

A study of the staggered introduction of an AI conversational assistant across 5,172 customer-support agents found an average 15% increase in issues resolved per hour. The research appears in the 2025 Quarterly Journal of Economics; the authors' updated paper describes substantial variation, with stronger benefits for less experienced workers and small quality declines among the most experienced and highest-skilled workers. Published study E2 and authors' paper

This provides evidence that giving employees useful guidance during service can improve productivity in an operating business. It was a different product in a different setting, studied through a staggered rollout; it was not a GuestMemoryOS trial. A 15% productivity gain is also not a 15% reduction in payroll. The case studies use separately labelled time-saving assumptions and require observed expense reductions before recognising cash savings.

What this means for the eight studies

These findings support the logic of testing better recommendations and better staff guidance. AFG's own reported operation adds evidence that its memory workflow has operated in a hospitality setting. None of these sources establishes that persistent guest memory causes the proposed rebooking, service or profit changes across all eight sectors.

The defensible claim is: there is evidence for relevant components of the mechanism, a reported operating example of AFG's product, and a transparent economic case for testing each customer application. Demonstrating sector-specific commercial performance is the next evidential step. Each following study defines what to measure and what would justify expansion.

1 Global hotel system

Service continuity across properties and owners

Anonymised opportunity study. Proposed outcomes are modelled; no deployment at, relationship with or endorsement by the benchmark operator is implied.

Commercial proposition: At an assumed deployment covering 10 million eligible occupied room nights annually, an additional one percentage point of paid-upgrade attachment at $50 per night would produce $5 million of additional room revenue. Better preparation could also release 50,000 staff hours in a separately defined arrival workflow.

The public benchmark

The benchmark is a global hotel system whose year-end 2025 portfolio contained 9,805 lodging properties and 1,779,936 rooms or units. Its loyalty programme had nearly 271 million members; member stays represented 68% of worldwide room nights. Its comparable worldwide 2025 average daily rate was $185.81 in constant dollars. These describe a large existing customer relationship and distribution system. Membership penetration is not a repeat-booking rate, and the total portfolio includes residential and other lodging products. 2025 results and operating tables H1

How GuestMemoryOS would create value

Consider a proposed workflow for a family travelling between properties. On one visit, a parent confirms a preference for connecting accommodation, an early housekeeping window and breakfast before an excursion. The family also explains which previous room arrangement worked poorly. These are scenario details, not a reported guest case.

1. The team connects each preference to the correct person and travel party, retaining its source and permission to reuse it.
2. Before the next stay, the family confirms what still applies. The property checks room inventory, entitlements and any additional charge.
3. Reservations can offer a suitable available upgrade. Housekeeping receives the confirmed preparation requirements and records completion.
4. Feedback updates the memory, so a later team can reuse the successful arrangement without repeating an obsolete one.

The paid upgrade earns revenue only when the guest chooses an additional benefit. Preparation improves the stay whether or not the guest buys anything. The proposed implementation builds on AFG's published hotel workflow. Hotel application

Service levels and repeat bookings

A useful service test is the share of eligible arrivals requiring a correction to a previously confirmed, feasible preference. An illustrative target of 12% falling to 9% means three fewer corrections per 100 arrivals, a 25% relative reduction. Neither rate is a disclosed operator result. Log corrections and guest feedback, rather than assuming a completed task proves satisfaction.

For a separate cohort of one million unique eligible booking customers, a one percentage point increase in completed repeat stays would mean 10,000 additional stays. At an assumed three room nights per stay and the reported $185.81 rate used as a proxy, that represents $5.574 million of room revenue before displacement and delivery costs. The cohort size, room nights and uplift are assumptions. This illustration is excluded from the core model until a 12-month comparison verifies incremental demand.

Hotel financial model and proof requirement

Model boundary: Existing booked stays in a proposed annual rollout. The 10 million eligible room nights, $50 net upgrade charge and attachment changes are assumptions. The model does not apply an uplift to the hotel group's consolidated revenue.

Calculation: 10,000,000 eligible room nights × attachment increase × $50. These are net new upgrades after subtracting upgrades that would have happened anyway, refunds, discounts and revenue lost by occupying a room another customer would have purchased. The charge is per upgraded night, so the unit matches the denominator.

What it could save

Assume one million eligible arrival-handling cases a year and three minutes of net staff time released per case. That produces 50,000 hours. At an assumed loaded cost of $30 per hour, the released capacity has a labour-cost equivalent of $1.5 million.

If 25% of that value reduces overtime, agency work or another identifiable expenditure, the cash saving is $375,000. The remaining $1.125 million is a capacity equivalent, available for better service. It is excluded from the financial benefit. The three-minute saving must already deduct the time required to capture, confirm and maintain the memory.

At the central revenue scenario, $3 million contribution + $375,000 cash savings = $3.375 million before recurring programme costs. This is conditional modelled operating value, not an AFG price or a reported profit improvement.

Who receives the benefit

The headline model measures property revenue and property operating value. Its assumed contribution margin includes management or franchise fees. An assumed 5% fee on $5 million of additional room revenue would transfer $250,000 to the brand. That is brand fee revenue, with its own delivery costs, rather than another $250,000 of guest spending. Assess the brand's return separately from the property's return. A combined owner-and-brand contribution analysis must eliminate the fee transfer and count the underlying costs once.

