Case study
How a construction firm used AI to modernize 60 years of hiring data
A custom document-intelligence agent reads 60 years of owner agreements, while an AI Recruiter quietly works the operational hiring pipeline across California and Arizona.
60 years
of contract history made directly queryable for the first time
2 agents
running in production: custom doc-intelligence and AI Recruiter
Zero
platform replacements: HubSpot, Procore, and Join stay in place
Overview
A U.S.-based construction firm with more than sixty years of experience on government and large commercial projects had built up an enormous, irreplaceable archive of contract history, owner agreements, project cost data, and lessons learned. The problem was that almost none of it was directly searchable. Years of value sat in PDFs, scanned signatures, contract addenda, and project folders, and the team had no realistic way to mine them when pricing a new bid or answering a question about a similar past project. Separately, the firm’s operational hiring (project managers and engineers, not union labor) was running through informal networks with no consistent outreach engine behind it.
The firm partnered with Squirrels.ai to deploy two purpose-built solutions side by side. The first is a custom AI agent that connects to the firm’s document repositories on Google Drive and SharePoint, reads owner agreements, outlines contract terms, and answers natural-language questions like “what did we pay for concrete on past large-format retail projects?” The second is an AI Recruiter focused specifically on operational hiring, project managers and engineers in California and Arizona, reaching out to candidate databases by phone, email, and SMS and booking meetings directly with hiring managers. Both agents respect the firm’s existing tech stack, including HubSpot, Procore, and Join.
Client at a glance
Industry
Commercial and government project construction
Workforce model
Union labor sourced through union halls; salaried project managers and engineers hired directly
History
60+ years in business, with deep institutional knowledge across a wide range of project types
Tech stack
HubSpot (CRM), Procore (project management and RFIs), Join (estimating with built-in AI)
Footprint
Operational staff across California and Arizona
Use cases deployed
Custom AI agent for contract and historical-document intelligence, plus an AI Recruiter for operational staff
The challenge
This was not a firm that needed help with AR collections (government contracts pay) or with prospecting (the leadership team is in daily contact with existing owners and reps). The real friction was elsewhere, in the parts of the operation that scaled with knowledge work and people, not with cash flow.
Sixty years of contract data, locked in PDFs
Across six decades of completed projects, the firm had accumulated thousands of owner agreements, change orders, cost summaries, and project closeouts. The information was real, but it was effectively invisible to anyone who was not willing to spend hours digging through folders. When a new bid required a quick reference to how a similar concrete pour had been priced in 2008, or which subcontractor had handled steel on a particular owner’s site, the answer existed somewhere on the network but was almost never retrieved in time to actually inform the decision.
The executive bottleneck
Inbound questions about historic projects, contract terms, and prior cost data all ended up routed to a small group of senior people who actually remembered the details. That made every “quick lookup” a real interruption, and the institutional knowledge stayed concentrated in a handful of inboxes and heads, rather than being made available to estimators, project managers, and engineers when they needed it.
Operational hiring with no real outreach engine
Union labor came through the union halls, which works. But the firm’s project managers, field engineers, and other operational hires had to come from somewhere else, mostly informal referrals and headhunter networks. There was no structured, repeatable way to work a candidate database, qualify interest at scale, or get hiring managers in front of qualified candidates quickly when a role opened up. Good people were available; the firm was just slow to reach them.
A tech stack worth respecting
The team already used HubSpot for CRM, Procore for project management and RFIs, and Join for estimating. Any new tooling had to integrate with what was already in production, not replace it. The firm wanted AI that fit into its existing workflows, not another platform demanding loyalty.
Additional constraints
- Contract data is sensitive: confidentiality, NDAs, and proper access controls were non-negotiable on the document-intelligence side.
- AI recruiting needed to apply only to non-union operational hires; union halls remained the path for craft labor.
- Senior leadership has limited bandwidth; AI tooling had to reduce inbound demands on their time, not add to them.
- Construction is a relationship business; any AI outreach had to sound credible and professional, never spammy.
The solution
Squirrels.ai deployed two production agents purpose-built for the firm’s operating model: a custom document-intelligence agent for the contract archive, and an AI Recruiter for operational staff.
Custom AI agent: 60 years of contracts, finally searchable
The first agent was custom-built rather than configured from a standard template. It connects to the firm’s document repositories on Google Drive and SharePoint, ingests contracts and historical project files, and acts as an executive-assistant-grade research tool for the senior team.
- Owner-agreement outlining: the agent reads new owner agreements end-to-end and produces a clean outline of obligations, retention terms, change-order language, payment milestones, and key risks for review.
- Historical cost lookups: estimators can ask plain-language questions about how the firm priced concrete, steel, or other materials on similar projects in past decades, and get answers grounded in the actual archive instead of guesses.
