AI Operating Philosophy · 6 min read · By Adam Golob · September 30, 2026
A customer asks for a copy of last month's invoice. Finding it and emailing it takes about five minutes. The customer gets it two days later. Nobody refused. Nobody forgot on purpose. The request just waited, over and over, between people who were each busy with something else.

Work time is the minutes someone actually spends doing the task. Elapsed time is how long the task takes from request to done, as the customer experiences it.
Owners tend to manage work time. How long does a quote take to write? How long does an invoice take to send? Customers only feel elapsed time. In most small businesses the gap between the two is wide, because the work is quick and the waiting is long. That gap, added up across every workflow, sets your company's clockspeed.
Here are five workflows almost every service business runs. The times below are illustrative examples, not measurements from any particular company. See if any of them sound familiar.
Pricing the job takes maybe 15 minutes. The request arrives by email Friday evening and gets read Monday. It needs one measurement, so someone emails the customer and waits. The answer comes Tuesday. Then the number waits for the owner to glance at it. Elapsed time: most of a week.
Scheduling takes ten minutes. The call hits voicemail at lunch. The callback lands in the customer's voicemail. The second callback books a slot the next day. Ten minutes of work, spread across two days of phone tag.
Replying takes two minutes. The web form lands in a shared inbox that three people each assume someone else is watching. By the time anyone answers, the lead has talked to a competitor. We cover this one in depth in why speed to lead decides who wins the job.
A purchase order needs a signature. Signing takes thirty seconds. The manager is out until Thursday, so the order sits, and so does the job that depends on it.
The job finishes Tuesday. The tech's notes make it back to the office Wednesday, get typed in Thursday, and the invoice goes out Friday. The payment clock starts four days late, before a single reminder has been sent.
None of these gets fixed by working faster. The five minutes is already fast.
Most process improvement goes after the task: a better template, a faster tool, a shortcut key. Useful, but it shrinks the smallest number on the page and leaves the biggest one alone. The leak is in the queues between tasks. Every handoff creates one. Every "I'll get to it" creates one. Every request that shows up after hours creates one that lasts until morning.
That is why our method for mapping and collapsing workflows starts with elapsed time. If you want to change a company's clockspeed, look where the clock keeps running while nobody is working.
No software required. A spreadsheet and an hour will do.
Then look for patterns. Does the gap grow on weekends? After 4 p.m.? When one particular person is out? Which step has the longest wait in front of it? That step is your leak. It is usually a handoff or an approval, and it is usually boring. Good. Boring is fixable.
Paper trail, four days
Invoiced the same hour
Where people stay: disputes, discounts and payment plans go to a person with the job history attached.
In the traditional lane, the notes travel back to the office, someone retypes them, someone checks them, the invoice goes out, and a reminder happens whenever someone remembers.
In the AI-native lane, the tech closes the job in the field app or by voice. An agent builds the invoice from the job record and your price list, sends it within the hour and runs the reminder schedule on its own. That is what connected automation work looks like in practice: the systems you already use, talking to each other with nobody in the middle retyping.
People still set prices, settle disputes, approve anything unusual and talk to the customer who is upset. The goal is to take people out of the waiting, not out of the work that needs judgment. When an exception shows up, it should arrive with the full history attached, so nobody has to go digging through three inboxes to understand it.
When elapsed time shrinks toward work time, several numbers move together: response time, completion time, handoffs per job, the volume your current team can carry and how quickly finished work turns into cash. Our page on AI operational efficiency covers how we track those after a build. The test is simple. Pull the same timestamps again a month later and see whether the gap actually closed.
The AI Audit times your real workflows step by step and shows exactly where work waits. It is a paid two-week engagement, and the $1,000 fee is credited toward your first build.
Book the AI Audit → Talk it through first →Work time is the minutes someone spends actively doing a task. Elapsed time is how long it takes from request to done, including every wait in between. Customers experience elapsed time, so that is the number that shapes whether they come back.
Ten recent examples are usually enough to see the pattern. You are not running a formal study. You are looking for the step where work waits longest, and that tends to stand out clearly once the timestamps sit side by side.
Yes. Remove steps that do not need to exist first, like duplicate data entry or an approval nobody really reads. Then automate what is left. Automating a broken process mostly makes the broken parts happen faster, which rarely helps anyone.
Related: Don't Automate a Broken Workflow · Your Business Has a Clockspeed. AI Can Change It.