Vision: Every decision at PPH is informed by AI. Every process has an intelligent layer. Not "we use some software" — the AI is the operating system.

Duncan — this plan lays out where I believe we need to go and how I intend to get us there. Everything below is my assessment of our current state, the opportunity in front of us, and the specific steps I'm proposing. I've built most of the foundation already. What I need from you is alignment on the direction and the restructuring required to execute it.

---

Where PPH Stands Today (March 2026)

The Gap

PPH today operates like most construction companies: our data lives in 6+ disconnected systems. Decisions are made by gut and spreadsheets. The AI layer (Paris Command) can report on what happened — but it can't yet drive what happens next.

AI-first means: the AI doesn't just answer questions — it watches, alerts, recommends, and eventually acts. We stay in control, but the AI is the co-pilot for every operational decision.

---

Operational Restructuring — Executing the Plan

The Problem

I'm currently running PSG day-to-day while simultaneously building the technology platform for the entire Paris group. That's unsustainable — and Duncan, you've rightly flagged that PSG hasn't been getting my full attention. But the answer isn't for me to go back to 100% PSG operations. The answer is to structure things so PSG runs well and the technology gets built.

PSG is a 3-person team. It does not need a full-time Service Manager. It needs a competent day-to-day operator with a high-level manager checking in. My highest-value contribution to the Paris group is not dispatching service calls — it's building the AI infrastructure that will make every company in the group more efficient, more profitable, and harder to compete with.

The Transition: My Role from 100% → 25% PSG

I'm not getting distracted from PSG — I'm being deployed where the ROI is highest. Every tool I build — the financial dashboards, the AI assistant, the automated alerts, the daily digest — serves PSG, PPH, and the holding companies. If I stay in PSG operations full-time, none of this gets built, and the entire group stays on spreadsheets and manual processes.

The Multiplier Effect: Duncan & Andrew

As the AI tools come online, they reduce workload across the entire management team — not just mine:

The result: As AI tools reduce your and Andrew's operational overhead, there's natural capacity for periodic PSG support if needed — not as a permanent assignment, but as flexible coverage. The tools make the whole management team more efficient, which makes the whole group more resilient.

Admin Overhead — The Strategic Decision

Admin salaries and wages are Paris's #1 expense. The AI-first plan forces a decision:

---

Why We Build — Not Buy

The Software Industry Is Collapsing

Construction SaaS companies like Procore charge $100K+/year because they employ thousands of engineers to build and maintain features. AI has fundamentally changed that equation:

• What took a 50-person team 6 months to build, one person with AI tools can ship in weeks
• The "moat" of complex software — that it's too expensive to build your own — is gone
• Vertical SaaS pricing will compress 80-90% over the next 2-3 years as AI-native competitors enter every market
• Companies that can't adapt will lose customers to cheaper, smarter alternatives

You're right that paying $100K/year for Procore today would be like signing a 3-year lease on a Blockbuster franchise in 2008. The thesis is correct — software costs are going to near-zero.

But the conclusion isn't "wait for someone else to build the AI-first CRM." The conclusion is: I'm already building it.

Build vs. Buy vs. Wait

	Buy Procore	Wait for "AI-First CRM"	Build in Paris Command RECOMMENDED
Annual Cost	$100K+/yr	$12K/yr Knowify while waiting	~$600/yr infra + Graham's time
Timeline	6-12 months to adopt	2-3 years? Maybe never	3-4 months per major capability
Data Ownership	Procore owns your data	New vendor owns your data	WE own it. Forever. In our DB.
AI Integration	Bolted on to legacy	Unknown	AI-native from day one
Customization	What they give you	What they give you	Exactly what PPH needs, nothing else
Competitive Advantage	Same as every other Procore user	Same as every other buyer	Proprietary. No competitor can buy this.
Risk	Overpaying for 3 years	Knowify data stays broken, no progress	Requires Graham's time investment

The AI-first CRM we've been talking about — I'm the one building it. Paris Command already has 2,072 customers, 5,019 invoices, 726 jobs, an AI assistant, and a daily Teams digest. That's 60-70% of what Knowify does, with better AI than any tool on the market.

The remaining pieces (job entry, cost tracking, change orders, progress billing) are 3-4 focused build phases — not a moonshot.

