Keller Williams Realty International (KWRI) · Case study

Building reusable AI agents with Twin

I identified Twin through independent research, initiated a relationship with its founders, and co-developed reusable AI agents for real estate work. I then helped turn those builds into teachable workflows and brought the opportunity to KW leadership.

My role: Research lead, workflow builder & vendor collaborator
Period: 2026
Status: Reusable agents shared; integration completed

Corporate headquarters · Industry & Learning. Research, prototypes, and enablement for the Keller Williams network of agents, Market Centers, and regional teams.

The starting point

Real estate agents had access to chat tools, but completing a task across research, content, and follow-up still required moving between systems. I saw an opportunity to test AI agents that could carry out that work rather than only generate an answer.

My responsibility

I led use-case research, built and tested agents, initiated vendor relationships, created setup guidance, and presented the work. Twin collaborated on the agent builds and packaging; KW’s integrations and partnerships teams handled their respective technical and commercial implementation.

Tools & methods

Twin · HeyGen API · OpusClip · Landbot · WhatsApp · Workflow testing

~57AI agents co-developed with Twin

Finding Twin and establishing the collaboration

I discovered Twin while researching agentic AI and made the first contact with its founder and co-founder. I was evaluating the opportunity in my role at KWRI headquarters.

I investigated its hosted browser, triggers, scheduling, and ability to run without the user keeping a device on. Working directly with Twin gave me access to test practical use cases and refine the builds with the people developing the platform.

Building and testing real estate workflows

I researched the use cases and tested each tool’s capabilities before connecting the tools into an agent. The work with Twin grew to approximately 57 agents across multiple tasks. Two workflows became recurring examples in my demonstrations and training.

Local market research to a social video

I built a Twin agent to gather local news and neighborhood information and produce a market-update script. The script then passed to HeyGen, using a selected avatar and cloned voice, and on to OpusClip for transcription, highlights, thumbnails, scoring, and social distribution.

I tested the individual functions and the complete sequence. Early avatar and voice results required new recordings and repeated checks; quality improved through those iterations and a HeyGen platform update.

Market Update Video Creator agent card describing local market research, a sourced script, and HeyGen API setup. — The Market Update Video Creator card packages local market research, script development, and HeyGen API setup as a workflow teams can inspect. Inspect this agent

WhatsApp lead qualification and follow-up

This was a separate workflow. I contacted Landbot, explained the use case, and began building and testing the qualification flow myself. A Landbot engineer then partnered with me to refine it.

The demonstration connected WhatsApp intake and qualification with Twin property matching, drafted follow-up, and booking support. I worked on making the flow repeatable and understandable enough to teach to other agents.

Making the agents repeatable and teachable

I partnered with Twin to package agents developed from my research into reusable tiles. API connections were a recurring setup difficulty for agents, so I wrote a one-page connection guide and arranged a dedicated Twin representative who could help over Zoom or email.

I taught selected workflows and how to build an agent to Gary Keller Mastermind groups and KW Worldwide Mastermind groups. The training included the workflow, its setup, and where to get help.

I also collaborated with Twin on a private agent library for Keller Williams teams. It brings workflows together as agent cards, with descriptions, connected tools, and recorded runs that teams can inspect.

KW × Twin agent library showing workflow cards for content strategy, listing monitoring, and seller lead research. — The private KW × Twin agent library, captured October 2026. The 72 entries shown are the library's current total; approximately 57 agents were co-developed through my collaboration with Twin. View full size

Bringing the platform opportunity to leadership

I worked with Twin on a KW-specific interface for evaluation and drove its presentation to Jason Abrams and Gary Keller. Jason, Gary, and I had test accounts. The dedicated KWxTwin interface remained a prototype.

I also connected Twin with KW’s integrations and partnerships departments. Those introductions and my recommendations advanced the separate Command integration and sponsorship agreement.

Outcome

Reusable agents and a completed vendor relationship

Approximately 57 agents were co-developed with Twin, and reusable workflows were packaged, shared, and taught. A private KW × Twin agent library is available on Twin’s website. The Command integration and sponsorship were completed and reported active in October 2026. The earlier dedicated KWxTwin platform proposal remained a prototype.

View the KW × Twin library (work email required)

Next project · Keller Williams Realty International (KWRI)AI tool research and vendor evaluation

Let’s connect

What could work better?

I’m interested in AI innovation and solutions roles where research becomes useful systems.

Start a conversation