← ALL CASE STUDIES
Construction & Real Estate

An AI agent for intelligent lead discovery

A regional construction firm kept finding high-value projects after the bidding window had effectively closed. We built a continuous AI agent that scans, understands, and delivers early-stage opportunities directly to their team.

90%Lead relevance
−90%Manual research time
150miCoverage radius, continuous
Regional construction firm
IndustryConstruction & Real Estate
Size~200 employees
Built byPromata — Blueprint → AI Build
StatusLive · running continuously

Does this sound like your firm?

You hear about the perfect project when bidding opens — alongside every competitor.
Someone spends hours a day refreshing portals, news sites, and planning boards.
Your “lead list” is a generic database everyone else also pays for.
Relationships win work — but you meet decision-makers after the shortlist forms.
The jobs you’d have won are the ones you never heard about.
IF YOU CHECKED THREE OF THESE — THIS CASE STUDY IS ABOUT YOU.
The Challenge

Missing early-stage opportunities.

bid-platform search — "schools" — 47 results, sorted by deadline
Elementary school additionPOSTED 61 DAYS AGOBID CLOSES 3 DAYS
Middle school renovation, Ph. 2POSTED 44 DAYS AGOBID CLOSES 6 DAYS
District athletics facilityPOSTED 39 DAYS AGO8 BIDDERS ALREADY
Charter school gymnasiumPOSTED 30 DAYS AGO11 BIDDERS ALREADY

47 RESULTS, SORTED BY DEADLINE — THE WINDOW HAD ALREADY CLOSED.

Construction firms live and die by timing. This team was consistently arriving too late — competing in crowded public bids against opportunities they should have known about months earlier.

  • Leads found too late

    Most platforms surface opportunities only after public bidding has opened — when competition is already at its peak and margin pressure is highest.

  • Manual research at scale

    The team spent hours monitoring websites, tracking fundraising campaigns, and re-checking the same sources repeatedly — time-consuming and prone to gaps.

  • Low relevance from existing tools

    Generic lead databases had poor niche filtering: no ability to detect the early signals — capital campaigns, grants, property acquisitions, institution-specific plans — that precede construction.

  • No update tracking

    Even when leads were identified, there was no visibility into progress. No alerts when funding milestones were reached or projects changed phase.

The Solution

An AI agent that thinks, filters, and tracks.

We built an AI-driven lead discovery and intelligence system that identifies early-stage opportunities, tracks them over time, and delivers only meaningful updates — directly into the team's existing workflow.

  • Geo-targeted lead discovery

    The agent continuously scans local news and institutional websites within a 100-to-150-mile radius of the target locations. Geographic relevance is enforced at the source, not filtered after the fact.

  • LLM-powered opportunity detection

    Rather than keyword matching, the AI reads and understands context. It identifies capital campaigns, grant announcements, and property acquisition plans — detecting pre-construction signals before competitors even know to look.

  • Niche-specific filtering

    The system targets the client's actual market — schools, churches, non-profits, and public institutions. Irrelevant institutions are filtered out before extraction, so the team only ever sees opportunities that match.

  • Lead intelligence engine

    A three-layer system prevents noise: duplicate detection ensures no lead is surfaced twice, continuous re-scanning tracks changes over time, and an LLM significance layer decides whether an update is worth a notification. Funding milestone reached? Alert. Minor page edit? Ignored.

  • Real-time delivery and feedback loop

    Structured lead cards arrive directly in the team's channel with consistent format: organisation, location, opportunity summary, source link. The team rates each lead with a thumbs up or down — and that feedback improves relevance over time.

What Makes This Different

Not a scraper. An intelligent agent.

  • Context, not keywords

    The system reads and understands source content the way a researcher would — identifying opportunity signals, not just matching terms. It catches what keyword tools miss entirely.

  • Early-stage detection

    Capital campaigns and grant announcements appear months before projects reach public bidding. The agent watches at that stage, giving the team time to build relationships before competitors know a project exists.