How to establish proof

Start with a defined set of properties and a concurrent comparison using the existing upgrade process. Where practical, assign travel parties to treatment and control and preserve that assignment across visits. If staff behaviour spills between groups, use enough independent properties or teams for cluster analysis.

Measure net room contribution per assigned eligible night, preparation corrections, guest feedback and net handling time. Keep pricing and inventory rules consistent; account for property, season, room category and trip purpose. Use completed repeat stays over 12 months for the retention claim. Expansion is justified only by positive incremental contribution after full costs, with service outcomes preserved or improved.

2 Multi-brand cruise group

More relevant purchases with dependable delivery

Anonymised opportunity study. Proposed outcomes are modelled; no deployment at, relationship with or endorsement by the benchmark operator is implied.

Commercial proposition: A 1% improvement within an assumed eligible half of a large cruise group's onboard and other revenue would be worth $27.095 million annually at its reported 2025 scale. The opportunity depends on improving the complete voyage's economics, including purchases made before departure.

The public benchmark

The benchmark group reported 9,446,010 passenger trips, $17.935 billion total revenue and $5.419 billion onboard and other revenue in 2025. These are consolidated group figures, not the results of one brand. Passenger trips are not unique customers. 2025 public filing C1

Nearly half of its onboard revenue was booked before sailing, and 90% of pre-cruise purchases used digital channels. This establishes a substantial existing route to the customer. AFG must improve outcomes against that existing capability. 2025 purchasing disclosure C2

How GuestMemoryOS would create value

In a proposed family workflow, one traveller wants early dining, another prefers a quieter excursion and a teenager enjoys active experiences. The reservation establishes who is travelling; confirmed individual preferences help the teams understand what each person wants from the voyage.

Before departure, the operator can offer available dining, excursion or spa options that fit the party's stated interests and timetable. Current inventory, age restrictions, entitlements and prices stay under the existing booking systems. A purchase produces a service requirement for the responsible team; delivery and feedback determine what should be retained.

If a guest changes their preference on day two, the change can inform another venue during the same sailing. At a later sailing, the new crew receives relevant, permitted knowledge for reconfirmation. The commercial mechanism combines relevance with reliable execution. It does not depend on asking guests to buy more of something they already declined. Cruise application

Service levels and future sailings

One service target could be first-attempt fulfilment rising from an illustrative 80% to 90% of confirmed, feasible service requests. That is a 10 percentage point improvement and halves the unfulfilled share from 20% to 10%. This is a target for testing, not a current performance assessment.

For an assumed cohort of 100,000 unique eligible customers, a one percentage point increase in completed repeat sailings means 1,000 additional passenger trips if each returning customer books one trip. Reported ticket revenue divided by reported trips gives a 2025 scale proxy of approximately $1,324.90 per trip. The resulting $1.325 million ticket-revenue illustration is excluded from the core spending model. It becomes incremental value only after allowing for capacity, displaced demand, sales costs and the actual passenger mix. Ticket revenue and operating statistics C1

Cruise financial model and proof requirement

Model boundary: $5.419 billion reported onboard and other revenue × an assumed 50% combined share for eligibility and deployment = $2.7095 billion. This share is not a measured eligible base. Discovery must exclude revenue that preferences cannot influence, such as relevant fees, and establish the participating brands and vessels.

Calculation: $5,419,000,000 × 50% × assumed uplift. The central outcome equals about $2.87 per reported passenger trip across the whole group. This is a scale comparison, not a predicted average for the reached guests. With only 25% effective coverage, the same 1% uplift would produce $13.548 million additional revenue.

The result must include all spending before and during the voyage, net of refunds and discounts. An excursion bought earlier is not additional revenue if it would otherwise have been bought onboard. The contribution margin must include fulfilment costs, commissions and purchases displaced from other venues.

What it could save

Assume two million eligible service interactions annually and two minutes of net handling time released per interaction: 66,667 hours. At an assumed $30 loaded hourly cost, that is $2 million of capacity equivalent. If 25% reduces an actual operating expense, cash savings are $500,000. The remaining capacity is excluded from financial benefit and can be used to improve service.

The central model therefore produces $14.048 million of annual operating value before programme costs: $13.548 million additional contribution plus $500,000 cash savings. A full rollout must deduct recurring technology, crew training, memory-review and support costs and account for implementation investment and adoption ramp.

How to establish proof

Begin with one vessel for workflow feasibility, then use enough comparable sailings or vessels to estimate commercial impact. Test one defined purchase category and the related service workflow. Randomise parties where practical; use voyage or vessel clusters when crew behaviour creates spillover.

The primary financial measure is total net incremental contribution per assigned eligible passenger trip. Stratify for cabin class, itinerary, voyage length, promotions and season. Measure first-attempt fulfilment and actual handling time, not simply the number of notes captured. Follow repeat purchases over 12 to 18 months and distinguish completed sailings from deposits or expressions of interest.

A brand rollout must use that brand's own eligible figures. Group revenue cannot be allocated to ships by ship count or berth share. The case for fleet expansion rests on replicated results and an operator-approved cost model.

3 Premium international airline

Continuity from lounge to cabin to the next journey

Anonymised opportunity study. Proposed outcomes are modelled; no deployment at, relationship with or endorsement by the benchmark operator is implied.

