Want to Know Where to Use AI — Follow the Friction

Every leadership team seems to be having some version of the same conversation right now.

“What are we doing with AI?”

It’s a fair question. Competitors are talking about it, employees are experimenting with it, and every week brings another tool promising to change the way we work. So naturally, organizations feel pressure to develop an AI strategy.

We’d argue that’s the wrong place to start.

When a business begins with “How can we use AI?” it tends to go looking for places to put the technology rather than problems to solve with it. A marketing department might sign up for a platform and launch a chatbot on the website because a chatbot feels like the obvious first move, only to realize six months in that almost no one uses it and nobody can say what it actually improved. Leadership ends up asking a different question than the one they started with: “What exactly did this do for the business?”

The companies getting meaningful value from AI aren’t necessarily the ones using the most sophisticated technology. They’re the ones that started with a problem worth solving — an accounting team buried in invoice data entry, a customer service team hunting through three applications to answer a basic question, an operations manager rebuilding the same spreadsheet every Monday morning. Those are business problems, plain and simple, the kind AI tends to be genuinely useful for.

At Swip Systems, we start with the problem our clients are facing, not the technology, and bring in AI only when it’s the right tool for the job.

Start With the Work You Wish People Didn’t Have to Do

The best AI opportunity in your organization probably isn’t hiding in an innovation meeting. It’s more likely sitting inside a process your employees complain about every Tuesday.

Picture the customer service rep who gets a routine question from a client — something she knows the answer to, technically. It’s just not anywhere obvious. It’s buried in an old email thread, or a PDF attachment, or a note someone left in a system she doesn’t usually open. So she opens four browser tabs, searches three different tools, and eventually pieces together an answer that should have taken thirty seconds. She does this a dozen times a day. So does everyone on her team.

One company we worked with wanted to have deeper insights into how their engineers were spending their time. The data was right there in the database, but no one in the company had the time to assemble it on the fly to answer the questions management was asking. We built an AI enabled inquiry tool that they could hand a question to and get an immediate answer. That’s one area where AI shines. Quick analysis of lots of data and a significant time saver.

Individually, these kinds of tasks seem small. A few minutes here, a few minutes there. It’s been part of the job for so long that no one thinks to question it. But multiply it across a department, and then across an organization, and the picture changes quickly. Those small tasks add up to real hours, real frustration, and real cost…even though nothing about them shows up on a strategy slide.

This is where AI becomes genuinely useful, not because it can produce an impressive answer to a clever prompt, but because it can increasingly understand, organize and act on the kind of information traditional software has always struggled with. Historically, a person had to make sense of that kind of information before software could do anything useful with it. AI is starting to change that math.

So the more useful question isn’t “What could AI do for our company?” It’s this: “What are our people doing today that we wish they didn’t have to do?” That’s where we’d start looking, and it gives us the thread that runs through everything else in this piece.

When Information Is Trapped in Documents and Emails

Think about how much of a business actually lives outside its official systems. Signed contracts sitting in an inbox. Vendor invoices arriving as PDF attachments. Intake forms filled out by hand and scanned in. A new client’s onboarding details, typed into an email instead of a database, because email was simply what was readily available at the time.

None of that is unusual. It’s just what happens when real work meets real people. But it also means that somewhere in nearly every company, someone’s job includes opening these documents, finding the handful of details that matter, and typing them into whatever system actually needs them. It’s careful, tedious, easy-to-get-wrong work — and it rarely feels like the kind of thing worth mentioning in a status meeting.

This is squarely AI territory. Not AI that writes something new, but AI that reads what already exists — extracting the invoice number and the amount due, classifying a document by type, summarizing a ten-page contract down to the three clauses that changed, routing a request to the person who actually needs to see it. The work doesn’t disappear. It just stops requiring a human being to do the reading first.

When Employees Spend More Time Searching Than Doing

There’s a particular kind of exhaustion that comes from knowing an answer exists somewhere and not being able to find it. Ask around most organizations and you’ll hear some version of the same complaint: the information is there, technically. It’s just scattered across a shared drive, three different platforms, an old Slack thread and someone’s inbox — and finding it costs more time than the question itself ever should.

A new hire asks how a particular client likes their invoices formatted, and the answer lives in the head of someone who’s been out sick all week. A salesperson needs the pricing exception approved for a similar deal last quarter and has to interrupt three people to track it down. None of this is a failure of effort. It’s a failure of retrieval — the organization knows more than any one person can hold, and its tools were never built to search across all of it at once.

This is the second place AI earns its keep: surfacing what already exists, in context, at the moment someone actually needs it. The value isn’t the novelty of the answer. It’s the minutes and the interruptions that it gives back to everyone else.

When People Are the Bridge Between Systems

Every growing business eventually collects a stack of software that doesn’t quite talk to itself. The CRM doesn’t sync cleanly with the order entry platform. The scheduling tool exports data in a format the billing system doesn’t recognize. Somewhere in between, a person becomes the connection — copying a client’s details from one place to another, reformatting a spreadsheet so a second tool will accept it, double-checking that the numbers still match after the manual handoff.

A traditional integration can solve this when the data lines up cleanly, field for field. But plenty of these gaps are messier than that. The information matches in spirit, not in structure… a name spelled slightly differently, a status update phrased in a sentence instead of a dropdown value, a note that needs to be interpreted before it can be entered anywhere. That’s exactly the kind of ambiguity that’s kept these processes manual for years, because rigid integrations don’t tolerate it well.

AI can sit in that gap in a way conventional automation never could — reading the messier version of the handoff and translating it into something the next system can use, without requiring every field to match perfectly first. That person gets to stop being the connective tissue between two systems that should have been talking to each other all along.

When Decisions Require Too Much Detective Work

Some of the most valuable AI use cases have nothing to do with automating a task and everything to do with helping someone decide faster. Picture a manager reviewing a stack of vendor invoices, trying to spot the one that’s priced differently than usual. Or an operations lead scanning a week of service tickets, trying to notice the pattern that three unrelated complaints are actually the same underlying issue.

That work is possible today. It’s also slow, and it depends heavily on whoever happens to be paying close attention that day. AI is well suited to this kind of pattern-matching… pulling scattered information into one place, flagging the exception, surfacing the anomaly a person would eventually catch, just later and with more effort… or may even miss.

The distinction matters here: this isn’t AI making the decision. It’s AI doing the detective work first, so the human making the call has what they need in front of them instead of buried across five systems.

A Simple Way to Think It Through

Everything above points to the same underlying pattern, and it’s worth naming directly, because it’s a very useful filter for evaluating any AI idea that comes across your desk. Start with the problem… the actual process pain, described the way an employee would describe it, not the way a vendor would pitch it. Look at the process currently used to deal with it, manual workarounds included. Look at the data involved, and whether it’s clean and structured or messy and scattered across formats. From there, the AI opportunity usually becomes clear on its own – often smaller and more specific than “an AI strategy” ever suggested. And finally, be honest about the business value: the time saved, the errors avoided, the capacity freed up for work that truly requires a person.

Problem, process, data, opportunity, value. It’s a solid framework you need going into any consideration for business process automation. It’s one that makes sense once you’ve found the friction, and a way to check your thinking.

The Bottom Line: Don’t Start With AI

Don’t start with AI. Start with the work you wish your people didn’t have to do. Then determine if and how AI is the right way to eliminate it. We feel this approach is considerably more useful to a business than another list of ways to use AI.

If you’re struggling with where and how to implement AI, we love to talk process automation geek speak!