Case Studies
May 2, 2024

Rentgrata's Test Driven Journey to a Production-Ready Chatbot

Anita Kirkovska
No items found.

Property management companies have a new ally—Rentgrata's revolutionary AI chatbot, Ari.

This chatbot helps them gain precise insights into what residents think about their living conditions. Using this valuable feedback, managers can make data-driven decisions to boost resident satisfaction and retention.

Rentgrata built this AI chatbot with a focus on security and validity, conducting numerous feasibility tests and evaluations before its release. But they also had an ally — they used Vellum throughout the entire development lifecycle, and were able to cut their projected 9-month development timeline nearly in half.

For those interested to learn how to build a production-ready AI chatbot, keep reading.

Who is Rentgrata?

Rentgrata is reshaping the renting experience by connecting prospective renters with current tenants to discuss living experiences. This platform rewards current residents for their insights, and new perspective residents get personalized and unique feedback.

Building on this success, they introduced Ari, a chatbot that uses advanced LLM technology to anonymously gather and analyze feedback from residents. This tool provides management companies with critical insights into tenant satisfaction, enabling them to make informed decisions and enhance the living experience for residents.

We sat down with Max Bryan, Rentgrata’s VP of Technology and Design, to learn more — here’s their test-driven journey from a simple concept to a production-ready AI chatbot that will revolutionize the proptech industry.

What Brought Them to Vellum?

Rentgrata had a vision to transform the way management companies understand and respond to resident feedback.

Early on, they tested their initial ideas by patching together models and code in Jupyter notebooks. However, they were missing a crucial element: "Evaluation.”

To build a truly reliable and actionable chatbot, they needed to evaluate their LLM features, and solve their early challenges:

  1. Overcoming the limitations of language models, such as context window size and math capabilities;
  2. Making the vast amounts of conversation data actionable and insightful for management companies.

That’s when they found Vellum.

How Does Rentgrata Use Vellum Today?

Today, Vellum is an integral part of Rentgrata's AI development lifecycle. As Max puts it, "We start with Vellum and end with Vellum.".

Their team started using Vellum for Evaluations, but soon enough started using it for every stage of their AI development:

Feasibility Testing: Before building a new feature, they use Vellum to confirm its feasibility, saving valuable time and resources.

Evaluation: Vellum's evaluation tools, both LLM-based and ground truth data, ensure that Ari's outputs are accurate and reliable.

Deployment: Rentgrata leverages Vellum's deployment capabilities to seamlessly connect Ari's backend to their user interface.

Production Monitoring: After deploying their system, they employ Vellum’s monitoring tools to gather feedback from end users and analyze the performance of their setup.

Building Ari: Actionable Renter Insights

The team has officially launched the outcome of their extensive development process: Ari, short for "Actionable Renter Insights”. This is a game-changing chatbot that enables management companies to understand what residents say about their community and what prospective residents want to know.

By analyzing conversation transcripts and presenting the data in an actionable format, Ari helps companies make data-driven decisions about marketing, management, and investments.

Preview of the Ari chatbot.

What impact has this partnership had on Rentgrata?

Bulletproof Accuracy

With Vellum, Rentgrata has achieved unparalleled confidence in Ari's performance. The numbers and insights provided by the chatbot are rigorously tested and evaluated, ensuring they are 100% accurate and trustworthy for their customers, the property management businesses.

Accelerated Development

Their small team initially estimated a 9-month timeline just for prompt engineering and evaluation.

By leveraging Vellum, they not only completed those tasks but also built complex AI workflows and managed deployment within an impressive 5 months, cutting their projected 9-month timeline nearly in half.

Actionable Insights

Ari empowers management companies to make informed decisions about where to invest their resources to improve resident happiness. In one case, a company was considering a $300,000 window renovation based on noise complaints. However, Rentgrata’s insights revealed that windows were not the primary concern, allowing the company to allocate funds more effectively.

Vellum has been instrumental in making Rentgrata's data actionable and reliable. With Ari, management companies can confidently make significant decisions based on quantifiable data, rather than relying on anecdotal reviews.

We're thrilled to collaborate with leaders like Max and his team, helping them bring their vision to life.

Want to Try Out Vellum?

Vellum has enabled more than 100 companies to build complex AI chatbot logic, evaluate their infra and ship production-grade apps.

If you’re looking to develop a reliable AI assistant, we’re here to help you. Request a demo for our app here or reach out to us at if you have any questions.

We’re excited to see what you and your team builds with Vellum next!


Join 10,000+ developers staying up-to-date with the latest AI techniques and methods.
Thanks for joining our newsletter.
Oops! Something went wrong.
Anita Kirkovska
Linkedin's logo

Founding Growth at Vellum

Anita Kirkovska, is currently leading Growth and Content Marketing at Vellum. She is a technical marketer, with an engineering background and a sharp acumen for scaling startups. She has helped SaaS startups scale and had a successful exit from an ML company. Anita writes a lot of content on generative AI to educate business founders on best practices in the field.

About the authors

No items found.

Related posts