A few days ago, Evolve told Skift that its AI is now deflecting 60% of guest inquiries before a human ever sees them. The company’s chief product officer described two years of moving AI from a pilot into core infrastructure, and Evolve cited their owners earning 18% more revenue and booking 9% more nights than the market average.
It is, on its face, an impressive number.
It’s also a useful moment to be very clear about how HostStarter is different — because we use a lot of AI too, and we still don’t deflect a single guest message away from a human.
Here’s why that’s a feature, not a gap.
What “60% inquiry deflection” actually means
Inquiry deflection is the industry’s polite term for “the guest typed a question and a bot answered it.” The guest doesn’t know — that’s the point. If the bot’s answer was good enough, the conversation ends there. If it wasn’t, the guest either re-asks, gives up, or escalates to a human.
For a property manager, deflection looks great on a dashboard. Fewer humans needed per booking. Lower cost per conversation. A line on an investor deck.
For an Airbnb host, the relevant question is different: what did the guest experience on the other side?
When that 60% figure includes:
- A guest who can’t find the WiFi password at 11pm and gets a recycled FAQ instead of someone walking them through it,
- A guest who has a question that seems like a standard question but is actually a sign of a real problem (a leak, a noise complaint, a missing item),
- A guest who is about to leave a 4-star review and could have been turned into a 5-star review with one warm reply…
…then “deflection” can be the same thing as “missed opportunity to keep this guest happy.”
We’re not anti-AI. We’re anti-confusing-the-host’s-revenue-with-the-manager’s-margin.
What HostStarter does with AI
It is genuinely 2026 and we use AI every day. Specifically:
- Drafting replies, not sending them. Our team uses AI to draft a response to every guest message. A real HostStarter teammate reads it, edits it, and sends it. We get the speed of AI with the judgment of a person who actually knows your property.
- Pricing. Dynamic pricing has been AI-assisted for years; we use it for every listing and tune it weekly.
- Listing optimization. We use AI to surface gaps in listing copy, photo coverage, and amenity tags, then a human writes the actual copy.
- Review analysis. AI flags themes in guest reviews so we can address recurring issues quickly.
- Lead routing. AI helps us route prospect questions to the right specialist on our team.
What we don’t do is hand a guest conversation to a model and walk away.
The 4 conversations where humans still win
In our experience managing properties across 33 markets, four kinds of guest conversations almost always pay for themselves several times over — and AI gets all four of them wrong in the same predictable ways.
1. The conversation that determines a 5-star vs 4-star review.
A guest pings you at 7pm saying the bedroom feels a little warm. The bot answers: “Our thermostat is set to 72°F for energy efficiency. You can adjust it on the wall panel near the entry. Have a great stay!” — and the guest now feels managed.
A person on our team answers: “Hey [name], so sorry about that — we’ll get it down right now. Two things while you’re settling in — there’s a fan in the hall closet if you want it tonight, and the upstairs windows do open if you want some cross-breeze in the morning. Anything else we can do?”
Same problem. Wildly different review.
2. The conversation that’s actually a problem in disguise.
“Just checking — is the front door supposed to stick?” is not a question. It’s a maintenance ticket. A human catches it in two seconds; a model trained on FAQs answers it as if it were a question.
3. The conversation that turns a one-time stay into a repeat booking.
The single highest-ROI sentence in property management is the one a guest gets when they say something nice about the property. That sentence cannot come from a bot. It has to come from someone who can say, “Glad you loved the kitchen — the owner is actually here in town and wants you to know about [the new coffee shop / the next listing they’re opening / a friends-and-family rate for your next trip].”
4. The conversation that prevents an Airbnb refund.
The single fastest way to lose a host’s money is for a guest complaint to escalate into an Airbnb ResolutionCenter case before anyone has tried to fix it warmly. Bots are statistically excellent at escalating complaints into formal disputes.
The number Evolve didn’t quote
Evolve cited 18% higher revenue and 9% more booked nights vs. the market average. That’s good — but the relevant comparison isn’t “Evolve owners vs. the average Airbnb host.” It’s “owners with a full-service manager vs. the average Airbnb host,” which most full-service managers can claim.
The real question for a host considering Evolve, Vacasa, Awning, or HostStarter is the net number: revenue minus management fees minus refunds minus replacement costs from unhappy guests.
That’s where the “human + AI” model has its edge: the same AI that helps everyone draft faster messages also lets a smaller, more accountable team handle the same volume — without the margin pressure that pushes other companies to lean on deflection.
That’s also why we charge a flat 12.5% rather than the 25–35% (or 35–45% with add-ons) you’ll see elsewhere.
What “human + AI” looks like on a 4-night stay
Here’s an anonymized real example from this month — a 4-night booking at one of our Dallas properties:
- Pre-arrival, T-48h. AI drafted check-in instructions; teammate personalized them with a note about the storm forecast and added an offer to leave the porch light on. Guest replied “thank you, please do!” Relationship started.
- Day 1, 9pm. Guest reported the dishwasher button wasn’t responding. AI flagged this as a service ticket; teammate dispatched our local handyman the next morning and sent the guest a $25 DoorDash code for the inconvenience.
- Day 3, 6am. Guest asked about late check-out. AI surfaced that the next booking didn’t start until 5pm, so a teammate offered a 2-hour late check-out for free.
- Post-stay, +6 hours. 5-star review. Guest specifically called out the late check-out and the handyman’s same-day fix.
Pure-AI deflection would have answered “check-out is 11am” and “please contact us if maintenance is needed.” Same operations, half the revenue impact.
How to evaluate a property manager in 2026
If you’re evaluating Evolve, Vacasa, Awning, or HostStarter (or any of the 40 other companies pitching you), here are the four questions worth asking:
- Who reads my guest messages? If the answer is “our AI handles routine inquiries automatically,” ask what counts as routine — and what your refund rate is when something gets miscategorized.
- What does a real review cycle look like? Ask to see two recent 5-star and one recent 4-star review on managed properties. The patterns are very visible.
- What’s the total fee, including add-ons? Industry average is 25–35% before add-ons, 35–45%+ with them. Anything bundled into “platform fees” or “guest fees” still comes out of your nightly rate.
- What happens when something goes wrong on a Saturday night? This is the entire job. Ask for a specific example from the last 30 days.
The HostStarter take
We think AI is one of the best tools to land in this industry in a decade. We use it everywhere. We just don’t think the right place to use it is as a substitute for a person reading your guest’s message.
If you want a manager that uses AI to make humans faster — instead of using AI to remove humans from the loop — we’d love to talk. Twelve and a half percent. Month-to-month. Every guest message read by a person.
If you’re currently with Evolve and curious about switching, we put together a direct comparison here. If you’re with Vacasa, start here.