We will name the systems.
Every firm bidding for your commercial AI programme will talk about agents, adoption and outcomes. Ask them to name the five systems the programme has to touch and who owns the integration contract for each. That is the first day of the actual work, and it is where most of these programmes quietly stall.
Commercial AI in a hotel is an integration problem wearing a model costume.
The levers a commercial programme is bought to move are ADR, RevPAR, direct mix, ancillary attach and loyalty margin. Every one of those is written in a system of record: rate and availability in the PMS and the CRS, pricing logic in the RMS, distribution in the channel manager, spend in the POS, and identity in the loyalty ledger and the guest CRM.
A greenfield data platform with an agent layer on top will produce excellent segmentation and very good outbound personalisation. That is a marketing outcome. Until the agent can write back to the system that owns the rate, the inventory or the loyalty transaction, it can recommend and it cannot transact, and the commercial P&L impact arrives a year later than the business case says.
We start from the systems and work outward.
The platforms, and what each one actually lets you do
| Layer | Platforms | What matters in an AI programme |
|---|---|---|
| PMS | Opera CloudMewsInfor HMSRMS CloudCloudbedsShiji | Where the folio, the rate plan and the reservation live. Write access, rate plan granularity and API rate limits decide what an agent can actually do. |
| CRS and distribution | SynXisAmadeusSabreWindsurfer | Two way writes are the hard part. Rate parity reconciliation is where most integrations are quietly one directional. |
| Revenue management | IDeaSDuettoAtomizePace | An agent that recommends a rate the RMS will override is theatre. The handshake has to be designed, not assumed. |
| Channel management | SiteMinderDerbysoftRate Gain | Mapping and derived rates. Where a clean model meets a messy reality. |
| POS and F&B | SimphonyInfrasysLightspeed | Ancillary attach lives here, and it is usually the worst governed data in the estate. |
| Guest CRM and loyalty | RevinateCendynEnterprise CRM platforms | Identity resolution across brands and entities. The loyalty ledger is a financial system, so it is treated like one. |
| Integration layer | HapiImpala pattern middlewareDirect APIs | Whether you own the integration or rent it decides your options in year three. |
We are not a reseller of any of these and we take no vendor commission. Where a platform is the wrong fit for your operating model, we say so, and we have.
A hotel back office
Five failure modes we see repeatedly
- The write back gap. The programme is scoped on insight and delivered as a dashboard, because nobody costed the two way integration. Fix it in scoping, not in phase two.
- Rate plan sprawl. An estate with hundreds of rate plans accumulated over a decade cannot be modelled cleanly, and no agent will fix that for you. Rationalisation is a prerequisite, and it is a commercial exercise, not a technical one.
- Guest identity that does not resolve. The same guest exists four times across brands, channels and entities. Personalisation built on unresolved identity produces confident, wrong output at scale.
- The revenue team does not change what they do. The single most common cause of a failed commercial AI programme. A recommendation that arrives in a system nobody opens on a Tuesday morning has moved the work, not removed it. This is why behavioural design is a named layer on our team rather than a change management afterthought.
- Long stay and mixed use models stretched onto transient logic. Serviced apartments, co living and extended stay run on lease like economics that a night centric PMS models with rate package gymnastics, manual journals and periodic billing workarounds. The workarounds compound at scale and they poison the data an AI programme depends on.
How a Glhip commercial AI engagement runs
Full detail is on the delivery page. In hospitality specifically, three things are fixed.
- Discovery starts in the systems, not the strategy. We ask for read access to a sample before we ask for a workshop. A rate export, a reservation extract, a loyalty schema. What is actually in the data decides what is actually possible, and that conversation is cheaper on day three than on month three.
- Every recommendation names its write path. If an agent is going to change something, we specify which system receives the change, through which interface, with what human checkpoint, before it is built.
- Adoption is designed, not hoped for. Behavioural design is one of our six layers. We design the coaching loop against the real cause of non compliance, which is usually ownership, incentive or a painful system step, not ignorance.
Who we serve
- Hotel groups and ownership companies, multi property, multi brand, where the integration surface is the hard part.
- Luxury and upper upscale operators, where guest data governance carries more weight than throughput.
- Mid scale and lifestyle, where the commercial team is small and the tooling has to work without a data function behind it.
- Serviced apartments, co living and extended stay, where the operating model is closer to leasing than to hospitality and most platforms fight you.
- Hospitality technology vendors, who need product, integration and go to market help from someone who has sat on both sides.