Launching 0→1 AI experience that transforms real estate agent workflows

ANYWHERE REAL ESTATE INC.

AT-A-GLANCE

I joined a team of 2 to design Ava AI, Agent Workplace's first AI experience built to accelerate real estate agent workflows. The solution supported document analysis, internal knowledge retrieval, and automation of CRM and marketing tasks.

Previously, agents wasted time navigating fragmented systems, switching between tools, and manually interpreting information before taking action.

ROLE & TEAM

Product Designer, working with 1 UXR, 1 PM, 4 Engineers

SKILLS

Interaction Design
User Research
Competitive Analysis
Product Strategy

TIMELINE

Oct 2025 - Feb 2026

IMPACT

👥

24%

Intended users adopted Team sharing in 1st month

⏱️

32%↓

Wasted time for coordination, saving teams 3–5 hrs/week

🧱

6 teams

Aligned by this foundation for future Teams work

IMPACT

👥

30%

User adoption within initial rollout

⏱️

74%

Positive user rating, through in-product feedback

🧱

2026 Honoree

Recognized by HousingWire for AI innovation in real estate

🚀

30%

User adoption within initial rollout

👍

74%

Positive user rating, through in-product feedback

🏆

2026 Honoree

Recognized by HousingWire for AI innovation in real estate

USER PROBLEM

The need to bridge information and execution

Real estate agents juggle tons of data from CRM, transaction documents, analytics dashboards, and internal knowledge across disconnected tools. They manually interpret contracts, search for policies, and draft repetitive follow-ups. This constant context switching increases cognitive load and slows execution.

What if AI could connect the dots and turn information into action instantly?

THE SOLUTION

AI bridges the gap between information and execution.

1/

Context-aware smart prompts

AI suggests relevant prompts based on the client, listing stage, and page context so agents do not have to figure out what to ask.

2/

One-click client follow-ups

Draft emails or texts with AI, edit them using quick shortcuts, and AI would autofill your fields to send instantly without leaving the page.

3/

Quick retrieval of internal enterprise database

AI pulls trusted internal guides, policies, and resources instantly so agents do not have to search across multiple systems.

BACKGROUND

Agents are always on-the-go and constantly context-switching

On mobile, they check to-dos, milestones, reply to clients, teammates, and communicate with legal partners
On desktop, they review contracts, send out transactions, analyze listings, and prepare marketing materials

DESPITE POWERFUL TOOLS, AGENTS FACE PAIN POINTS LIKE

🧠

#1

Cognitive load remembering info

🐢

#2

Slow execution switching tabs

🤖

#3

AI disconnected from context

Real estate is built on user trust and time sensitive. Delays impact deals and client trust

That's why AI could not be another tab. It needed to live where agents already work. Ava AI will support both quick mobile interactions and deeper desktop tasks. I then mapped out the AI entry points that embed intelligence directly into the workflow.

Earlier discovery has already revealed that collaboration needs to be lightweight, while team management need to be durable and governed. Treating both as one feature would be overly complicated.

After posing pros and cons to PM and Engineers, we decided on building both Global and Contextual AI.

Earlier discovery has already revealed that collaboration needs to be lightweight, while team management need to be durable and governed. Treating both as one feature would be overly complicated.

So how did I got here? It all started with…

DEFINING USE CASES

When and how would AI meaningfully reduce friction?

To avoid building isolated features, we anchored discovery around this one simple question. Through interviews with 5 users, I identified recurring tasks that clustered into 5 core phases of the home buying/selling journey. These phases reflected how agents naturally structure their work and think about their day-to-day workflow.

Earlier discovery has already revealed that collaboration needs to be lightweight, while team management need to be durable and governed. Treating both as one feature would be overly complicated.

Prospecting

Lead nurturing

Marketing

Transactions

Support

To surface the potential of our AI assistant for new users, I translated these workflow phases into suggested prompts, giving agents clear starting points and reducing the hesitation of a blank chat box.

Earlier discovery has already revealed that collaboration needs to be lightweight, while team management need to be durable and governed. Treating both as one feature would be overly complicated.

design ITERATIONS

We tested the design quickly and iterated on feedback

"These prompts feel too generic. This doesn’t feel relevant to what I’m working on right now.” :(

5/5 users

1/ Contextual > Popular prompts

Agents wanted prompts that felt specific to their listing, their client, and their current task without setting up prompts repeatedly. I partnered with engineering to explore whether the AI could detect page-level context, allowing AI to dynamically generate relevant, personalized prompts.

Global AI still shows groups of questions in phases, with questions more personalized.

I explored a couple of contextual AI ideas on each page would cater to the context of that specific lead.

2/ Drafting follow-up is the most frequent task with AI

User often open up external AI chat to draft message, refine couple times, before copy, switch tabs to email/message to send their reply. I learnt from existing text drafting interface from AI tools and explore ways to improve this experience further.

Referencing different version

Back, forward and page

Refine button

shortcuts

Direct to email/message

Autofill fields

IMPACT

🚀

30%

User adoption within initial rollout

👍

74%

Positive user rating, through in-product feedback

🏆

2026 Honoree

Recognized by HousingWire for AI innovation in real estate

IMPACT

The final design addressed siloed, solo workflows by introducing a streamlined team-sharing experience with clear permissions, enabling agents to collaborate safely, efficiently, and without duplicated work. In the 1st month, we achieved:

👥

24%

Intended users adopted Team sharing in 1st month

⏱️

32%↓

Wasted time for coordination, saving teams 3–5 hrs/week

🧱

6 teams

Aligned by this foundation for future Teams work

WHAT I LEARNED

🤖

Leveraging AI in collaboration and the design process

Experimenting with vibe coding showed me how AI can accelerate collaboration and reduce handoff friction. I’m excited to keep exploring how AI can meaningfully supercharge our workflows while staying grounded in real user needs.

💡

Advocating for design and growing as a product thinker

I grew more confident as a product thinker by proactively influencing the roadmap. By presenting research insights, PMs are more receptive of ideas, and convinced on the need of bigger Teams initiative.

WHAT I LEARNED & next steps

💡

Using AI to prototype AI

Designing AI is a quick, experimental experience. I leveraged AI a lot throughout the process, especially with brainstorming and prototyping to explore its capability and spot usability issues that I'd never caught in static mocks

💡

Using AI to prototype AI

🤖

Leveraging AI in collaboration and the design process

🤖

Leveraging AI in collaboration and the design process

I'm excited to continue bringing AI experience with real estate agents a step further, into voice and agentic AI, helping user getting information, carryout actions seamlessly without touching the phone.

You've reached the end… how about another story?

Designing for navigation and a11y at scale for Call of Duty players

ACTIVISION BLIZZARD / 2024