Commercial proposition: For a long-haul airline carrying 53.2 million passengers annually, an assumed programme reaching 10% of passenger trips could generate $7.98 million of additional fare revenue if it creates a one percentage point increase in eligible paid-upgrade conversion at a $150 net fare increment.

The public benchmark

The benchmark is a Dubai-based international airline with a large premium-cabin operation. For the year ended 31 March 2026, it reported 53.2 million passengers and AED130.9 billion revenue, approximately US$35.7 billion. Its network covered 152 cities in 80 countries at year end. These figures refer to the airline, not the wider group; airline revenue also includes cargo and is not a pure passenger-fare base. Annual results A1

How GuestMemoryOS would create value

A proposed workflow begins with a passenger who confirms that they prefer to sleep soon after takeoff, receive a short rather than extended service introduction, and use a particular communication channel for service follow-up. The record belongs to the individual and distinguishes a recurring preference from the circumstances of one trip.

With permission, relevant information could be available to the lounge and cabin teams. Today's meal choice, seat entitlement and service conditions remain authoritative. A resolved complaint can be handed over accurately so that the passenger does not need to explain it again.

For a later flight, a preference for more personal space may justify presenting an available paid cabin upgrade. The airline's pricing and inventory systems decide what can be offered. An accepted purchase is followed through to service delivery. A declined offer is respected; essential care does not depend on purchase.

AFG's airline application is designed to assist service alongside operating systems. It does not authorise changes to safety procedures, medical protocols, departure decisions or aircraft configuration. Airline application

Service levels and repeat travel

For eligible, reconfirmed noncritical service preferences, an illustrative miss rate falling from 10% to 7% would mean 3,000 fewer misses per 100,000 assessed opportunities. The target is a 30% relative reduction. It must be measured from actual service and customer feedback; the public disclosures do not establish these starting rates.

An assumed cohort of 500,000 unique eligible customers with a 0.5 percentage point increase in completed repeat journeys would generate 2,500 additional journeys. At an assumed $800 net fare per journey, the gross revenue opportunity is $2 million before displacement and costs. This is excluded from the core model. Neither the fare nor the repeat effect is a disclosed airline figure.

The service case is strongest where a preference survives a real handover between teams and makes the next journey easier for the customer. The pilot must establish that the current systems leave such a workflow to improve.

Airline financial model and proof requirement

Model boundary: 53.2 million reported passenger trips × an assumed 10% eligible deployment share = 5.32 million eligible passenger trips. This does not mean 10% premium passengers, 10% unique people or the airline's actual upgrade-eligible population.

Calculation: 5,320,000 eligible trips × conversion increase × $150 net additional fare. The $150 increment and conversion changes are assumptions. Net value must deduct cannibalisation of existing paid upgrades and the opportunity cost of seats that would otherwise sell at a higher fare.

The model uses eligible trips rather than multiplying total airline revenue by a percentage. An expansion in routes, aircraft capacity or base fares is not an AFG effect. Reported annual scale is a historical reference, not a forecast of the coming year's flying programme.

What it could save

Assume one million eligible ground-service or customer-care cases annually, with two minutes of net handling time released per case. At an assumed $30 loaded hourly cost, 33,333 hours have a capacity equivalent of $1 million. Converting 25% into reduced paid overtime, outsourced handling or another actual expense would save $250,000.

This model does not assume that cabin crew headcount or mandatory crew ratios fall. Released time can also be used to improve care. All time measurements must include the work needed to capture, review and confirm preferences.

At the central scenario, additional contribution of $4.788 million plus $250,000 cash savings produces $5.038 million before recurring programme costs. Integration with passenger, loyalty, service and booking systems must be scoped and costed before calculating net return.

How to establish proof

Select comparable routes and a defined noncritical service workflow. Assign eligible customers or travel parties consistently to the pilot and comparison groups; allow for crew and flight-level spillover. Preserve fare controls and revenue-management rules so that an apparent conversion gain is not simply a larger discount.

Track contribution per eligible assigned trip, service-preference fulfilment, customer-care handling time, complaints and opt-outs. Segment for route, cabin, customer status, load factor and disruption. AFG should receive no credit for unrelated schedule, fleet or pricing changes.

Measure completed repeat journeys over an agreed 6 to 12-month window for customers with a realistic opportunity to travel again. Exclude re-accommodation after a disrupted flight from loyalty-related rebooking. Expansion requires financial benefit after full costs and an acceptable passenger and staff experience.

4 Membership private aviation

Preserving client knowledge across flight teams

Anonymised opportunity study. Proposed outcomes are modelled; no deployment at, relationship with or endorsement by the benchmark operator is implied.

Commercial proposition: A 1% improvement in an assumed eligible half of a US private aviation platform's reported flight revenue represents $3.113 million of annual additional revenue. The mechanism is retaining or winning additional suitable trips through easier planning and consistent delivery.

The public benchmark

The benchmark is a US private aviation platform offering memberships and charter, with a strategic relationship to a major scheduled airline. It reported $622.688 million flight revenue in 2025, within $736.495 million total revenue. Its disclosed 44,694 live flight legs are completed one-way revenue-generating private flights. 2025 public filing P1

Its reported private-jet gross bookings were $833.904 million. Gross bookings and recognised revenue are different measures; this model uses flight revenue. The operator has also publicly described a more integrated, personalised commercial service model. AFG must demonstrate additional value against that existing investment. Results and definitions P2

How GuestMemoryOS would create value

For a proposed repeat-client workflow, the client confirms catering preferences, the desired cabin service style and how their assistant should receive updates. Family travel and executive travel remain separate contexts. A preference from a family holiday should not automatically shape a board delegation's flight.