  • Update intelligence

    Discovering a lead is only half the job. The system tracks it, detects meaningful changes, and sends alerts only when something actionable has happened — without manual follow-up.

  • Zero-noise architecture

    Duplicate prevention and significance analysis work together to ensure the team is never alerted about something they've already seen or something that doesn't matter. Every notification is worth opening.

Architecture

Built for scale and reliability.

A lean, cloud-native stack designed to run continuously without intervention — scanning, reasoning, deduplicating, and delivering at scale.

WEB SOURCESNEWS · SITES · REGISTRIES
SCRAPING ENGINEGEO-SCOPED
LLM DETECT + FILTERCLAUDE
SIGNIFICANCE SCOREDEDUPE · TRACK
DATABASELEAD MEMORY
SCHEDULERCONTINUOUS
TEAM CHANNELSTRUCTURED CARDS
WEB SOURCES → SCRAPING → LLM DETECTION → SIGNIFICANCE → LEAD MEMORY → SCHEDULER → YOUR TEAM
Results & Business Impact

Early access. Less work. Better leads.

The team no longer spends hours on manual research — and no longer learns about projects after the window has closed. The system works continuously so they can focus on winning work.

  • 90% lead relevance

    Highly targeted, niche-specific opportunities. Irrelevant results are eliminated at the source, before they ever reach the team.

  • 90% reduction in manual research time

    Automated discovery replaced the repetitive monitoring that consumed team hours across multiple sources, every single day.

  • Early access to high-value projects

    The team engages with prospects before public bidding opens — building relationships while the competition is not yet watching.

  • Zero duplicate and low-value alerts

    Only meaningful updates are delivered. Every notification is worth reading — no noise, no repeats.

From manual research to an intelligent opportunity engine.

The firm no longer waits for the market to tell it what's available. It finds projects early, tracks them intelligently, and engages at the right moment. That's not a productivity gain — that's a competitive advantage that compounds.

90%Lead relevance
−90%Manual research time
150miTerritory, watched continuously

What we didn't solve

Sources behind a paywall or a login are out of scope and are not monitored — that is a stated boundary, not a gap being worked on. The agent finds and scores opportunities; deciding which to bid is still the estimator's call.

Have a process worth automating?

Tell us about your operation. We'll show you what a system like this could do — and what it's worth — before you spend a dollar.

Book a discovery call
The Solution

An AI agent that thinks, filters, and tracks.

We built an AI-driven lead discovery and intelligence system that identifies early-stage opportunities, tracks them over time, and delivers only meaningful updates — directly into the team's existing workflow.

Geo-targeted lead discovery

The agent continuously scans local news and institutional websites within a 100-to-150-mile radius of the target locations. Geographic relevance is enforced at the source, not filtered after the fact.

LLM-powered opportunity detection

Rather than keyword matching, the AI reads and understands context. It identifies capital campaigns, grant announcements, and property acquisition plans — detecting pre-construction signals before competitors even know to look.

Niche-specific filtering

The system targets the client's actual market — schools, churches, non-profits, and public institutions. Irrelevant institutions are filtered out before extraction, so the team only ever sees opportunities that match.

Lead intelligence engine

A three-layer system prevents noise: duplicate detection ensures no lead is surfaced twice, continuous re-scanning tracks changes over time, and an LLM significance layer decides whether an update is worth a notification. Funding milestone reached? Alert. Minor page edit? Ignored.

Real-time delivery and feedback loop

Structured lead cards arrive directly in the team's channel with consistent format: organisation, location, opportunity summary, source link. The team rates each lead with a thumbs up or down — and that feedback improves relevance over time.

Want results like these?

Tell us where your team is losing hours. We'll show you exactly what automation can do about it — and what it's worth, before you spend a dollar.

30 minutes · No pitch deck · An honest first read

Case study: AI agent for intelligent lead discovery — 90% lead relevance