Practical automation for Texas businesses that still run on paper and spreadsheets
What we automate first
- Quoting and estimating
- Scheduling and dispatch
- Follow-up on open quotes and open invoices
- Data entry between systems that don’t talk
The framing
Technology without the technology project.
A technology project has a name, a budget, a steering committee and a go-live date. It asks the company to stop what it is doing and become a different company on a Monday. In a business this size that is a serious risk, and owners are right to be suspicious of it.
The scar is usually specific. An ERP that took eighteen months and never got used. A CRM the field team quietly abandoned. A quoting tool the estimator opens once a quarter to keep someone happy. The money is gone, the old process is still running, and nobody wants to be the person who suggests trying again.
Technology without the technology project: no eighteen-month rollout, no rip-and-replace, no six-figure license signed before anyone has watched the thing work. One process at a time, usually inside the tools you already own.
Be honest about the numbers
Adoption is real, and it is not arriving evenly.
17–20%
of US businesses report using AI in any business function, measured fortnightly from December 2025 to May 2026
37%
of firms with 250 or more employees use AI, against under 20% of firms with four or fewer
Both figures are the Census Bureau’s, not a vendor’s. The size gradient is the part worth sitting with: the companies you compete with for crews, for margin and eventually for buyers are adopting faster than you are, and the gap is widening in the direction of scale.
A note on this number, because we used to publish a different one. The Census changed the question in November 2025, from AI used “in producing goods or services” to AI used “in any business function.” Figures below 10%, which were accurate under the old wording and are still widely quoted, are not comparable to these. We were quoting one of them until 2026-08-10.
Why the gap is that wide
Buying the tool is the easy part.
A subscription can be bought in an afternoon. Getting a crew to change how they record a change order takes months, and it only happens if the new way is faster for the person doing it than the old way was.
That is why adoption surveys and production measurements disagree so sharply. The software was purchased. The work never moved into it. We assume this is the default outcome and design against it: nothing counts as done until the old method has been switched off and nobody has asked for it back.
We also start where the hours are, not where the technology is interesting. Before anything gets built we spend time watching how work actually moves through the office. The process people describe in a meeting is rarely the process running on a Thursday afternoon.
The four we start with
Where the hours actually go.
Quoting and estimating
Usually the most valuable hour in the company, spent by the person who can least afford to spend it. We build the cost model that already exists in his head into something that produces a consistent quote, then let him spend his judgment on the jobs that need it.
Scheduling and dispatch
The board lives in one person’s head, a whiteboard, or a spreadsheet emailed every morning. Making it visible to two people is worth more than making it clever. Automation here means the schedule updates itself when something moves, not that an algorithm decides who goes where.
Follow-up
Open quotes go cold and open invoices age because following up is nobody’s named job. This is the cheapest automation in the building and often the one that pays for the rest of the work in the first quarter.
Data entry between systems
The same job gets typed into the quoting sheet, the schedule, the field app and the accounting package. Every retype is an hour and a chance to be wrong. Most of these handoffs can be connected without replacing anything on either end.
What changes
How we know it worked.
Not by logins. By whether the old spreadsheet is still open on somebody’s second monitor six months later.
The measure is hours returned to the people who were spending them, and whether the person who used to do the task by hand would choose to go back. If he would, we built the wrong thing and we say so.
- One process automated at a time, each one live before the next begins
- The old method switched off deliberately, not left running in parallel
- Hours measured before and after, on the same task, by the same person
- No new license signed until a working version has been used on real jobs
- Every automation documented well enough for your team to change it later
Honestly
What we don’t do.
We don’t run a company-wide software program. We don’t automate a process that is broken. That just makes a bad process faster. And we don’t put AI in front of your customers to save an internal step.
Kevin Aris leads this work at Rusk & Co., on site with partner companies in Houston and across Texas.