The relationship team can prepare a relevant proposal using the current party and itinerary. Once the client accepts, authorised information goes to the teams responsible for cabin preparation, catering and ground transport. The responsible staff verify delivery and record a change if the client's requirements differ.

On the next flight, a new host or relationship manager can continue the service relationship. The client gains less planning work and more consistent delivery; the operator can compete for future trips on that service quality. Aircraft availability, operational approvals and contractual commitments remain in the aviation systems. These are proposed uses of AFG's private-aviation memory design. Private aviation application

Service levels and repeat flying

For eligible noncritical preparation requirements, an illustrative first-time-right rate rising from 90% to 95% halves the error share from 10% to 5%. Measure cabin, catering and communication preparation separately from technical dispatch reliability. Preference memory does not itself demonstrate fewer flight cancellations or delays.

For an assumed cohort of 1,000 client accounts, a one percentage point increase in retained accounts means ten accounts. If each produces four genuinely additional flight legs, that is 40 legs. The reported flight revenue divided by live legs gives a blended proxy of approximately $13,932 per leg; the revenue illustration is about $557,290. It is not a quoted charter price or a verified client value.

Retained flying is a possible driver of the core revenue increase, so this amount must not be added to it. Measure account-level completed flying, cancellations and actual itinerary contribution.

Private aviation financial model and proof requirement

Model boundary: $622.688 million of reported annual flight revenue × an assumed 50% combined share for eligible clients and deployment = $311.344 million. Wholesale activity and other unsuitable business must be removed when the actual population is identified.

Calculation: $622,688,000 × 50% × assumed uplift. Revenue must arise from additional completed flying or demonstrably improved economics, rather than merely receiving a prepaid balance earlier.

The 25% incremental margin is an assumption requiring particular scrutiny. Additional trips can require repositioning, purchased lift, extra crew costs or scarce peak-period capacity. The operator's historical gross margin, adjusted contribution margin and the marginal profit of one more itinerary are not interchangeable.

What it could save

Assume 10,000 eligible flight-preparation cases a year and 15 minutes of net coordination time released per case. That is 2,500 hours. At an assumed loaded cost of $50 per hour, the capacity equivalent is $125,000.

If one-quarter reduces an actual expenditure, the cash saving is $31,250. The remaining 1,875 hours can support service. Do not treat a faster client conversation as a wage saving when the staff schedule and expense remain unchanged.

The central scenario therefore creates $809,610 of annual operating value before programme costs: $778,360 incremental contribution and $31,250 cash savings. Aircraft acquisition, ordinary fleet transformation and unrelated commercial growth are outside this model.

How to establish proof

Pilot with a defined set of account teams and matched or randomised client accounts. Account-level assignment is preferable to switching the same repeat client between incompatible service processes. Compare against the current personalised relationship service, including existing notes and client-management tools.

Measure completed-flight contribution per eligible account, net preparation time, first-time-right service and client feedback. Control for aircraft category, route, stage length, peak dates, pricing, repositioning and capacity restrictions. Distinguish a retained account from a member who prepays but does not fly during the measurement period.

Allow an observation window aligned with the accounts' booking and renewal cycles, typically assessed over 6 to 12 months in the proposed design. The actual sample and duration must follow booking frequency and statistical power.

Rollout is commercially justified only if the extra flying has positive incremental contribution after capacity and delivery costs. Service continuity is valuable, but a higher booking total does not by itself establish higher profit.

5 Integrated resort group

Connected hospitality across dining and guest services

Anonymised opportunity study. Proposed outcomes are modelled; no deployment at, relationship with or endorsement by the benchmark operator is implied.

Commercial proposition: A 1% increase within an assumed eligible half of a major integrated resort group's food and beverage revenue would produce $3.22 million annually. This study focuses on hospitality spending and service continuity.

The public benchmark

The benchmark group operates major resort destinations in Singapore and Macao. Its 2025 results reported $644 million food and beverage revenue, $1.422 billion rooms revenue and $13.017 billion total revenue. The reporting categories are materially different, so the model isolates dining rather than applying a percentage to the whole business. 2025 issuer results R1

These venues already have host, loyalty, reservation and service systems. Operational discovery must establish whether a specific handover leaves useful information unused, and whether GuestMemoryOS can improve it.

How GuestMemoryOS would create value

In a proposed workflow, a returning resort guest confirms a preference for quieter dining, an early show and concise communication through the concierge. Another member of the party has a different dining schedule. The preference belongs to the relevant person and occasion, not to the whole party by default.

The host can coordinate an available dinner and entertainment sequence that fits the guest's timetable. Restaurant staff receive the confirmed service details, and guest services can see whether a previous service issue was resolved. A shift change does not require the guest to start the same conversation again.

The revenue mechanism is more relevant paid hospitality combined with fewer failed arrangements. Recommendations must be based on customer-stated interests, available capacity and current permissions. The model excludes wagering, gaming-loss targeting, complimentary entitlements and restricted customers from commercial targeting. No gaming revenue uplift is claimed. Integrated resort application

Service levels and return visits

An illustrative service target is repeat contact about an already confirmed request falling from 10% to 7%. That is three fewer repeat contacts per 100 eligible requests, a 30% relative reduction. Measure the customer's reason for recontact: a new request is not a failure, and a closed ticket does not prove the original request was fulfilled.

