Why most operators hit a 60–70 per cent automation ceiling in guest messaging

Reading Time: 7 minutes

Worldwide: [Sponsored content] AI-powered platform for property managers HelloHost discusses discusses why guest messaging automation typically stalls at 60–70 per cent and how newer multi-agent systems are helping operators move closer to 90 per cent by automating routine inbox communications.

AI-powered guest messaging is now embedded in the short-term rental industry. PMS providers are rolling out their own assistants, from “AI Replies” to “Auto Replies”, offering suggested responses to guest messages. Meanwhile, a list of third-party tools continues to grow, helping guest experience teams respond faster.

Yet despite widespread adoption and inbox automation reaching 60-70%, many operators say the same thing: the inbox still consumes hours every day. Routine questions may be faster to handle, but teams still monitor messages closely for edge cases. In a majority of cases, a human needs to review, edit and click “send”.

That’s where a smaller set of operators are becoming the exception. They’re pushing beyond “AI that drafts” and rules-based logic to true autopilot: an AI that can take ownership of guest replies, decide when not to respond, double-check its sources, revise messaging when needed, and escalate only the situations that require human judgment. HelloHost’s Autopilot is one example of this approach. After years of mapping and resolving the edge cases, its users say they’re now truly hands-free from guest messaging and lengthy AI setup is no longer required.

The promise (and limits) of early AI guest messaging

Guest messaging has always been one of the first operational areas targeted for automation. It was repetitive, time-sensitive, and measurable. Early tools were designed to generate suggested replies for common questions we get all the time: WiFi details, check-in times, parking instructions, and amenity confirmations.

This approach delivered immediate ROI. Routine questions could be answered in seconds rather than minutes or hours. Staff no longer had to type the same responses repeatedly. For smaller, family-run types of portfolios, this alone represented meaningful time savings (during daytime, that is).

But with the portfolio of a scaled PM, a new pattern emerged. While AI could draft responses, humans still had to supervise the inbox. Messages needed to be reviewed before sending. Edge cases still required manual handling. And operators (including virtual assistants) remained responsible for ensuring tone, accuracy, brand alignment, and compliance with internal policies.

So, in practice: AI reduced typing, but it didn’t really address the attention divide, nor the responsibility behind guest messaging.

Why most operators are stalling at 60-70% automation

Today, many professional operators report that roughly 60–70% of guest messages can be automated. The remaining 30–40% are where time and attention concentrate.

But this part isn’t made of rare or extreme scenarios. They are actually everyday situations:

• An early check in or late check out request that requires checking the calendar with a negotiable fee
• A smart lock with a dead battery and a guest who’s locked out
• A guest verification process that fails to work
• A complaint that requires empathy rather than information

Each case may be manageable on its own. But together, at scale, they keep guest-facing teams anchored to the inbox.

This is perhaps why inbox automation often feels limited. Monitoring remains constant because the cost of missing an important message is high. A delayed response can escalate into operational disruption, if not into refunds and negative reviews.

In these cases, the AI isn’t “learning” how to handle the situation, which means that a team member will keep needing to jump in and edit an answer before pressing “send”, tomorrow as much as today.

From “AI replies” to true autopilot

This automation ceiling that many face comes down to how early AI systems were designed.

“In the very beginning, AI guest messaging was kind of just tapping a generative AI model for an answer. You get the guest message, you send it to ChatGPT along with your data, and it comes back with a response,” says Annie Li, co-founder of HelloHost.

This approach works well for isolated questions. This is how most platforms “suggest” a draft response for approval. It is still largely a “rules-based” system, with very little effort going into the accuracy and quality of the message. It will share the code to the parking garage when the guest asks about parking, without checking the reservation dates.

Operators are still the gatekeepers to messaging. But hospitality messaging is rarely isolated. Context matters. Policies matter. Brand tone matters. And so does knowing when not to respond automatically or when not to make things up – which AI did, a lot.

“Now you have multi-agent systems that work together, that can check multiple data sources, synthesise them and quality-assure their own work. These systems are capable of producing accurate, high-standard responses on their own,” Li adds.

That shift, from a single response generated by one AI agent to coordinated decision-making between multiple AI agents for a message sent without human review, is what we call “autopilot”. The term is often used for marketing purposes, but guest messaging is truly on “autopilot” only when teams no longer need to review every message within minutes of it coming in.

This is when AI can negotiate a fee, reject an early check-in based on same-day turnover, help a guest regain access during a lock-out, or flag situations that require human input, then correctly communicate that decision back to the guest.

The new standard for guest messaging automation

The next phase of guest messaging automation is not about writing better replies, nor about handling routine inquiries. It is about ownership of more guest communications and operational scenarios by AI, with humans reviewing only the most value-added opportunities and critical issues, faster and better.

