I run Descent IT, a managed services company, and SSCITServices, a small AI practice. You're probably here because someone sent you.
We built a client dashboard where, as we configure their system, a tech can instruct the bot to create the documentation directly. Not afterward. Not from notes. As the work happens. That shift cuts the lag between doing something and having documented how to do it — from weeks to near zero.
Most MSPs calculating AI ROI are measuring labor hours saved. That number is real, but it's the wrong denominator. The correct denominator is capacity ceiling — the point at which you'd have to hire to grow the book of business. That's the number that changes the conversation with ownership.
Weekly project budget reviews used to eat half my Monday. Built a simple workflow that pulls timesheet data, matches it to project budgets, and flags variances above ten percent. Takes ninety seconds now instead of two hours. The interesting part wasn't the time saved — it was realizing how much project drift we were missing because the manual review was too painful to do properly.
Ticket close documentation dropped from an average of 8 minutes per ticket to under 90 seconds. We'd been treating documentation as a minor overhead — a few minutes here, a few minutes there. When you run 200 tickets a week across a team, those minutes are actually a full-time position.
For 20 years I watched techs batch-enter timesheets on Friday afternoon. Guessed hours. Skipped entries. The fix wasn't a better timesheet tool — it was removing the ask entirely. AI generates time entries from screen recordings and activity patterns. Compliance went up because the ask got smaller. Not because we enforced harder.