· AI
Before You Automate a Task, Price In the Time You'll Spend Minding It
The question is not whether a task is boring enough to automate. It is whether you have priced in the hours you will spend watching the automation.
The task a business owner wants to automate first is almost always the one driving them crazy right now: the inbox, the follow-ups, the same quote retyped for the tenth time. That instinct is not wrong, exactly. It is just missing a line item. Every automated task still carries a labor cost after it ships, in the hours someone spends feeding it context, checking its output, and fixing what it got wrong. Skip that line item and the project can look like a win on the surface while losing money underneath.
My position: before you automate anything, estimate the supervision time it will require, and treat that number as part of the real price, not an afterthought.
The case for just automating the annoying thing
I want to state the other side fully, because it is not a strawman. If a task eats five hours a week and you are certain it is driving you up the wall, waiting for a perfect readiness assessment before touching it has a cost too: five more hours lost every week you delay. Most of what a small business automates really is routine enough that some supervision is fine, and a partial win now beats a theoretical perfect win later. Speed is often the cheapest way to compete, and a business that waits for certainty on every task gets outrun by one that does not.
What the supervision time actually looks like
The trouble is that the supervision cost is not small, and it is not evenly distributed across tasks. Glean's Work AI Institute surveyed about 6,000 full-time digital workers across the US, UK, and Australia between December 2025 and January 2026 for its Work AI Index 2026. Eighty seven percent of workers use AI at work, and they believe it saves them about 11 hours a week, over a quarter of a standard workweek. That is the number that gets quoted. The number that gets left out, reported independently by IT Brief UK, is that the same workers spend 6.4 hours a week on what the report calls botsitting: supplying context the tool lacked, supervising its output, debugging mistakes, and cleaning up afterward, which the report says adds up to 37 percent of all AI-related time. Line those two self-reported figures up and a rough, honest estimate of the real weekly gain drops from 11 hours to closer to 4 or 5, not the quarter of a workweek the headline implies. Both numbers are self-reported perception, not logged minutes, so treat the exact figures as directional. The direction is still the point: the sticker number and the take-home number are not close to the same thing.
It gets worse before it gets better. The same report found 69 percent of AI users admitted to what it calls botshitting: 41 percent sometimes delivered AI output they could not actually explain, and 28 percent blamed the AI for a mistake that was their own. Only 13 percent said AI had made their organization perform significantly better, despite nearly nine in ten using it. That gap, between "I personally save time" and "the organization is measurably better off," is exactly where an unpriced supervision cost hides. One honest detail in Glean's own data: workers with better context built into their tools cleaned up after AI about a third less often than workers without it, and delivered unexplainable work about half as often. The supervision tax is not fixed. It depends heavily on how much the tool actually knows about your business before you ask it to act.
What this means before you build anything
None of this is an argument against automating. It is an argument for estimating supervision time with the same seriousness you would estimate the subscription price, before committing a task to it. A task with a narrow range of inputs and few ways to go wrong, a quote template, a scheduling confirmation, costs little to supervise because there is little to check. A task with wide variance, an inbox that spans angry customers, new vendors, and one-off requests, can cost you most of the time it claims to save, because every unusual case still needs a human to catch it. The same automation can be a clear win on one task and a wash on another, and the only way to know which is to estimate the babysitting time honestly before you build it, using your own experience with how often that specific task throws something unusual at you.
That estimate, task by task, is the actual first step of the readiness work behind MojoAI's automation practice, because a tool that looks cheap on the invoice can still be expensive on the clock. If you want a second set of eyes on whether a specific task will pay for its own supervision, that conversation is worth having before you buy the subscription.
Sources
References used in this article. Links also appear alongside the relevant claims.
Let's make it happen.
You don't need a polished brief. A couple of lines about where you want your company to go is plenty, and we'll come back with what we'd tackle first.
We scope the work around your goals and timeline before quoting anything, so you know exactly what you're getting.