In Park City and beyond, two Utah firms are moving past policy debates to deploy AI directly inside businesses — and the early results are measurable. MyAdvice has built Maya, bespoke small language models for individual medical and legal practices, while GCommerce is pushing hotel inventory into AI channels to reclaim direct bookings. Deployments report conversion rates roughly tripling and review-response times falling from more than 20 days to under 24 hours, signaling operational shifts rather than theoretical change. These pilots highlight both practical benefits and persistent trust questions for staff, clients and customers.
Key Takeaways
- MyAdvice (Park City) has built Maya, custom small language models trained on each client; the firm serves more than 1,000 health-care and legal practices.
- When practices switch their website chat to Maya, conversion rates have roughly tripled compared with previous chat setups.
- Average review-response time across MyAdvice clients fell from over 20 days to under 24 hours; nearly 100% of reviews now receive a response.
- GCommerce (24 years in hospitality) manages online presence for about 2,500 properties and is integrating hotel inventory into AI platforms to avoid third-party commission channels.
- Industry studies note long planning journeys for travelers — an average of 141 pages visited over 45 days — and 66% of travelers report disliking the planning process, creating an opening for AI-driven booking experiences.
Background
Conversation about AI in Utah has been dominated by legislators and regulators debating transparency, accountability and risk management. Those policy discussions matter, but they are unfolding in parallel with practical deployments by local companies experimenting with applied AI inside everyday business workflows. The contrast matters: regulating an emerging technology differs from observing it operating in customer service queues and booking engines.
Historically, generic, large-scale chatbots have struggled with hallucinations — producing plausible but incorrect answers — which undermines trust for businesses that must protect reputations and comply with rules. One response has been to train smaller, domain-specific models on verified, local data to reduce error rates and align outputs with a brand’s voice. That trade-off — narrowly scoped accuracy versus broad generality — is driving different implementation choices in health care and hospitality.
Main Event
Shawn Miele’s MyAdvice built Maya as an in-house platform that produces a separate, custom small language model for each client practice. Rather than routing questions through a single generic model, each deployment is tuned to the practice’s services, tone and verified information. Miele says that tailoring reduces hallucinations and yields consistent quality levels compared with mixed human-plus-AI chains of work.
The operational effects are immediate for some clients: websites using Maya report conversion rates roughly three times higher than before, and review management that previously lagged by weeks now responds in under 24 hours with nearly complete coverage. For clinicians and small-business owners who previously answered messages between appointments, this changes daily workflows and frees staff for higher-value tasks like patient care.
In hospitality, Scott Van Hartesvelt’s GCommerce is addressing a structural challenge: when travelers use AI assistants to plan trips, those assistants commonly direct bookings through third-party platforms such as Expedia or Booking.com, which capture commissions and the customer relationship. Van Hartesvelt frames this as a repeat of an earlier disruption after Sept. 11, 2001, when aggregators inserted themselves between hotels and guests.
GCommerce’s approach is to push hotel inventory and pricing directly into AI systems using emerging data protocols so that an AI assistant can present a hotel’s live availability rather than steering the customer to a commission-bearing aggregator. The goal is to restore the direct hotel–guest relationship and retain margin that third-party sites currently extract.
Analysis & Implications
These Utah pilots illustrate a broader strategic choice for businesses: adopt narrow, verified AI models that reduce factual errors and protect brand voice, or rely on generic models that are more flexible but more prone to hallucination. For service providers such as dentists and hotels, factual accuracy and tone are core to customer trust; errors carry reputational and potentially legal risk.
Operational gains reported by MyAdvice — tripled conversion and near‑instant review response — translate into measurable revenue and time savings. If clinicians can see more patients because administrative tasks are automated, practice throughput and patient access both improve. For smaller practices, these are not incremental workflow tweaks but structural changes to job design and resource allocation.
For hotels, reclaiming direct bookings could shift billions in annual commission flows if data protocols and AI integrations scale. However, achieving that requires technical standards, cooperation from major AI platforms, and robust mechanisms to ensure pricing parity and inventory integrity. It also raises new questions about data governance: who controls the feeds, how are cancellations handled, and what transparency is offered to consumers?
Comparison & Data
| Metric | Pre-AI (MyAdvice clients) | Post-AI (Maya) |
|---|---|---|
| Website conversion | Baseline | ~3× baseline |
| Average review-response time | >20 days | <24 hours |
| Share of reviews replied | Variable | ~100% |
The table contrasts key operational metrics reported by MyAdvice before and after Maya deployments. Separately, industry research cited by local firms indicates travelers visit an average of 141 pages over 45 days before booking, and 66% of travelers dislike planning — statistics that create demand for streamlined AI-driven booking flows. These figures help explain why hotels are racing to insert themselves into the emerging AI ecosystem rather than ceding the space to aggregators or general-purpose assistants.
Reactions & Quotes
“AI is both our biggest opportunity and an existential threat; we chose to build the capability in-house and train models per client to protect accuracy and brand voice.”
Shawn Miele, MyAdvice (digital marketing for health and legal practices)
Miele emphasized control and trust as primary motivations for in-house development. He framed the decision as risk management: a narrowly trained model reduces false or misleading answers that could harm a practice’s reputation or confuse patients.
“This shift is faster and more disruptive than anything I’ve seen in 24 years. If hotels don’t get inventory into AI, third parties will keep taking the customer relationship.”
Scott Van Hartesvelt, GCommerce (hospitality marketing)
Van Hartesvelt warned that without direct integrations, hotels risk repeating past disintermediation. His firm is working on technical pipelines to feed real-time availability into AI platforms so hotels can appear in assistant-driven bookings with their own terms and pricing.
Unconfirmed
- Long-term durability of the reported conversion lift across diverse client types has not been independently verified beyond initial deployments.
- It remains unproven whether AI-driven direct inventory feeds will fully eliminate booking commissions from third-party platforms at scale.
- Potential downstream effects on staff roles and employment at scale are plausible but not yet measured or quantified.
Bottom Line
MyAdvice and GCommerce demonstrate that the AI debate is not only legislative; it is operational. Tailored, small models and direct inventory integrations can deliver measurable gains in conversion, responsiveness and customer ownership, but they also surface trust, governance and labor questions that policy-makers and business leaders must address.
The businesses that combine technical controls, transparent disclosures and workforce planning are likeliest to capture durable advantage. For Utah — a state with deep technology adoption and an emerging AI ecosystem — these pilots could become a template for how local firms scale trustworthy, business-focused AI.