How Buildertrend Took Its Customer Help Experience From Keyword Search to AI, on Its Own Salesforce Org

Buildertrend is an end-to-end construction management software for established home builders, remodelers, commercial and specialty contractors. Great support means giving customers answers when and how they need them. In addition to its award-winning customer service team, the help center empowers customers to find answers, troubleshoot issues, and keep moving with easy access to self-service knowledge articles.

Buildertrend's help center ran on the standard Experience Cloud template. Its search bar matched keywords in article titles and text, so customers had to guess the right words to find an answer. At the same time, the team was building a service bot on its website that sends customers to those same articles, and it wanted the path from bot to help center to feel like one experience.

Tylene Lucas, Staff Solutions Architect at Buildertrend, wanted search functionality that worked more like Google: ask in natural language, get a short AI summary, see the articles behind it. Tylene started her career as a developer and knows many languages, but Apex isn't one she writes fluently. She built it with Tribal, after an introduction from Jules O'Donnell, a Senior Data Engineer on Salesforce's Private Equity team who had helped Buildertrend with its first Agentforce setup.

Starting with a feature built mid-meeting

When Buildertrend came back to Jules wanting more from its Agentforce setup, the team's program manager listed what they hoped to add. Jules used Tribal to build one of those features, a type-ahead search, while the meeting was still going. The team's reaction was immediate: how did you do that?

Tylene had recently taken over Buildertrend's Salesforce technology. She saw what Tribal could do and brought her own list.

From keyword matching to answers in plain English

Tylene described what she wanted in a chat with Tribal, and Tribal turned it into a requirements document with acceptance criteria for each item. She edited some herself and asked Tribal to adjust others.

The result is a Lightning Web Component that sits on top of Buildertrend's knowledge articles and Salesforce AI prompt templates. When a customer asks something like "Why can't my sub see the purchase order I sent?", a quick query returns matching articles right away. While the LLM reads those articles and writes a short answer, the summary builds above them. When it lands, the articles it used are marked, and the rest stay visible below.

Customers no longer have to guess a title. They ask the way they'd ask a person.

Building with a human in the loop

Tylene worked in Tribal in stages. She approved requirements first, then checked the architecture view, which marks each component as created, modified, or read-only. If a component that should change isn't set to modify, she corrected it before the build starts.

During the build, she followed Tribal's steps as they ran. When Tribal reaches for a Salesforce best practice that doesn't fit a legacy piece of Buildertrend's org, she stops it and redirects. For UI components, preview mode lets her see and adjust the design before anything is deployed.

After testing the search by hand, Tylene asked Tribal to write the unit tests for the Apex behind it, including the guest user authentication. The first pass landed near 70% coverage, and a few iterations later every class cleared Salesforce's 75% requirement.

Each deployment produced release notes, which Tylene shared with internal partners so they know what changed in their day-to-day.

Tylene still uses Claude regularly. In her experience, Tribal brings Salesforce best practices and knowledge of where the platform is heading, and applies them to Buildertrend's own metadata. When she asked how ready the org was for headless and MCP-based experiences, Tribal recommended adding descriptions to fields and objects. It also recognized a custom object as the one used for quoting, without being told.

Moving Bailey, the service bot, to Agent Script

The service bot began as one of Buildertrend's first Agentforce bots. Tylene asked Tribal to convert it to an Agent Script agent. She then added routing based on business hours, plus controls to switch escalation to human agents on and off.

The bot is now live in production. It answers customer questions and hands customers to human agents when needed.

The bot is now live in production. It answers customer questions and hands customers to human agents when needed.

The impact

• AI-written answers in the help center, with source articles marked beneath each summary, replacing keyword-only search.
• Unit test coverage above 75% on every class, generated and improved within Tribal.
• A legacy Agentforce bot converted to Agent Script, with business-hours routing and escalation controls, now in production.
• Release notes for each deployment, shared with internal partners.
• An org readiness review for headless and MCP experiences, run from Buildertrend's own metadata.

Customers get a faster path from a question to an answer. Internal partners get a clear record of every change. Tylene, who owns the org, can build components, controllers, and tests in far less time than writing them by hand.

"I can create so many more things within Salesforce, and faster. A project like this, with a prompt template, LWC components, and Apex controllers, would have taken me a lot longer by myself, especially without Claude." Tylene Lucas, Staff Solutions Architect at Buildertrend

‍