- Project precedent search: questions like “what did we do on the last large-format retail project we built for this kind of owner?” return real precedent from the archive, including approximate dates, project scopes, and notable details.
- Executive-assistant workflow: the senior team can hand off research requests that used to consume hours of someone’s day, and get summarized answers back, with citations to the underlying documents for verification.
- Repository integration: the agent works against the firm’s existing Google Drive and SharePoint folders directly. No mass migration, no parallel system to maintain.
- Access controls preserved: the agent operates within the access rules already configured on the document repositories, so confidentiality on sensitive contracts is enforced by default.
AI Recruiter: operational hiring across California and Arizona
The second agent is the standard Squirrels.ai AI Recruiter, configured specifically for the firm’s operational hiring scope and geographies. Union craft labor remains sourced through union halls and is excluded from the agent’s outreach.
- Targeted scope: outreach is limited to project managers, field engineers, and other salaried operational roles where the firm hires directly. Union positions are excluded by design.
- Geographic focus: candidates are sourced from databases concentrated in California and Arizona, matching where the operational staff actually work.
- Multi-channel outreach: the agent runs a coordinated sequence across phone, SMS, and email to qualify candidate interest.
- Hiring manager scheduling: interested candidates self-book meetings directly with the appropriate hiring manager’s calendar, so the firm stops losing good people to slow scheduling.
- Status sync with HubSpot: lead status, conversation logs, and meeting bookings flow back into the firm’s HubSpot CRM so the team has a single view of every candidate.
- Professional, on-brand tone: construction is a relationship business, and the agent’s tone was tuned to match: confident, brief, respectful, and never spammy.
Working alongside the existing stack
- HubSpot continues to own the CRM relationship. The AI Recruiter pushes updates back; the firm did not change how anyone uses the system.
- Procore continues to own RFIs, drawings, and project workflows. The document-intelligence agent reads from the contract archive, not from the active project workspace, so day-to-day project management is unchanged.
- Join continues to own estimating and its own built-in AI capabilities. The Squirrels.ai contract agent surfaces historic cost data that estimators can use as an input to the bidding process.
- No platform replacement, no migration project, no disruption to existing workflows.
Dashboards and audit trail
- Real-time visibility into AI Recruiter activity: candidates contacted, responses received, meetings scheduled, meetings completed.
- Conversation logs and call transcripts available on demand for every candidate interaction.
- Document-intelligence queries are logged with citations to source documents, so any answer the senior team uses on a bid can be traced back to the underlying contract.
- Clear separation between the two agents, so each can be monitored and tuned on its own.
Results and impact
The firm now has two production AI agents working alongside its existing tech stack: one mining 60 years of contracts for answers, the other running a structured operational hiring pipeline.
What changed in the business
- Six decades of contracts became a working asset. Estimators and project leads can now query the firm’s full historical archive in plain language and get sourced answers in seconds, instead of either guessing or chasing down a senior partner.
- Senior leadership got time back. Routine “what did we do on that project ten years ago?” questions are now answered by the AI agent, freeing the senior team to spend their time on judgment calls and client work.
- Operational hiring runs on a real engine. Project manager and engineer candidates in California and Arizona are reached, qualified, and routed to hiring managers automatically, instead of waiting on a referral or a recruiter callback.
- Union sourcing stayed exactly as it is. Craft labor continues to come through the halls. The AI Recruiter is scoped tightly to operational positions, respecting the firm’s labor relationships.
- Existing tools kept their place. HubSpot, Procore, and Join continue to run the workflows they were built for. The AI agents fit into the existing system, not on top of it.
- Institutional knowledge is no longer trapped in a few heads. The risk of losing critical historical context when a senior team member retires has dropped substantially, because the agent can surface the same information directly from the archive.
- Hiring conversations stay professional. Tone is confident and brief on every candidate contact, which matters in a construction market where reputation travels fast.
TESTIMONIAL
Why Squirrels.ai
Squirrels.ai deploys AI agents that handle calls, emails, and texts at scale, and builds custom agents tailored to deep, industry-specific problems. For this construction firm, the platform delivered:
- A custom AI agent that turns 60 years of contracts, owner agreements, and project files into a queryable knowledge base, with citations back to source documents.
- Direct integration with Google Drive and SharePoint, respecting existing access controls and confidentiality requirements.
- An AI Recruiter tightly scoped to non-union operational hiring (project managers and engineers) across California and Arizona.
- Two-way sync with HubSpot so the firm’s CRM stays the source of truth for candidate relationships.
- Zero disruption to the existing tech stack: HubSpot, Procore, and Join all stay in place and keep doing what they do best.
- A professional, brief, on-brand voice across every candidate contact, fitting a relationship-driven construction industry.
Decades of documents. Operational hiring that needs to move. We build the AI agents that fit your business.