Start With the Stuck Work, Not the Software

Automating unclear work spreads mistakes faster and makes it harder to see who is responsible. To decide what to automate first, start with the work that is blocking a result you care about, then ask whether it needs simpler instructions, a clear owner or technology. Customer follow-up, scheduling, records and coordination are the usual places to look, but none belongs at the top of every business's list.

Find what is actually blocking the work

An operating constraint is the obstacle that limits the result you want: serving customers reliably, finishing work without repeated clarification, or giving the owner room to manage. Naming it ties a technology conversation to the business it's meant to help.

An operator who has bought and led businesses and has a long background in healthcare went looking for a healthcare services business where technology would change what limits growth. Adding locations and clinicians is one way healthcare businesses grow, so the operator wanted a model where technology could reach more patients. Replacing an old tool wasn't the point. The same test works in any service business: does this tool move the thing that stops the work?

Picture a service business whose customers keep asking for updates. A messaging tool makes sending updates easier. But if staff can't tell whether a job is ready, the obstacle is an unclear job status, and faster messages leave it in place.

The same complaint can have different causes. Customers may be waiting because nobody owns the reply, because the answer needs a manager's judgment, or because someone has to dig through scattered notes. All three get called "slow follow-up", and each needs a different fix.

Ask the people doing the work what stops it. Listen for missing information, conflicting instructions and decisions still sitting with the owner. The manual work audit walks through that conversation.

These are the places to look first. The order isn't a ranking.

The table can't tell you which matters most in your business. It moves you from a general wish to automate toward a specific account of what a customer or employee is going through.

Name the work and its owner

"Automate the back office" leaves too much undefined. "Customers can't tell who is handling their request" gives the team something to discuss. Describe where the work starts, what information people need and what a good result looks like.

The owner is the person responsible for the result. They may not do every task, but they know when the work is incomplete, what authority they have and where a question goes when they can't answer it.

Ownership matters even when the software works. A reminder can arrive on time while the request behind it stays unresolved. A record can be copied perfectly and still be out of date. The action and the result need separate attention.

A standard operating procedure describes how recurring work should be done. It helps only if it reflects the work people actually face, including the cases that need judgment. An instruction to send a routine appointment reminder can be clear. A customer whose circumstances have changed needs a conversation. Treat those as the same work and the system looks consistent while the service gets worse.

Ask the employee who does the work what an outsider would miss: a customer preference, an informal promise, a dependency that isn't written down. That tells you whether the work is understood well enough to discuss a tool at all. If the team can't say who handles an unresolved request today, automating the handoff won't settle it.

Try the simpler fixes first

A work problem can deserve attention without needing new software. Dropping an unnecessary request, clarifying an instruction or using a feature you already pay for may clear the obstacle.

Back to the customer updates. If staff mean different things by "ready", agreeing on the meaning may help more than a message generator. If the meaning is clear but hard to find, an existing shared record may do. If someone has to interpret an unusual situation, a person stays central to the answer.

Compare options against the obstacle. Can the existing tool show who's responsible? Would simpler instructions remove a handoff? Would connecting two existing systems stop repeated copying? What work is left for staff after the change? Build, buy or integrate software covers those choices from the work outward.

Count what it takes to live with the choice. Someone maintains instructions, fixes records, supports colleagues and explains changes. A tool that removes one task can create a responsibility somewhere else, and that ongoing owner attention is part of the cost.

Keeping a process that works is a legitimate outcome. So is deciding the work has to change before any software decision makes sense.

AI use is not the same as operating value

Using AI is an activity. Operating value is whether the business works better for customers, staff or the owner. An AI-written summary can be convenient and still leave a job waiting on the same missing decision. Before scaling whatever you pilot, check whether it actually paid for itself; feeling faster is not the same as being faster.

Predictability is a better test. Can people find the status they need? Does a routine request reach the right person with less confusion? Do staff still have to reconstruct missing context before acting? Those questions are about the work, whether or not AI is involved. An AI chief of staff can support coordination, but the title gives it no management authority; someone still interprets what it produces.

Adoption surveys show interest in AI; they can't pick your next change. The US Census Bureau's Business Trends and Outlook Survey reported that overall AI use by US businesses hovered between 17% and 20% from December 2025 to May 2026¹. That covers all US businesses and says nothing about the payoff from any particular task.

In the Goldman Sachs 10,000 Small Businesses Voices survey, only 14% of 1,256 small business owner respondents, surveyed from January 27 to February 4, 2026, said AI was fully embedded in core operations².

The two surveys cover different populations, periods and questions, so don't compare the figures directly. Neither tells you what automating customer follow-up or scheduling would do in your business. Reading small business AI adoption statistics goes into why.

Choose one question to investigate

A certified business guide tells owners to name the game they're trying to win and the barrier in the way before choosing an operating system. The same goes for automation: a good next question names the result you want and what you don't yet understand about getting there.

"Which AI tool should we buy?" leaves the problem unstated. "Why can't staff tell customers whether their job is ready?" gives the people involved something to investigate, and leaves room for the answer to be information, ownership or judgment.

Questions for that conversation:

• What does the customer or employee experience when this work gets stuck?
• What result would make the work more useful?
• Who owns that result and can explain what's blocking it?
• What do the records show, and what's unclear?
• Would clearer instructions or a tool you already have fix it?
• Where does the work need a person's judgment?
• What does the team need to understand before choosing a change?

An open question is useful when it shows you what you don't know yet. A clear description of the obstacle gives the team a shared subject and makes the next conversation specific.

Write down the one piece of work that's stuck, who owns the result and what you still need to find out.

Source notes

• US Census Bureau, Business Trends and Outlook Survey, America Counts story¹.
• Goldman Sachs 10,000 Small Businesses Voices survey².
• Guest remarks are paraphrased from interviews with a healthcare operator and a business guide; examples are our own.