A Build Spec Without a Payback Period Is Just a Scope of Work
After the Operations Map, the question isn't whether to automate. The audit usually makes that obvious. Phase 2 answers a narrower question: which workflow do we build first, what exactly does it do, and how fast does it pay back?
The deliverable is two documents per workflow. A build spec. An ROI model.
What Goes Into the Spec
A workflow spec has five fields, and none of them are vague:
Trigger: what starts the automation. A web form submission, a job status change in the field tool, a time-based schedule, a CRM record update. One clear event.
Source data: what the automation reads before it does anything. Customer record, job history, quote template, current material pricing. The spec names each source and confirms the data exists in a usable format.
Transformation: what changes. The AI drafts a quote in the owner's voice. An invoice PDF is generated. A follow-up message is written with the original job date in it. One action.
Output: where it lands. A draft in Gmail, a row in Airtable, a PDF in a shared folder, a Slack message to the owner.
Human gate: what the owner reviews before anything leaves the system. In early builds, the gate is almost always there. After a few months of trust, most get removed.
Writing the spec before the build starts kills scope creep. There's no "I thought you'd also handle X." The spec said what X was and whether it was in.
The ROI Model
Every spec comes with three numbers.
Build cost: the flat fee for that specific automation. Not the engagement total. Just this build.
Annual savings: hours per week from the audit, times the hourly rate for whoever currently does it, times 52. The audit already has the hour ranges as a typed pair. We use the midpoint and a conservative wage.
Payback period: build cost divided by annual savings, in months.
A quote follow-up workflow built at $3,200, saving 10 hours per week at $30 per hour: annual savings of $15,600, payback period of 2.5 months. If the owner's time is worth $80 per hour instead of admin time at $30, the payback period shortens to under a month.
The model uses the audit's numbers directly. If the business owner corrects the volume assumption, the model updates. That correction is what makes the call worth having.
The Ordering Logic
Three workflows make it to the Build Plan. Payback period isn't the only ranking factor.
Dependency order: invoice automation requires job completion tracking to exist first. The spec for build 01 is the one with no upstream dependencies. The client shouldn't pay to build something that requires another build to work.
Data readiness: job history memory only works if the CRM records are consistent enough to query. If they're not, that gets flagged in the spec. The first step might be a data normalization pass before the automation layer goes in. That's in the document, not hidden from the price.
The client sees the ordering and the logic behind it. They can push back before anything starts. That conversation is better at the spec review than three weeks into a build.
Why Payback Period Over ROI Percentage
A two-year ROI of 400 percent is accurate. It also sounds like marketing.
Payback period maps to cash flow. "This build pays for itself in two and a half months" is a sentence a business owner can evaluate against their current position. It's also a sentence they can repeat to a partner without explanation.
We use 12 months as the threshold for a clear yes. Under 6 months is straightforward capital allocation. Under 3 months is a decision that requires almost no persuasion.
The spec gets that number on the page before the proposal conversation. The conversation stops being about whether to do it and starts being about which one first.