What the paper studied
Research on smart tourism destinations and data-driven service innovation tends to assume abundant data, mature infrastructure and capable institutions, conditions that emerging destinations rarely meet. This review asks how data-driven service innovation supported by AI can help emerging destinations move toward becoming smart destinations when data is scarce.
The authors ran a PRISMA 2020 systematic review of records exported from Scopus, Web of Science and ScienceDirect, which produced a corpus of 219 papers. They analyzed the corpus with bibliometric indicators, VOSviewer, Bibliometrix and dictionary-based content coding, and added a documentary analysis of Riohacha, Colombia as an illustrative case.
Key findings
- Research output in the corpus is growing at 18.7% per year.
- Two bodies of work develop largely in parallel: 43 studies link data and AI to service innovation, and 25 study innovation under scarcity.
- Only seven studies combine both, a co-occurrence close to what would be expected if the two topics were independent (lift = 0.96).
- None of those seven studies adopts the destination scale.
The proposed framework
The authors propose a layered model that takes scarcity as its starting point. It has five dimensions:
- Scarcity conditions
- Collaborative data governance
- The minimum viable data ecosystem
- Frugal and augmentative AI capabilities
- Co-created smart services
Inclusive co-creation and a learning loop cut across these dimensions. Seven design principles guide how the model is put into practice: data sufficiency, source bricolage, frugal and augmentative AI, collaborative data governance, inclusive and culturally relevant co-creation, incremental scaling and continuous learning.
Why it matters for hospitality
Much smart destination thinking starts from conditions that many hospitality markets do not have. This review puts numbers on that gap and offers a model built for limited data. For hotels, tour operators and destination organizations in emerging markets, the message is that waiting for complete data is not the only path: useful AI-supported services can begin with sufficient data, shared governance and small steps that scale.
Practical takeaways
- Start from the data the destination actually has, and decide how much is sufficient for the service in question.
- Combine the sources that are available rather than waiting for one complete dataset.
- Favor frugal AI that augments people's work.
- Set up data governance collaboratively across destination stakeholders.
- Co-create services inclusively so they stay culturally relevant to the destination.
- Scale in increments, and build in a learning loop so each stage informs the next.
Limitations
The corpus comes from three databases, and dictionary-based coding depends on the terms chosen. Riohacha is an illustrative case drawn from documentary analysis rather than a field test. The abstract presents the model as a proposal and reports no implementation results.