One Retirement Away from Starting Over

Tuesday morning. A framing crew was supposed to show up at the Leaside project at 7 AM. By 9 AM, nobody had arrived. Ray, the project manager, called the sub. Voicemail. Texted. Nothing.

By 10 AM, he was calling down his contact list. He needed a framing crew available this week, insured, within 30 minutes of the site. He tried four subs before someone said yes.

The crew that showed up had never worked on one of their sites. They were available because another GC had dropped them after quality issues. Ray didn't know that. He picked them because they were the fifth call and the first yes.

Three weeks later, the framing inspection failed. Rework cost: $8,400. Project delayed 9 days. The client was furious.

"I didn't have a better option at 10 AM on a Tuesday," Ray said. He was right. Not because better options didn't exist. Because his system for finding them was his phone and his memory.

General contracting firm in Toronto. Residential renovations and custom builds across the Greater Toronto Area. The owner, two project managers including Ray, four full-time carpenters, one office admin. Eight to twelve active job sites at any given time.

They use 60-plus subcontractors across 14 trades: framing, electrical, plumbing, HVAC, roofing, drywall, painting, flooring, tile, landscaping, concrete, demolition, insulation, window and door installation. The core crew handles rough carpentry. Everything else goes to subs.

Ray has been with the company 18 years. He handles most sub booking. His method: a mental ranking by trade, quality, reliability, location, and availability. Calls his top pick first, works down the list if they're booked. Sometimes checks insurance status. Sometimes doesn't. Depends on the urgency and how well he knows the sub.

No centralised database. No performance tracking. No insurance monitoring. No certification system.

The past 12 months: two inspection failures ($8,400 plus $5,200 in rework), one insurance lapse discovered mid-project ($3,600 in delays), three scheduling conflicts ($11,000 in cascading costs). Total sub-related waste: roughly $28,200.

Ray is 57. When he retires, 20 years of subcontractor knowledge walks out the door. Who's reliable, who cuts corners, who's great but overbooks, who answers the phone at 7 AM. All of it gone in an afternoon.

The owner's response when asked about Ray's retirement: "I try not to think about that."

The owner started a shared spreadsheet three years ago. Sub name, trade, phone number, notes. Twenty-eight entries. Ray never used it. "I already know who to call." Abandoned within months.

Asking subs about their own insurance sounds reasonable. "You're covered, right?" Most say yes. Some aren't current. The GC is liable if an uninsured sub gets injured on site. The $3,600 work stoppage came from a sub who said "yeah, we're covered" during the two-week gap between his renewal date and his broker processing the payment. Asking isn't verifying.

Reference checks happen for new subs but never get updated. A sub who was excellent three years ago might've expanded too fast or lost key people. Without ongoing tracking, initial references are all you've got.

Construction management software includes sub directories, but the company uses a basic project tool and isn't migrating. The sub matching feature alone doesn't justify $400 to $800 per month for a platform they'd use 20% of.

Hiring a coordinator was discussed. $50K-plus per year. Still depends on Ray for initial setup. Still one person. Just a different single point of failure with less institutional knowledge.

Four stages, built around capturing what's in Ray's head and keeping it current.

Stage 1: Subcontractor registry

Centralised database of all 63 subs. Each record includes: trades covered, service area, insurance status with expiry date, certifications with expiry dates, WSIB (workplace safety insurance) status, crew size, typical availability lead time, rate range, and contact details.

The initial data came from two brain-dump sessions with Ray. Three hours total, spread across two mornings. He sat with a coffee and talked through every sub he works with. Who they are, what they do well, where they're weak, which sites they're good for, which ones to avoid on tight timelines. Twenty years of accumulated judgment, captured in a structured format for the first time.

The registry isn't static. It gets updated every time a sub is booked and every time a job closes.

Stage 2: Performance tracking

After every job, the project manager logs a quick score. Quality: 1 to 5. Reliability: showed up on time, yes or no. Communication: 1 to 5. Any issues noted. Takes 60 seconds.

Over time, this replaces Ray's mental ranking with documented history. Not a feeling that "Mike's crew has been slipping lately" but a trail showing quality scores dropping from 4.5 to 3.2 over six jobs and an on-time rate of 60%. Patterns that are hard to see across 60 subs and dozens of jobs per year become visible in a profile.

