Once a business decides to "use AI," the instinct is to look for the most impressive thing to build. That's backwards. The best first automation is almost never the flashiest, it's the most boring, frequent, and predictable task you do, the one nobody enjoys and everybody forgets.
Picking well is the whole skill. Automate the wrong task and you spend money making something rare slightly faster. Automate the right one and you hand a full day a week back to your team. This is a simple way to find the right ones.
The two questions that matter
For any task, ask two things:
- How often does it happen? Daily and hourly tasks compound. A two-minute job done 50 times a week is more valuable to fix than a two-hour job done once a quarter.
- How rule-based is it? Could you write down the steps so clearly that a new hire could follow them without judgment calls? The more "if this, then that" a task is, the cleaner it automates.
High frequency plus high predictability is the sweet spot. That combination is also exactly where the time goes: in workplace surveys, the tasks that eat the most hours are email, data collection, and data entry, the definition of frequent and rule-based.Smartsheet, automation in the workplace survey. smartsheet.com
Green flags: automate it
A task is usually a strong candidate when it's:
- Repetitive and high-volume: done many times a day or week, the same way each time.
- Rule-based: the decisions follow clear logic, not gut feel.
- Time-sensitive: value drops if it's slow (replying to a lead, answering a call, sending a reminder).
- Error-prone by hand: manual data entry and re-keying where a typo costs you.
- Stuck between tools: anywhere a person copies information from one app into another.
Red flags: leave it alone (for now)
Automation isn't free, and some work resists it. Be cautious when a task is:
- Rare or one-off: building a system to do something twice a year rarely pays back.
- High-judgment: strategy, negotiation, sensitive calls. AI can assist, but it shouldn't be the decision-maker.
- Relationship-critical: the human touch is the value; automate the prep around it, not the moment itself.
- Constantly changing: if the process is redefined every month, lock it down before you build on it.
An honest detour: build vs. buy
Before you build anything custom, ask whether an existing tool already solves it. A huge amount of common work, scheduling, email marketing, basic CRM, e-signatures, is handled well by off-the-shelf software you can turn on this afternoon. If that's your situation, the right move is to buy it, not to commission a build.
The goal isn't to build the most. It's to fix the most, for the least.
Custom work earns its keep in the gaps: when your process is specific enough that no tool fits, when several tools need to talk to each other in a way they don't out of the box, or when the volume is high enough that a tailored system pays for itself quickly. That's the honest line we draw with clients, and it's the line worth drawing for yourself.
Start where it's already breaking
If you want a shortcut past the framework, start where you can already feel the pain:
- Lead response: anything that makes a new lead wait. Speed has an outsized effect on whether they ever convert (see Speed-to-Lead).
- Missed inbound: calls and messages that go unanswered while you work (see The True Cost of Missed Calls).
- The weekly report: data someone assembles by hand every week that a pipeline could generate on its own.
- The handoffs: the moments work passes between people or tools and things fall through.
The takeaway
You don't need to automate your whole business. You need to find the two or three tasks that are frequent, predictable, and quietly expensive, confirm a tool doesn't already solve them, and fix those first. Do that and the next opportunities tend to reveal themselves.
If you'd rather not guess, that prioritization is the entire point of a free AI audit: we map your operations, score the opportunities, and tell you plainly which are worth building and which aren't.
Sources
- Smartsheet, automation in the workplace survey: email, data collection, and data entry are among the manual tasks consuming the most worker time; over 40% of workers spend at least a quarter of their week on repetitive manual work. smartsheet.com
- McKinsey & Company, "The economic potential of generative AI": value concentrates in everyday functions, customer operations, marketing and sales, software engineering, and R&D, rather than exotic use cases. mckinsey.com