By hand in spreadsheets
Bids built line by line
Each bid was assembled by hand over days, then re-versioned by email until it was hard to say which one was current.
Case Study
Urban Design Landscape
From spreadsheets to a single operating system.
We built Urban Design Landscape a custom platform that turns a pool and landscape design-build business into one connected system. A construction bid that used to take days now arrives fully itemized in minutes, every customer and project lives in one place, and an AI assistant can actually do the work across the app, not just talk about it.
Newport Beach, CALuxury landscape design-build
[THE CLIENT]
About Urban Design Landscape.
Urban Design Landscape is a landscape design-build firm in Newport Beach, California, crafting high-end outdoor spaces from pools and hardscape to planting and lighting. Like most design-build companies, they ran on a patchwork of spreadsheets, inboxes, and tools that did not talk to each other.
[THE CHALLENGE]
The patchwork we set out to replace.
Scattered records
Customers in two places
Customer details lived across inboxes and phones, with the same person saved in more than one place.
No consistent capture
A dozen channels, no capture
Leads arrived through a dozen channels with no consistent way to record what each person actually wanted.
Status in someone's head
No shared view of a project
Where a project stood lived in someone's head rather than on a record the whole team could see.
Estimates never checked
Bids never met actuals
The numbers so carefully estimated were rarely compared to what the work actually cost.
[WHAT WE BUILT]
One platform, tailored to how a design-build company actually operates.
AI-assisted bid creation
Produce a fully structured construction bid in minutes instead of days, straight from a conversation.
Construction bids
Versioned and organized by trade, editable inline, with a built-in budget check and full revision history.
Guided design workflow
Every project moves through clear stages, with a one-click approve step that locks in a snapshot and an audit trail.
Design proposals
A separate, professional design contract with deposit terms, included revisions, and site visits.
Visual project pipeline
A drag-and-drop board that mirrors how the business really runs, end to end.
Customer management
Every customer organized, deduplicated, and linked to their projects and bids.
Portfolio and wishlist
Prospects browse finished work, heart what they love, and send it all in with one inquiry.
AI assistants throughout
From a public intake agent that turns a conversation into a project to an internal assistant that builds bids and moves projects forward.
An operations layer
Receipts and bid-vs-actual variance, invoicing, crews, timesheets, and a shareable client status page.
[HOW IT FLOWS]
Every project moves through seven clear stages.
Submitted
A project enters the system with its own sequential reference, like UDV-001.
Reviewing
The team confirms scope and reviews the request before design work begins.
Designing
Work happens in a focused design view built around the bid.
Concept Sent
The concept goes to the client for their first look.
Revising
Feedback comes back and the design is refined.
Approved
One approve step locks in a snapshot of the bid totals and records who signed off and when.
Completed
The build is delivered and the project closes out.
[A CLOSER LOOK]
Three places the system does the heavy lifting.
AI-assisted bid creation
- Describe the job to an AI assistant and a fully itemized bid lands inside the platform.
- Organized into trade sections with every line populated.
- Days of manual assembly become minutes, and it is safe to re-run without ever making a duplicate.
A visual project pipeline
- A drag-and-drop board lays every project across the seven workflow stages.
- Move a card to the next column and the project advances.
- Every project gets an automatic reference ID like UDV-001 the moment it is created.
A public portfolio with a wishlist
- Prospects browse finished work and heart what they love, from a whole project to a single material.
- Those favorites collect into a wishlist that travels with their inquiry.
- The first conversation starts from what they loved, not from scratch.
What changed
From a patchwork to one system.
Before
After
Bid creation
Assembled by hand in spreadsheets over days.
A fully itemized construction bid in minutes.
Source of truth
Scattered files and long email threads.
One connected system with versioning and audit trails.
Leads
A vague inquiry with no context behind it.
An inquiry that arrives with the prospect's wishlist attached.
Visibility
Project status lived in someone's head.
A live, at-a-glance pipeline of every project.
Estimating
Bids were rarely checked against real costs.
Bid-vs-actual variance on every completed job.
[THE RESULTS]
One system that wins, designs, and delivers the work.
Bid creation went from days to minutes.
- The biggest time sink is now a few-minute, AI-assisted task.
- Safe to iterate on without ever making a mess.
One source of truth.
- Bids, designs, customers, and projects in one connected system.
- Versioning and audit trails in place of scattered files.
Leads that qualify themselves.
- Browsing becomes a wishlist that arrives with every inquiry.
- The team knows what a prospect wants before the first call.
A business you can see.
- A drag-and-drop pipeline and clean CRM in one place.
- A live, at-a-glance view of every project and customer.
AI that acts, not just chats.
- Inside the platform, it builds bids and advances projects.
- On the public side, it turns conversations into real leads, with read-only guardrails where data should stay safe.
The estimating loop closed.
- Bid-vs-actual variance on every completed job.
- Each finished job makes the next bid sharper.
[UNDER THE HOOD]
Built on a Next.js application backed by a Supabase (Postgres) database, with AI delivered through a governed AI gateway for consistent control over models and cost. The backbone is a clean match-or-create pattern: the AI generates structured information and the database enforces correctness, so the same action can run again and again without ever producing a duplicate. Let the AI generate, let the database keep it correct.
Let’s start building together.