For an assumed cohort of 100,000 unique eligible resort customers, a one percentage point increase in completed return visits would mean 1,000 additional visits. At an assumed $500 net room-and-hospitality basket, that is $500,000 additional revenue before displacement and costs. This separate illustration is excluded from the core dining model.

An extra return visit only adds economic value to the extent it generates incremental contribution. On a sold-out date, replacing another guest at the same economics may add no room revenue. Any reduced acquisition cost must be demonstrated separately.

Resort financial model and proof requirement

Model boundary: $644 million reported food and beverage revenue × an assumed 50% effective eligible deployment share = $322 million. The actual pilot must identify eligible venues, paid transactions, guests and service opportunities.

Calculation: $644,000,000 × 50% × assumed uplift. Food and beverage disclosures may include allocations for complimentary services. The eligible base must exclude those allocations where no additional customer payment or incremental economic benefit results. The assumed 50% share is not evidence that this adjustment has already been measured.

Revenue should be measured across the relevant resort venues. Moving a guest from one restaurant to another does not add group revenue unless the total economics improve. Deduct discounts, refunds, displaced purchases, food costs and additional service labour when estimating contribution.

What it could save

Assume 500,000 eligible guest-service handling cases annually, with two minutes of net staff time released per case. This produces 16,667 hours. At an assumed $30 loaded hourly cost, the capacity equivalent is $500,000.

A 25% conversion into actual expense reduction produces $125,000 cash savings. The other hours can improve service availability but are excluded from the cash case. Staff time used to capture and reconcile preferences must be deducted from the saving.

The central annual operating value is $1.574 million before programme costs, comprising $1.449 million contribution and $125,000 cash savings. Spending on the platform, integrations, venue training and ongoing review must fit inside a proven benefit with an acceptable return.

How to establish proof

Start with a defined hospitality workflow across a small number of connected venues. Use comparable guest cohorts or enough independent venue and time clusters to address staff spillover. Keep promotions, access rules and complimentary policies consistent between groups.

The financial outcome is total incremental hospitality contribution per eligible assigned customer over the visit. Supporting measures include request fulfilment, repeat contacts, time to resolution, customer feedback, refunds and net handling time. Track cross-venue purchases so that transfers are not mistaken for growth.

Follow completed return visits over an agreed 6 to 12-month period and account for dates, room availability and substituted demand. There is no basis in the cited facts for claiming that GuestMemoryOS has already improved retention at the benchmark group.

The successful case would show that more consistent hospitality earns additional profitable custom while making service easier for the guest and the team.

6 Luxury retail group

Keeping the client relationship when the advisor changes

Anonymised opportunity study. Proposed outcomes are modelled; no deployment at, relationship with or endorsement by the benchmark operator is implied.

Commercial proposition: A 0.5% increase within an assumed quarter of a major luxury group's physical retail sales would represent approximately €19.90 million of annual additional net sales. The opportunity is continuity of relevant advice and fulfilment across appointments, advisors and locations.

The public benchmark

The benchmark is a Swiss luxury group with major jewellery houses and specialist watchmakers. For the year ended 31 March 2026, it reported €22.420 billion sales. Physical retail represented 71% of group sales; direct-to-client sales, including online retail, represented 77%. These reported percentages are rounded. Annual results L1

The scale of direct client interaction is established. The disclosures do not show whether a particular house lacks effective clienteling or whether AFG would improve its existing systems.

How GuestMemoryOS would create value

A proposed client workflow begins when an advisor learns why a previous piece was chosen, which fitting adjustment mattered, and how the client prefers to be contacted. A gift purchase is identified as a gift; it does not automatically become the purchaser's personal style preference.

With appropriate permission, a later advisor can prepare for an appointment using current, verified context. The stock system confirms which suitable pieces are available, and the client confirms whether the earlier occasion, budget or sizing information still applies. The advisor can then give relevant service without reconstructing the whole relationship.

After a purchase or alteration, the team records the outcome. A service problem and its resolution remain accessible to the authorised team, so that the client does not need to repeat it after changing boutique or advisor.

The financial mechanism is a better conversion rate or more retained profitable purchases, net of returns. Recommending unavailable products, sending more messages or advancing the timing of an inevitable purchase does not establish incremental sales. AFG's published retail application addresses style, fit, purchase context, communication and service recovery. Luxury retail application

Service levels and repeat purchases

An illustrative service target is appointments prepared with verified, relevant client requirements increasing from 80% to 90%. This is ten more prepared appointments per 100 eligible appointments. Measure whether the requirements were correct and useful, not merely whether a profile existed.

For an assumed 100,000 unique eligible client cohort, a one percentage point increase in completed repeat purchases would mean 1,000 additional orders. At an assumed €2,000 net retained basket, the opportunity is €2 million of net sales before delivery costs. This is a potential driver of the main net-sales uplift and must not be added again.

Respectful contact and accurate memory are essential to this proposition. Measure opt-outs, inappropriate recommendations and corrections as well as sales.