The Knowify Transition

We won't rip out Knowify overnight. We absorb it gradually:

The Payroll Switch

Ceridian PowerPay has no API — it's a dead end for an AI-first operation. The plan:

1. Request full historical data dump from Ceridian (employee records, pay history, deductions, T4s) — they're required to provide this
2. Switch to a Canadian payroll provider with an API (Humi or Rise People — both handle CRA, ROEs, T4s)
3. Build the importer — one-time historical data load into Paris Command
4. Auto-sync going forward — payroll data flows into Paris Command, AI can analyze labor costs per job, overtime trends, utilization

Cost: ~$5-10/employee/month (roughly what PowerPay costs anyway)
Gain: Payroll data becomes part of the AI layer — labor cost analysis, overtime prediction, capacity planning all become possible

---

The 5-Layer AI-First Model

Where PPH is now: Layer 1 is solid. Layer 2 is live. Layer 3 just started (Teams digest). Layers 4-5 are the destination.

---

KPI Intelligence — Solving the Measurement Problem

Why KPIs Have Never Worked at Paris

We've been trying for years to get the controller and management team to pick a few KPIs, track them, and improve on them. It's never stuck. I don't think this is a people problem — it's a systems problem:

How Paris Command Fixes All Four Problems

Problem 1 → Auto-Calculated KPIs. The data already flows in from QBO (both companies synced), Knowify (248 customers, 726 jobs), and soon HCP and Monday.com. KPIs compute automatically from the unified database — nobody pulls anything. The number is just there, updated every sync cycle. Zero manual effort.

Problem 2 → Single Source of Truth + Data Confidence Scoring. Every KPI traces back to the unified data layer. Revenue comes from QBO. Job costs come from Knowify/HCP. There's no "which spreadsheet is right?" because there's only one database.

Each KPI displays a Data Confidence Indicator:

Nobody can dismiss a KPI by saying "those numbers are wrong" without the system either confirming or denying it. The confidence indicator kills the excuse.

Problem 3 → Role-Based KPI Menus. Each role gets a curated menu of KPIs relevant to their function. You pick 3-5 from your menu. You see what matters to you — nothing else.

Each person selects their KPIs once. They see them every time they open Paris Command. They see them in their morning Teams digest. The KPIs are theirs — relevant, personal, persistent.

Problem 4 → AI Accountability Loop

The AI doesn't forget. It doesn't get busy. It doesn't avoid the awkward conversation.

Why This Is Different From Every Other KPI Attempt

Every previous attempt required someone to do the work of tracking KPIs. This system does the work automatically. The only human action required is:

1. Pick your KPIs (one-time, 5 minutes)
2. Set your targets (one-time, 10 minutes with your manager)
3. Read your morning digest (daily, 2 minutes)
4. Act on what the AI flags (this is the actual job)

The controller doesn't compile reports. The PMs don't fill out spreadsheets. You don't chase people for numbers. The system collects, calculates, validates, and delivers. People just have to look at it and act.

And if they don't act? The AI keeps flagging it. Every morning. "This KPI is still off target. Here's the trend. Here's what's causing it." The system doesn't give up, doesn't get distracted, and doesn't accept "I'll get to it next week" as an answer.

---

Stress-Testing This Plan — The Hard Questions

Before presenting this plan, I ran it through an adversarial review — four competing perspectives arguing for and against the AI-first approach given our constraints. I want to be transparent about the tensions and how I've resolved them.

The Three Irreconcilable Tensions

1. Speed vs. Depth. I can't revamp all systems and processes in 6 months as a solo builder. Even the most aggressive position in the review had to concede that. But we're not on a hard clock — we're willing to take 3-5 years to reach maximum value. That means I can build this right instead of building it fast. The phases are sequenced by dependency and impact, not by an arbitrary deadline.

2. EBITDA vs. Multiple. Every dollar of EBITDA matters at exit — a buyer pays 4-6x on earnings. But the multiple itself is the bigger lever. A traditional mechanical contractor at 5x EBITDA caps out at $15M on $3M earnings. A contractor with proprietary AI ops tools and demonstrably better margins becomes a 6-7x company. On $3-4M EBITDA, that's $3-8M of incremental value. The resolution: build tools that DIRECTLY improve financial metrics. When we're ready to sell — whether that's in 2 years or 5 — the data proves itself.

