Finding the work where AI actually belongs.
The useful place to start is the work your team repeats, the questions it answers again and again, and the handoffs that slow everyone down.
Giving a team access to AI can feel like progress because everyone can open the tool and start using it immediately. But if the same handoffs, repeated questions, and slow decisions are still there a month later, it is worth looking more closely at where the work was supposed to change. A field experiment with 515 high-growth startups found that firms prompted to search across their operations for AI use cases identified more opportunities and saw stronger outcomes than the comparison group. The results belong to those startups, but the question is useful for a smaller business: how do you find the work where AI can make a difference?
At Ruby, we start by finding out how the work actually gets done. We talk with the people who own each step, document the handoffs and recurring questions, and trace where someone keeps repeating manual work. That gives a team a map it can act on. From there, we decide where an AI workflow can remove a step, help someone prepare a better decision, or give an employee time back. We look at whether that change saves time, lowers cost, or improves productivity so the team can see what actually worked.
Think about a question that comes in every week. One person knows where the answer lives, but everyone else has to ask them. They may search through old documents, check whether the answer is still current, and write back. It looks like a small interruption, yet the same sequence happens over and over. If you only give the team a chatbot, the person still needs to know which information it can trust and what to do when the answer is incomplete. If you first document the question, its source, who checks it, and how it reaches the next person, you have something concrete to improve.
That discovery also shows which steps should stay with people. Some need a person's judgment, especially when the information is sensitive or the cost of being wrong is high. Other steps may be repetitive but too rare to justify a new system. The useful opportunities are the ones that happen often enough to matter, have a clear input and output, and leave employees doing work they would rather put toward customers or decisions.
From there, the workflow can be designed around the team. A tool might gather the right information, prepare a first pass, or route a question to the person who can resolve it. Employees should know when to review the result, where to correct it, and how the process will change when something breaks. Building that with the people who do the job is how the workflow becomes usable after the first demo.
The INSEAD study calls the challenge of discovering where and how AI creates value the “mapping problem.” It studied high-growth startups, so its reported performance gains should not be read as a forecast for every small business. Its practical lesson is that finding the right work deserves as much attention as choosing the technology.
For a company with 10 to 100 employees, the first step can be simple: ask which question is answered again and again, which handoff delays a customer or colleague, and which manual step keeps showing up on a busy person's calendar. Document one of those paths, then decide what a better version would save in time or reduce in effort. That is where a useful AI workflow begins.
Sources & further reading
Start with the work. Build the right system around it.
We help teams understand where AI is useful, choose the right tools for the job, and build the confidence to keep improving.
Book a Call