Retail financial model and proof requirement

Model boundary: €22.420 billion sales × reported 71% physical retail share × assumed 25% effective eligible deployment share = approximately €3.980 billion. The arithmetic retains unrounded intermediate values, but the estimate is approximate because the source percentage is rounded.

Calculation: €22,420,000,000 × 71% × 25% × assumed uplift. The eligible share must be replaced by the actual participating houses, boutiques, clients and sales categories. Permission to remember a client within one house does not automatically permit sharing across every house in the group.

The 50% contribution margin is an assumption after cost of goods and incremental selling, fulfilment, alteration and return costs. A disclosed group gross margin is not evidence of this marginal rate. Measure retained sales across channels and a sufficient time window to identify purchases displaced from another boutique or pulled forward from a later date.

What it could save

Assume 500,000 eligible appointments a year and five minutes of net preparation time released per appointment. That equals 41,667 hours. At an assumed €40 loaded hourly cost, the capacity equivalent is €1.667 million.

If 25% reduces an actual expense, cash savings are approximately €416,667. The rest can support more useful service. If that released capacity is what enables the additional sales, its full labour value cannot also be counted as a separate financial benefit.

The central model yields approximately €10.366 million before recurring programme costs, combining incremental contribution and the assumed cash saving. Inventory scarcity, client permissions and advisor adoption can materially reduce the eligible opportunity.

How to establish proof

Pilot within one house before testing portability more widely. Compare eligible clients receiving the workflow against a concurrent group served by existing clienteling tools and practices. Randomise at client or advisor-team level where feasible and account for boutique clustering.

Use net retained contribution per eligible assigned client as the financial measure. Track prepared appointments, wrong recommendations, returns, advisor preparation time, opt-outs and customer feedback. Hold promotions and discount authority constant. Evaluate cross-channel and cross-boutique sales together to prevent internal transfers appearing as growth.

A 6 to 12-month purchase window may be needed in the proposed design, with a longer period for infrequent categories. The actual duration depends on purchase frequency and the sales effect the operator needs to detect. Evidence must show that the relationship produces additional retained economic value, not just more recorded interactions.

7 Branded residences

Reliable household service over years

Anonymised opportunity study. Proposed outcomes are modelled; no deployment at, relationship with or endorsement by the benchmark operator is implied.

Commercial proposition: At the scale of a disclosed 160-residence property, an assumed programme serving 120 participating households could produce $43,200 annual operator fee revenue from one additional requested service per household per month, alongside $10,400 cash savings if a defined part of released staff time reduces expenditure.

The public benchmark

The benchmark is a global luxury hospitality operator that reported 61 residential properties across 20 countries in January 2026. An operating residential tower in Boston lists 160 private residences in its official property facts. These establish a real portfolio and property scale; they do not disclose participating households or residential service revenue. Residential portfolio B1 and property facts B2

How GuestMemoryOS would create value

In a proposed workflow, a household confirms preferred service windows, who should receive routine communication and how the residence should be prepared before the owners return. The system distinguishes the owner, other residents, an authorised assistant and a visiting guest.

Concierge can offer a relevant optional service when requested or permitted: household preparation, transport or an event arrangement. The team checks the current price and availability, assigns delivery and verifies the result. Maintenance can retain context about a resolved issue while the building's work-order system remains authoritative for the asset and technical repair.

When an employee changes, the household should not need to rebuild the same service instructions from memory. When an owner, authorised contact or access arrangement changes, the record must change too. Historical preferences never substitute for current permission to enter a residence.

AFG's residential application is aimed at continuity across concierge, maintenance and other service teams over a long relationship. The commercial case belongs to the actual operator or contracted service provider receiving the fees. Residential application

Service levels and the equivalent of rebooking

An illustrative target is avoidable repeat contact on eligible routine service requests falling from 20% to 15%. That is a 25% relative reduction. Technical faults that require a legitimate second visit should not be reclassified as a memory failure or suppressed to improve the metric.

The relevant repeat-business measure is renewal or repeated use of optional resident services. Ownership is not an annual hotel booking. For 120 assumed eligible households, a five percentage point increase in optional package renewal means six more annual subscriptions. At an assumed $600 per package, that represents $3,600 of retained service revenue, before substitution and fulfilment costs.

That renewal illustration is excluded from the core model. A renewal of a service already counted in the monthly transaction model cannot be added again. No home-price premium, property-sales uplift or brand-wide residential profit is inferred from these figures.

Residential financial model and proof requirement

Model boundary: One property with 160 disclosed residences. Assume 75%, or 120 households, are participating and service-eligible. Each additional service transaction has an assumed $150 customer price, of which the operator retains a 20% fee as agent. The remaining amount belongs to the fulfilling provider.

Calculation: 120 households × 12 months × extra monthly transactions × $150 × 20%. These transaction frequencies, prices, participation and fees are assumptions, not disclosed terms. Gross supplier transactions are not operator revenue.

At an assumed 80% contribution margin on the retained fee, the central $43,200 revenue produces $34,560 incremental contribution. This assumes the operator acts as an agent. If it instead delivers the service as principal, both recognised revenue and delivery costs need a different model.

What it could save

Assume two eligible routine requests per participating household per week for 52 weeks, with five minutes of net handling time released per request. The resulting 12,480 requests release 1,040 hours. At an assumed $40 loaded hourly cost, the capacity equivalent is $41,600.

