EliseAI builds for customers while scaling support with Assembled AI

EliseAI’s all-in-one property management automation platform and conversational AI agents handle the entire renter’s journey — from first leasing inquiry to renewal. The platform can answer prospect and resident questions, book tours, unlock doors for AI-guided property visits, and handle maintenance requests — all without human intervention.

For hundreds of property management companies, EliseAI acts as a tireless, 24/7 team member.

The challenge

When Margaret Kim joined EliseAI on the Strategy & Operations team, she found herself working at an AI company without AI-powered support. With 16+ years of experience scaling operations at companies like Goldman Sachs and Ellevest, she quickly identified the gap — and the obstacles:

• High complexity tickets: Fewer than 30% of EliseAI’s customer inquiries were simple, Tier 1 support cases. As a complex B2B platform, most tickets require looping in stakeholders, backend changes, or engineering support.
• Minimal voice-of-customer visibility: Different teams were gathering rich customer feedback via support tickets, client calls, and surveys, but there was no efficient way to capture and analyze trends and share them with product and engineering. 
• Manual triage and data lookup: Staff jumped between multiple systems to find customer context, classify ticket types, and route tickets to the right team.
• Valuable time spent on administrative tasks: Team members were spending hours drafting responses and writing interaction summaries for handoffs and analysis.

For EliseAI, delivering exceptional, concierge-level support isn't just about operational efficiency — it’s critical to customers' success. When property managers can't quickly resolve issues, it directly impacts their ability to serve their residents and prospects effectively. This made enhancing EliseAI’s support operation not just an internal priority, but a competitive necessity.

Why a company with their own AI agents chose Assembled

EliseAI knew that building AI agents for support in-house would divert resources that were better spent building their core product. With that insight, the company launched a vendor search, starting with 10 platforms and narrowing to Ada, Fin, Zendesk AI, and Assembled.

Their evaluation process was deliberately focused:

1. Take real tickets from EliseAI’s backlog
2. Feed them into each platform with no special workflows or training
3. Measure how well the platform generated accurate answers from existing help content

This test revealed which platforms had the best, out-of-the-box generative responses — versus those dependent on extensive configuration.

Three factors made Assembled stand out:

• Superior generative AI that wasn’t dependent on a perfect knowledge base or specific workflows — critical for EliseAI’s rapid release cadence
• Complex case handling for high-touch, technical Tier 2 cases
• Partnership approach that felt collaborative from day one

"The thing we loved about Assembled when we initially met the team is that we felt that they would really partner with us and build with us. And that has definitely proven to be the case,” Margaret reflected.

Enter AI support agents

EliseAI rolled out Assembled gradually, starting with Assembled’s AI Chat Agent and AI Copilot, then expanding to AI Email Agent.

While some teams hesitate to deploy AI agents without a “perfect” knowledge base, EliseAI proved it’s not a prerequisite for building a successful AI system. Satisfied with the baseline responses Assembled AI agents delivered out-of-the-box, EliseAI launched AI chat quickly. The team then used the Assembled QA review feature to evaluate reply accuracy and style on every ticket to rapidly improve accuracy. "We QA every single chat and automated email response in Assembled." Margaret explained. "When we first launched, the accuracy was around 40%. Today we are consistently in the 90s.”

Within 3 months:

• Accuracy jumped from the low 40s to 90s
• AI powered chat achieved 70% containment rate, resolving most issues immediately
• QA reviews exposed product and documentation gaps

Just as importantly, chat and email interactions became a diagnostic tool. "We use customer feedback as a roadmap of what we need to improve in our product and in our help center,” Margaret noted. This feedback loop had a compounding effect: Every time they addressed a gap, it not only improved AI accuracy but also reduced future inbound volume by proactively answering questions before they became tickets.

Automated data lookup, classification, and routing 

Before Assembled, triage was a tedious, multi-step process. Agents manually searched for the property name, cross-referenced it in a separate internal tool to find the customer’s relevant system details (e.g., integrations), manually classified the ticket after a thorough review, and then assigned it to the right specialist or team queue. 

Now, with Assembled, the customer name, property, and system details automatically populate in Zendesk when a ticket comes in. AI-driven case classification analyzes the ticket content and determines the urgency, sentiment, and category of inquiry (e.g., user education vs. billing). Automated routing then sends the ticket to the correct team or agent without human intervention.

The result? About 50% of triage is automated, reducing human error and freeing agents up for more complex cases.

Copilot: From “nice-to-have” to essential

The Assembled Copilot is an AI assistant that helps agents draft responses, summarize cases, and handle administrative tasks directly within their help desk platform. At first, Copilot was an optional add-on. The team described it as a “nice-to-have.”  Some agents resisted, feeling the tone “didn’t sound like them.” But many were already pasting tickets into ChatGPT to polish responses — an extra step that took them out of Zendesk and created inefficiencies.

Copilot brought that workflow, along with even more useful features, in-house:

• Compose from notes: Agents drop in bullet points, and Copilot turns them into polished, on-brand responses.
• Reply polisher Agents can improve the tone, clarity and professionalism of drafted responses (that can still “sound like them”) with one-click, without starting from scratch. 
• Generative wrap-ups: Agents can auto-draft a summary of their customer interaction, with key customer information auto populated for seamless handoffs and internal documentation. 

When wrap-ups became required for root cause analysis, adoption soared.

Partnership that delivers

What distinguishes Assembled isn't just technology — it's the partnership that drives continuous improvement. "In our weeklies, I’d bring a long list of feedback and the team was always receptive and even thankful for the insights, and it never felt like lip service — they delivered,” Margaret shared. “Sometimes I’d mention something in passing, and the next day it was in the product."

From UX tweaks to roadmap alignment, EliseAI and Assembled now operate as strategic partners, not just vendor and customer.

Ready to see what’s possible?

Even for a high-complexity, high-touch B2B system, AI can deliver measurable impact — when paired with the right platform and partnership. EliseAI’s journey proves that AI isn’t magic; it’s a disciplined, iterative process that, when done right, transforms operations.

If you're looking to elevate your support operation, Assembled gives you everything you need to deliver superhuman support at any scale.

Want to see how? Set up time with us to learn more.