AI tools are commodity. The wrap is the value.
Every professional services firm I talk to has access to the same AI tools. Their CRM has AI. Their ERP has AI. Their project management tool has AI. Microsoft sells them Copilot. ChatGPT runs on every laptop in the building.
What’s rare isn’t the tool. What’s rare is leverage — measurable improvement in how work actually gets done.
The reason is structural. Every business process has the same shape: an input goes into a process, which produces an output. AI lives inside the process step. But a generic AI tool — Claude, GPT, Copilot, the AI button buried inside an ERP — is only half the puzzle. It’s powerful, but it doesn’t know the specific business, the specific process, the document types, the people, or what “good” looks like.
The other half of the puzzle is someone who owns the process. Someone who maps it, builds the system around the tool, runs it in production, and improves it over time.
Without that wrap, AI is a clever feature that doesn’t move the P&L.
What the wrap actually is
When I built AECFirmAI’s Title Commitment automation, the AI part was maybe 15% of the work. The other 85% was:
- Figuring out which documents matter (Schedule B, Part II — and only the exceptions, not the requirements)
- Parsing the recorded document links out of the source PDF
- Building a downloader that handles county recorder sites with their varied authentication quirks
- Running OCR against the downloaded docs
- Building a parcel-level applicability check using Township / Range / Section comparisons
- Wiring the structured output back into the firm’s workflow so a surveyor could actually use it
- Building the audit trail so we’d know what was processed when
- Making the whole thing idempotent so reruns didn’t double-charge
- Handling the edge cases (blanket easements, items requiring field verification, missing source documents)
The AI does one specific thing in that pipeline: it reads the legal text of each exception and classifies it. That’s a job a model is good at. Everything around the model is engineering that has nothing to do with AI — and that engineering is what turns “we ran the document through GPT” into “we processed 41 exceptions in 17 minutes, work that used to take two days.”
That engineering is the wrap. The wrap is the value.
Why this matters for how you buy AI
If the tool is the value, you should be paying for tools. Pay your Copilot license, train your team, hope it sticks.
If the wrap is the value, you should be paying for the wrap. That looks different. You’re paying for someone to sit inside the work, map a specific process, build a system around the AI, and run it. The AI license is a line item inside that, not the thing you’re buying.
Per-seat pricing makes sense when the thing being sold is access to a tool. It makes no sense when the thing being sold is an operational outcome. You don’t pay per seat to have your accounting done — you pay for the work to be done.
So that’s how I price. Utilization, not seats. Pay for what the system produces. If it produces nothing, you pay nothing. If it produces a thousand outputs, you pay for that. Cost tracks value because the tool’s cost already does — and your bill should follow the same logic.
The test
If you’re evaluating an AI vendor or an internal AI initiative, here’s the question that cuts through the noise:
What part of this is the wrap, and who owns it?
If the answer is “we’ll figure that out after we sign,” the leverage isn’t real yet. You’re buying a tool and hoping someone — probably you — will build the wrap.
If the answer is concrete — “the wrap is this specific process, this specific data flow, these specific integrations, and I own running it after we ship” — that’s the part that moves the P&L.
The tool is downstream. The wrap is the work.