If 25% reduces actual expenditure, cash savings are $10,400. Annual operating value before programme costs is therefore $44,960: $34,560 contribution plus $10,400 cash savings. Reduced expenditure may accrue to the residents' association, the management company or a supplier, depending on the contract. It should be assigned to the correct beneficiary.

This is a property-level case. Multiplying it by 61 properties would assume identical unit counts, participation, demand and contracts across the portfolio. The public evidence does not support that extrapolation.

How to establish proof

Use a consented household cohort and comparable service teams. If one property's small population cannot establish commercial impact precisely, extend the test to comparable buildings rather than claiming certainty from a handful of renewals.

Measure operator contribution per eligible household, correctly fulfilled routine requests, repeat contacts, optional-service renewals and net handling time. Record who receives every fee and bears each cost. Test whether additional transactions replace purchases the resident would have made through the same operator anyway.

Follow service use and renewal through an annual cycle, including seasonal absences. A household that has sold its residence or has no opportunity to renew should not remain in the same renewal denominator.

The practical result to demonstrate is consistent household service that residents value and the operator can deliver profitably. A portfolio investment decision needs each property's actual participation and contract economics.

8 Private members club network

Member recognition across houses and teams

Anonymised opportunity study. Proposed outcomes are modelled; no deployment at, relationship with or endorsement by the benchmark operator is implied.

Commercial proposition: A 1% improvement within an assumed eligible half of a global club network's in-house revenue would produce $2.408 million annually. The case combines relevant use of the club with recognition that survives changes in staff and location.

The public benchmark

The historical benchmark is a members club network with 45 houses and 212,447 house members at 29 December 2024. That year's disclosed in-house revenue was $481.613 million, and membership revenue across its membership businesses was $418.026 million. Its reported global waiting list exceeded 112,000 applicants. These are a dated full-year baseline, not a statement of the network's current size. 2024 public filing M1

The house-member count and total membership revenue have different scopes. Dividing them would not establish an accurate annual fee per member. The waiting list also changes the economics of retention: a departing member might be replaced quickly.

How GuestMemoryOS would create value

In a proposed workflow, a member confirms a preferred dining setting, how they like to receive relevant invitations and what resolved a previous service issue. The record can help an authorised host at another house prepare an appropriate welcome without disclosing information beyond that host's role.

An existing reservation or event system checks capacity and entitlement. Staff can suggest an available dinner, stay or activity that fits a stated interest; the member remains free to decline. The service team records what happened and updates preferences when the member changes them.

The value comes from making the membership easier and more satisfying to use. The profile is useful only when it informs dependable service. It should not create an entitlement to a table, room or event that the current membership and reservation rules do not provide. Private club application

Service levels and membership renewal

For confirmed, feasible member requests, an illustrative correct-fulfilment rate rising from 90% to 95% means five more successful outcomes per 100 requests. It halves the failure share. Verify the result through service records and member feedback, while monitoring inappropriate contact or recognition.

For an assumed cohort of 100,000 eligible paying members, a one percentage point improvement in renewal means 1,000 additional retained members. At an assumed $2,500 annual fee, that protects $2.5 million of gross dues associated with those members.

That is not automatically $2.5 million of extra revenue. Replacement members, joining timing, existing capacity and different fee levels determine the net effect. A readily available replacement could make the incremental dues benefit close to zero. Acquisition savings or improved member spending must be measured. The renewal illustration is excluded from the core financial model.

Club financial model and proof requirement

Model boundary: $481.613 million reported 2024 in-house revenue × an assumed 50% effective eligible deployment share = $240.8065 million. The actual participating houses, customers, transactions and permission boundaries must be established before rollout.

Calculation: $481,613,000 × 50% × assumed uplift. Figures are rounded to the nearest dollar after calculation. In-house revenue is kept separate from membership dues and the group's other businesses.

Measure net spending across participating houses, not one venue in isolation. Moving a member's dinner or bedroom booking from one house to another does not itself create network revenue. Included benefits, discounts, refunds, capacity displacement and variable delivery costs must be reflected in the contribution figure.

What it could save

Assume 500,000 eligible member-service cases a year and two minutes of net handling time released per case. That produces 16,667 hours with a capacity equivalent of $500,000 at an assumed $30 loaded hourly cost.

Converting 25% into an actual expense reduction would save $125,000. The remaining capacity can support recognition and service, but it is excluded from the cash benefit. Any time used to enter, maintain and reconfirm memory has already been deducted in this assumed net saving.

The central case produces approximately $1.329 million annual operating value before recurring programme costs. This consists of approximately $1.204 million contribution and $125,000 cash savings. It does not include gross retained membership dues.

How to establish proof

Pilot within selected houses and compare eligible members with a concurrent group served using the current tools and host practices. Preserve assignment across visits where possible, and account for members visiting several houses. Use enough houses or independent teams if staff behaviour affects both groups.

Measure incremental contribution per eligible assigned member, request fulfilment, complaints, net handling time and member feedback. Track engagement quality and opt-outs as well as visits. Holding membership rules and promotions constant helps distinguish a service-memory effect from a different offer.

Observe renewal through a complete renewal cycle. Deduct the contribution that would have been earned from replacement members and allow for the timing of dues. A higher renewal rate may improve continuity without producing the full gross dues amount as incremental revenue.

