How a 35-Property Manager Hit 96% Automation with AI

TL;DR

• Achieved 96 percent automation rate within 18 months, starting from 80 percent out of the gate
• Zero staff reductions—same team size, elevated roles focused on upselling and business development
• Response times dropped to under one minute, compared to inconsistent human-based responses
• Enabled focus on 20 to 30 percent additional revenue through upsells and extended stays
• Improved guest ratings across multiple channels through consistent, accurate around-the-clock responses

The Challenge

After 13 years building a portfolio of 35 short-term rental properties, Erwan Le Roy faced a problem: the guest communication bottleneck was pulling him back into operations.

"It was really about answering the guest on time with the right information," Le Roy explained. "We're in an industry where it's here now. It's not like, oh, you have this question, don't worry, we'll get back to you in five hours. No, it's here now."

The pressure for instant responses was relentless. Guests expected the same immediacy they got from Uber Eats—anything less triggered complaints. Le Roy's virtual assistant team struggled to maintain consistent response times across a true around-the-clock operation.

Le Roy found himself drifting from owner to operator—and eventually, as he put it, becoming "the mom."

"I would end up paying more and more attention when they forget and then talk to them," he said. "I told my virtual assistant team, I'm not your mom. I'm not here to micromanage you."

SOPs existed, but they only delivered value when the team executed them perfectly every time. Human attention couldn't sustain the continuous vigilance required across all time zones. The team spent their days reading every email, fielding mundane questions—Wi-Fi codes, parking instructions, and information already in the listing—while higher-value work went untouched.

The Solution

Le Roy implemented Conduit approximately 18 months ago, feeding the AI with his existing Airbnb listings and SOPs. The knowledge base became the foundation for everything.

The key insight: the AI needed training like any new team member.

"It's like if you have a baby and the baby starts walking and the first time the baby falls, you're not going to say, 'Well, you're just not a walker,'" Le Roy explained. "We knew we had to train the AI."

He assigned one of his top virtual assistants to identify gaps—using Conduit's interface to see exactly what information was missing when the AI couldn't answer a question.

"When AI cannot answer, we know why. AI will not improvise things—it tells you that's what I'm missing," he said. "I had a virtual assistant focusing on providing what's missing."

The approach was systematic: identify the gap, fill it, measure the improvement, repeat.

The Transformation

Le Roy kept his entire team and elevated their roles.

"I tell my team, 'You're way too smart to give the Wi-Fi code to a guest. You're way too smart to resend the check-in instructions,'" Le Roy said. "I still have the same number of staff. I did not let go anyone. Instead, I elevated their job."

The mundane copy-paste work that had consumed the team's attention now happened automatically, around the clock, within one minute of each guest inquiry. The team's new mandate: business development and upselling.

"Now we have this freedom to test—really business development," Le Roy explained. "Every week, every month we test new things. Some of them work, we implement them, some of them we drop."

The virtual assistant assigned to train the AI became deeply invested in the metrics.

"Any A player loves to be measured. They would look at Conduit—what is my automation rate? 'Oh I'm at 82 percent. See everyone, I was at 80 percent, I'm at 82 percent, now I'm at 85,'" Le Roy recalled. "I set the goal at 90 percent. Now we're at 96 percent. They have a new goal—it's going to be 98 percent."

One of the most effective workflows Le Roy implemented was sentiment-based review requests. Conduit automatically detects guest sentiment, and the workflow is simple: if a guest is flagged as unhappy, don't request a review. If they're not unhappy, ask for a five-star review—and ask again.

"Why in the world would we ask a guest who is unhappy to submit a review? I still don't understand it," Le Roy said. "The AI measures emotion without putting human emotions in the way. It's really scientific."

The Results

• 96 percent automation rate—up from 80 percent at launch, with sights set on 98 percent
• Sub-one-minute response times, calibrated to feel human (not millisecond-fast, which guests find suspicious)
• Zero headcount reduction—team roles elevated from reactive support to proactive revenue generation
• Improved ratings across multiple channels due to consistent, accurate around-the-clock responses
• Focus shifted to upsells, which Le Roy estimates should represent 20 to 30 percent of revenue for any property manager
• Better virtual assistant retention through more engaging, strategic work

Looking Ahead

Le Roy's next frontier is voice AI, extending the same principles to phone calls and enabling guests to handle requests like late checkouts through natural conversation.

Le Roy's advice to property managers still hesitant about AI:

"Right now you don't have an option anymore. We see a divergence of hosts now: the hosts that use AI and the hosts that don't use AI. If you compete human versus AI, I'm sorry, you lost the game. It's not an option anymore."

"Get started and don't expect to be perfect on day one. Put all your knowledge in it. The AI is going to know as much as you give it content. Start at 75 percent or 80 percent and then you grow from here—just teaching the AI."

"All you need is 10 seconds of courage to change your business. That's all you need."

Erwan Le Roy operates 35 short-term rental properties and has been in the property management industry for 13 years. He is based in San Francisco and has been a Conduit customer for approximately 18 months.

‍