Work – Bright

Work

Two builds, one on each side of the value equation: AI that reduces cost and AI that grows revenue.

Project 1

AI-powered customer support for a ride-hailing platform

Industry: Ride-hailing, Philippines
Function: Customer support (live chat)
Value lever: Automate tasks, lower cost
Built on: Claude API

The challenge.

A ride-hailing platform's support team handles over 2,000 cases a day across channels, with live chat carrying the highest volume. A team works through every chat, and solving even simple issues means checking several systems and trackers by hand, so resolution times run long.

What we built.

An AI support agent that works inside the support team's existing chat platform. It picks up chats right away, pulls the information it needs from multiple sources, and drafts a reply that a second AI model reviews for accuracy and safety (LLM-as-a-judge). Support staff approve every draft at first, and each type of reply is automated once it has proven reliable, with the AI review still in place.

How it works

A support chat, step by step

Customer
AI
Your team

Customer

Chats in

On the channel they already use

Understands the request

Sorts the concern and works out what’s needed

Gathers the facts

Pulls only the data the answer needs

Drafts and checks a reply

Answers only what the facts support

Customer

Gets an answer

Fast, in the same chat

Your team

Or takes over

With a summary ready, so no one starts from zero

Built-in safeguards.

Guardrails are enforced in code, not just in prompts. The AI agent only sees what it is allowed to, and sensitive or urgent cases go straight to a person. The client's team can step in or switch it off at any time.

What it is designed to deliver.

Faster answers for drivers, less manual work for support staff, and fewer handoffs over time as the system learns.

Project 2

An AI sales agent for a resort membership developer

Industry: Leisure real estate and resort membership, Philippines
Function: Sales and lead conversion
Value lever: Add capacity, personalize at scale
Built on: An agentic workflow on Claude

The challenge.

A premium resort developer receives a high volume of new leads every day. Sales reps reached out to each one by hand, follow-ups depended on when a rep had time, and conversations were scattered across channels.

What we built.

An AI sales agent that reaches every new lead quickly, answers questions using only the client's approved sales content, and follows up based on how interested each lead is. When a lead is ready to talk to a person, a rep takes over with the full context.

How it works

From inquiry to closed deal, step by step

Lead
AI
Your team

Lead

Inquires

From an ad, your website or a form

Reaches out right away

While interest is still fresh

Answers and follows up

Using only your approved information

Spots who’s ready

Tells the browsers from the buyers

Your team

Takes the warm leads

With a brief on what the lead wants

Your team

Closes the deal

Calls, viewings and signing stay with your people

Personal at scale.

Each reply responds to what the lead actually asked, draws on the full conversation, and sounds like the client's own sales team.

Built-in safeguards.

The AI agent never makes commitments only a person should make. Every message has its key facts matched against the client’s approved information, and an independent AI check confirms it’s accurate before it’s sent. During rollout, a human signs off on each message. Over time, replies that consistently meet the standard go out automatically, while the checks continue on every message.

What it is designed to deliver.

A fast response to every lead, consistent follow-up, and one complete record of every conversation.

Have a similar problem?

Tell us about the work that is slowing your team down.

Book a discovery call