The commercial case succeeds when members receive measurably better service and the network earns additional contribution or reduces real costs. Membership counts and a large waiting list alone do not prove either outcome.

The standard for proving commercial impact

Separate mechanism from causation

There are three different questions. Can the product capture, govern, recall and act on useful information? Can those functions work within the customer's operating systems and policies? Do they cause a commercially meaningful improvement relative to the best practical alternative? The live operating evidence addresses the first question in one hospitality environment. Each new sector requires evidence for the second and third.

Public financial disclosures and transparent arithmetic can establish the size of an opportunity. They cannot establish that GuestMemoryOS causes a particular uplift. A patent filing describes an invention; it is not proof of customer return or commercial efficacy. The appropriate outcome of these studies is a controlled, costed test with an explicit decision rule.

Define the population before measuring the result

Identify eligible people, bookings or service opportunities; the metric determines the denominator. Passenger trips, room nights, unique customers, households and paying members are not interchangeable. Repeated visits by the same customer need to be recognised in the analysis. Group and brand boundaries must remain consistent.

Specify a primary financial outcome and a primary service outcome before launch. Assign treatment and a concurrent comparison where practical. Use customer or party assignment when it prevents mixed service within a group. Use property, team, vessel or venue clusters when staff behaviour can affect both arms, and allow for that clustering in the uncertainty estimate.

Compare against existing operations and tools, not an artificially weak process. Analyse all eligible units assigned to the programme, including people who do not engage. If the measured effect already includes nonparticipation, do not reduce it a second time by multiplying by an adoption factor.

Record the complete chain

For an assessed service opportunity, retain the link between the permitted source, the correct customer, the current confirmation, the action assigned, the delivery outcome and relevant feedback. A captured observation is not a completed action. A task marked complete is not automatically a successful guest outcome.

Revenue measurement needs the sale, refund, discount, fulfilment cost and any displaced purchase. Savings measurement needs net staff time and the actual expense reduced. Incremental capture, human review and error-correction work belong in the cost base. Memory that generates extra work can have negative net productivity.

Use an observation window that can answer the question

Size the test using the operator's baseline rates, spending variability, minimum valuable effect and operating clusters. A short feasibility pilot can demonstrate usable workflows; it may be too small to detect a one percentage point retention change. Report confidence intervals and economically meaningful uncertainty, rather than declaring success because a number moved upward.

Keep the comparison through the relevant booking or renewal cycle. Control for season, prices, capacity, promotions and unrelated operational changes. Replication across teams and locations is necessary before extrapolating one successful test to a global enterprise.

Calculation rules and investment decision

Rebuild every case from its inputs

The payback formula assumes steady annual benefit. An actual investment case must model the rollout ramp, seasonality, cash timing and renewal cycle. No programme price, implementation cost or guaranteed payback is supplied by these studies.

Avoid double counting

The revenue sensitivities in each case are alternative scenarios. Do not add them. Repeat-business illustrations are outside the headline operating values; where they explain a revenue uplift, they are components of that uplift rather than another benefit. Contributions are included within revenue, and the cash-realisation portion of released staff time is included within the full capacity equivalent.

If more time with customers is the reason for additional sales, do not also claim the same time as a wage reduction. If additional guests displace equally profitable demand, calculate the difference in contribution, not the full booking value. If bookings move between a group's brands, properties or channels, measure the whole relevant economic boundary.

Stress the result before expanding

Revenue-base cases scale linearly with the verified eligible share and the measured effect. Halving either halves the revenue opportunity. Conversion cases behave similarly with eligible volume and net sale value. For every case, test zero revenue uplift, no realised cash saving and additional capture or integration costs. Negative results should remain visible.

The assumed 25% cash-realisation share is a transparent sensitivity choice. It is not an industry benchmark. A programme may produce valuable service capacity with no cash saving, or may reduce specific outsourced costs more directly. Finance must reconcile realised savings to expenditure without degrading the intended service.

Expand only when the measured contribution after full costs meets the customer's investment hurdle, service outcomes are acceptable, the workflow is adopted and the effect is reproducible. Record the tested scope and uncertainty when publishing an achieved result. These opportunity studies can then be updated with actual evidence from that sector.

Source and assumption register

Public business disclosures

Source titles below use neutral identifiers. The linked original disclosures allow verification of the underlying benchmarks. Access date is 16 September 2026. Annual periods differ and are stated explicitly; historical bases are not forecasts.

Additional disclosures and product sources

AFG product sources: live operating evidence, platform, governance, enterprise integration and the individual industry pages linked in each case. Product evidence is company-reported; the studies do not represent it as an independent financial audit.

Research sources E1 and E2 are linked and described in the independent-evidence section. They support specific mechanisms in other settings; their effect sizes are not used as assumptions in the financial models.

Assumptions requiring customer validation

Every eligible deployment share, eligible cohort, uplift, service target, upgrade price, basket value, contribution margin, task volume, net time saving, wage rate and cash-realisation share is an analytical assumption unless specifically identified as a reported fact. Those assumptions appear beside the calculations they control.

There are no claimed achieved GuestMemoryOS revenue, retention or savings figures in this publication. Proposed customer journeys illustrate how the mechanism would operate; they are not fabricated accounts of actual customers. The public benchmark does not establish that the operator has a service deficiency.

AFG Holdings Ltd • GuestMemoryOS • 16 September 2026