3. Operational Independence. The tools aren't the risk — any developer can pick them up. The real bottleneck is operational. Right now, Duncan, you ARE the operations. Estimation runs through you. PM oversight runs through you. Major job decisions run through you. If we ever want the option to sell, a buyer doing diligence won't ask "what happens when Graham stops coding" — they'll ask "what happens when Duncan stops working 55-hour weeks?" If the answer is "the business slows down," the valuation takes a hit. The AI tools are part of the solution here: they systematize the knowledge that currently lives in your head, they automate the oversight you do manually, and they create a management layer that doesn't depend on any single person — including you. That's not about replacing you. It's about making the business resilient enough that it doesn't need any one person to function.

What the Review Changed in My Approach

The original plan was comprehensive — 8 phases, full transformation. The stress test narrowed my focus:

The Guardrail

Nothing in this plan risks PSG profitability. If the choice is between shipping an AI tool and hitting our financial targets, the financial targets win every time. The AI work should serve profitability, not compete with it. Where possible, the first tools I build are the ones that directly reduce overhead or improve margins — automated dispatch, job costing visibility, compliance tracking.

Owner Independence — Building for Resilience

Even if we never sell, every tool I build needs to work without me in the room. I could get hit by a bus, go on vacation, or shift focus. The tools can't break when I'm not around. My approach:

• Paris Command runs on Cloudflare Workers — zero-maintenance infrastructure. No servers to patch, no databases to babysit. It runs like a utility.
• The tools are operational, not experimental. They pull data from QBO, format it, and present it. They don't require an engineer to operate — they require someone who can read a dashboard and act on alerts.
• Documentation is built as I go. Every tool gets a runbook. The AI assistant answers questions about the business data — it doesn't require someone to maintain the AI.
• Ongoing maintenance is minimal. Cloudflare's platform handles scaling, uptime, and infrastructure. If a connector breaks (QBO changes their API), that's a $5K-10K contractor fix — not an existential dependency.

If we do decide to sell, the story writes itself: "You're acquiring a mechanical contracting platform with AI-driven ops that produce X% better margins than industry average. The tech is production-proven, runs on zero-maintenance infrastructure, and is operated by the management team — not dependent on the founder." But more importantly, this resilience makes the business stronger whether we sell or not.

---

The Transition Plan

---

The Transformation Timeline

---

What Makes This "Bleeding Edge"

The competitive advantage isn't the software itself — it's the speed of decision-making. A company that catches a margin problem on day 3 instead of day 30 saves real money. A company that quotes in 45 minutes instead of 4 hours wins more work. A company that never misses a cert renewal or insurance lapse never gets shut down on a job site.

---

The Software Cost Revolution

The question isn't "can we afford to build this?"
The question is "can we afford not to?"

Every year PPH waits is a year competitors could adopt this approach. The ones who move first build a compounding advantage — more data, smarter AI, faster decisions, better margins. The ones who wait are buying yesterday's software at yesterday's prices.

---

Investment Required

ROI Examples

• Catching one over-budget job 2 weeks earlier = $5K-50K saved
• Processing invoices 80% faster = bookkeeper redeployed to higher-value work
• Quoting 3x faster = more bids submitted = more wins
• One avoided compliance lapse = avoided project shutdown
• Eliminating Knowify = $12,000/year saved permanently
• Eliminating future Procore consideration = $100,000+/year never spent

---

Build Priority — Impact-Driven Sequencing

Not every phase carries equal weight. Here's how I've prioritized — by financial impact and dependency, not by an exit clock:

Every phase makes the business more profitable, more efficient, and more valuable. We build them all. The sequencing is about dependencies and impact — Phase 5 needs Phase 1's data layer, Phase 7 needs Phase 2's alerting framework, and so on.

---

Immediate Decision Checklist

1. Confirm PSG profitability path is intact before committing building time — the financial targets are non-negotiable
2. Align on the PSG restructuring — Aisley takes day-to-day, I step back to 25%, so I can execute the buildout
3. Agree on the admin overhead approach — hold steady and reassess at Month 6 with data, or address performance issues now
4. Start documenting KPI baselines today — we need "before" numbers for every metric we plan to improve, so we can measure the real impact
5. Set the hard rule: every AI tool must map to a financial metric — this keeps me focused on what improves the business, not what's technically interesting

---