Stage 3: Smart matching

When a sub is needed, the agent matches against: trade required, proximity to site, availability based on recent booking patterns, insurance status (verified, not self-reported), valid certifications, and performance score. Returns a ranked list. Best fit first, not first to answer the phone.

The Leaside framing crew would've scored differently. No performance history. No positive references. Low match score. They'd still appear on the list, but at the bottom, flagged as unvetted. Ray would've had the signal that they were an unknown quantity before committing.

Stage 4: Compliance monitoring

The agent monitors insurance expiry dates and certification renewals for every sub in the registry. Thirty days before any policy or certification expires, it sends an alert: "ABC Electrical's liability insurance expires April 15. Verify renewal before next booking."

This is the stage that would've prevented the $3,600 work stoppage. The sub whose insurance had lapsed between renewal and processing would've been flagged before booking, not discovered mid-project when the owner asked for a certificate.

Ray's brain dump was the most valuable part of the entire project. Not for the agent. For the company. Twenty years of accumulated judgment, captured in two sessions, documented for the first time. The owner sat in on the second session and said afterward: "I've worked with Ray for 18 years and I didn't know half of this." Which is exactly the point. The knowledge was locked in one person. It was invisible to everyone else until someone took the time to extract it.

The 60-second performance log needed to be truly 60 seconds. The first version asked for too much detail. Written comments, detailed scoring across five dimensions, timeline accuracy percentages. The PMs filled it in for two jobs and then stopped. We cut it to four fields: quality score, on-time yes/no, communication score, and a single notes field. Completion rate went from about 30% to over 85%. Quick and rough beats thorough and abandoned every time.

Insurance verification was harder than expected. You can't call an insurer and ask if a policy is active. The sub has to provide a certificate, and certificates have expiry dates. The agent tracks dates but can't independently verify status. We built a workflow: the agent requests updated certificates 30 days before expiry and flags any sub who hasn't provided one by 14 days out. Not perfect. But dramatically better than "you're covered, right?"

One sub fired the company. A drywall sub who'd been getting regular work saw his performance scores and took offence. "I've been doing this for 25 years, I don't need a report card." He stopped taking their calls. Ray's take: "His scores were low because his work was average and he was late three out of five jobs. If he doesn't want to be measured, that tells me something." They found a replacement within a week. The replacement scored higher across the board.

Estimated annual savings: CAD $20,000 to $30,000. Avoided rework, prevented insurance incidents, reduced scheduling delays, faster sub booking that keeps projects on timeline.

CAD $260 per month. $3,120 per year. Against $20,000 to $30,000 in avoided waste.

But the number the owner keeps coming back to isn't the savings. It's the continuity. If Ray gave notice tomorrow, his replacement would inherit a documented sub database with performance histories, insurance tracking, and compliance monitoring. Not a blank phone and a "good luck." Ray's 20 years of knowledge is now the company's knowledge. It lives in a system, not a person.

Ray's response when he saw the registry for the first time: "I didn't realise I knew this much." He paused. "Actually, that's terrifying. What if I'd just quit?"

If you're running a contracting company, trades business, or any operation that relies on external partners, and those relationships live in one person's head, two things are true.

You're not selecting the best fit. You're selecting the first to answer. The gap between those two choices is the rework, the delays, and the insurance incidents that show up as "just part of the business" instead of what they are: a matching problem with a measurable cost.

And your most valuable operational asset has a retirement date. When that person leaves, the business doesn't lose an employee. It loses the infrastructure.

The agent doesn't replace Ray's relationships. It captures them, documents them, and makes them available to everyone. Ray still makes judgment calls. The agent makes sure the company keeps what he knows after he's gone.

Want to see if your subcontractor or vendor relationships are a business continuity risk? The AI Bottleneck Audit takes 5 minutes and shows you where the knowledge gaps are hiding. No pitch.

Want to see 25 agent architectures across different industries? Download Unstuck. It includes blueprints for pipeline tracking, transaction coordination, referrals, equipment lifecycle, and more.

by CC
for the AdAI Ed. Team