To do so, instead of suggesting responses, some systems can now take responsibility for sending messages directly within the PMS inbox. They assess the situation and determine whether AI needs to intervene and, if so, generate a response and act. They continuously evaluate whether escalation to a human is required.

Behind the scenes, this often involves multiple checks for accuracy, policy alignment and tone, allowing the AI to revise its own draft, verify references, criticise itself and improve multiple times, before ultimately choosing whether to escalate or respond.

In this model:

• All routine messages are handled automatically
• Far more edge cases are handled automatically
• Escalations, when they do happen, are channelled internally through existing tools such as Slack or WhatsApp, with teams notified only when judgement or action is required, supported by AI-provided context and suggested actions
• Teams speak directly with an AI assistant to handle escalations
• The AI continuously learns from these interactions, enabling it to handle more scenarios over time in a way that aligns with how operators would manage them themselves

By all means, 90% inbox automation is not an easy standard to achieve. It requires purpose-built tools focused on guest messaging that are also easy to use without hiring in-house AI specialists.

HelloHost, for example, differentiates itself from single-AI setups within PMS platforms, as well as broader AI systems that attempt to manage everything from upselling to financial workflows, or require extensive configuration and testing before becoming useful.

In deeper guest messaging workflows, users now call an AI agent inside WhatsApp or Slack threads to pull in external context and carry out actions such as resolving maintenance requests.

Operator perspectives on multi-agent AI (small and large portfolios)

It helps to understand the impact of full inbox ownership by looking at how it is used across different portfolio sizes.

Smaller, family-run operations

For smaller teams and hosts, time reclaimed from guest messaging often translates directly into better hospitality, and removes the mental load of being online 24/7.

“It takes away all of the low-end work that our team was spending most of their time on. We now have 90% of our guest messaging automated, with inbox triggers in the PMS and HelloHost replying to our guests for us,” says Michael Chang, founder and CEO of TrustBNB.

Beyond speedy replies, the shift is in how teams reallocate their attention. “Our internal team is spending five times more time on bigger problems and making a real difference for our guests,” says Chang. For these operators, automation is less about scale and more about sustainability and avoiding burnout.

Large property managers at scale

At larger portfolio sizes, the economics become unavoidable.

“Before this, I had 14 virtual assistants working around the clock just to handle guest messages. They had to be available full-time, and we had to constantly check up on them,” says Humza Zafar, CEO of Stay Luxora.

For larger operators, the value lies not only in cost reduction but also in consistency in customer service standards. AI does not rotate shifts, miss messages, lose context, or get mixed up across properties. Rather than replacing guest experience teams, it helps them declutter daily tasks and focus on higher-value hospitality work.

Why setup time has become the hidden barrier

One reason many operators fail to reach higher levels of inbox automation is setup friction. Imagine you’ve been using ChatGPT for three years and someone asks you to delete all your data and start again. Would you do it so easily and quickly?

In the same way, most AI messaging solutions may require weeks of configuration: building workflows, training models, defining intents and maintaining rule sets. While powerful, these systems demand ongoing attention and technical ownership. For this reason, property management companies with more than 1,000 listings now hire technical AI experts in-house.

HelloHost’s AI Autopilot is designed to go live with near-zero setup, typically after a short onboarding conversation. Rather than asking operators to train the system extensively, it is built to operate immediately using existing data sources, including months of inbox conversations, unit or regional policies, and team training materials such as SOPs.

This emphasis on time to value reflects a broader shift in hospitality technology. As teams become leaner and more distributed, tools that require constant tuning are increasingly difficult to sustain.

What this means for hospitality teams

As the short-term rental industry matures, guest expectations continue to rise. Response speed is now assumed and expected. Empathy and professionalism are non-negotiable, let alone accuracy.

AI tools can support this, but only when they reduce cognitive load for teams rather than adding to it. With guest messaging, when automation is implemented effectively, teams can stop reacting and start prioritising. They intervene less frequently but more meaningfully. Instead of scanning inboxes, they focus on adding a hospitality touch, resolving issues, supporting owners and improving operations.

The difference is significant. AI changes the way teams work, but it does not replace hospitality professionals. It creates the conditions for them to do their best work.

Highlights / Key takeaways

• AI guest messaging adoption is widespread, but most operators remain stuck at 60–70% automation when they use AI drafts or simple rules to respond.
• Multi-agent AI systems now enable fuller inbox automation and ownership, with operators automating up to 80–90%.
• HelloHost focuses exclusively on guest messaging autopilot, requiring minimal setup and escalating only when needed. You can start a free 30-day trial here.