Intelligent automation can’t fix a bad process

A conversation with business strategy expert Sam Ransbotham

Workflows are the regular tasks and routines that employees perform. They define how work gets done inside any organization. Productivity can nosedive when companies use outmoded technology to help manage their workflows, or rely on ill‑conceived processes to begin with.

Companies can streamline a wide variety of workflows today using business process automation (BPA), robotic process automation (RPA), and machine‑learning algorithms. Whether the workflow falls in IT, sales, manufacturing, or HR, the goal is usually to reduce costs, increase efficiency and create better experiences for employees.

However, automation isn’t the best solution for every process glitch, cautions Sam Ransbotham, associate professor of information systems at Boston College’s Carroll School of Management. For example, many healthcare and insurance companies still rely on image‑recognition software to digitize faxes for billing and electronic records. That is just a Band‑Aid for a terrible process. Companies don’t need to make faxes easier to use; they need to get rid of them.

In a conversation with Workflow, Ransbotham argued that companies need to rethink legacy workflows before rushing to digitize them using process automation and AI.

Given the massive influx of investment in AI, do you think there is too much of a rush to adopt it everywhere?

AI is the “cool thing” certainly. There’s a shiny new tool aspect of it, but that doesn’t mean that it’s the right tool for everything. For example, I did an interview with Airbus last year. They said one thing that I really like: “We don’t invest in AI. We have processes that use AI, but that’s because it was the best choice for solving a problem.”

You say that AI can often be a Band‑Aid for a bad process. What does that mean?

A Band‑Aid just masks a symptom. It doesn’t actually heal the cut. For example, think about a business process you can make more efficient through AI or automation, like using chatbots for customer service.

Sure, you can cut costs this year and next year, but what if you flip the problem and ask if there’s a completely different way of customer interaction that AI would enable, but you don’t do it because you’re focused on the short‑term solution? Well, you may have missed out. People with a long‑term approach are thinking less about cost savings now and more about what new process AI could enable—because the opportunity for a new way of doing things might not come again.

What’s an example of a bad process that AI might hide rather than fix?

Everybody schedules meetings and, let’s face it, there’s too much time spent on scheduling them. Too many emails flying back and forth between too many people.

Today there’s a lot of AI emerging around automating meeting scheduling. That’s a good thing, right? I’m not so sure. Given the supply‑and‑demand curve, making scheduling easier means that the number of meetings will increase, not decrease, and you’ll just end up with more meetings. I agree with using AI to make the scheduling process smoother and easier, but it doesn’t address the real problem—which is that there are a lot of meetings that simply don’t need to happen.

Instead, AI could be part of the solution. Better analytics could shorten discussions. AI could even take over routine decisions so we don’t need to have meetings about them. Meetings could become the exception rather than the rule, so that we only meet when a decision is too complex for AI to handle.

Why are legacy workflows so hard to let go of?

Change is always hard. A significant shift in business processes entails risk. A safer approach, in the short term, is to make a series of small improvements to a process. But processes rarely, if ever, exist in a vacuum—they affect all sorts of other processes.

When you decide to make incremental changes, it can insulate related processes from change. AI can work well here by optimizing legacy processes in ways that reduce risk and keep other processes in place, but the benefit is short term rather than longer term. While you’re making your safe, incremental improvements, a competitor may be making riskier, more fundamental improvements.

How should companies audit existing processes to know where AI and automation makes sense?

You first need to step back and ask, “Does this process need to exist?” It’s the difference between getting to the top of a local hill versus a global hill. You can use an algorithm to get to the top of a local hill faster than your competitors—by, say, getting a chatbot up and running for customer service.

Or you can use AI to figure out what the biggest hill is before you start climbing, by investing in big data analytics to predict market trends. For example, how do you meet demand for new products in emerging markets? Step back before you start optimizing a legacy process. Ask, “Do we need to take this approach?” Just because you’ve always done it one way doesn’t mean you need to optimize doing it that way.

Think about where industry competition will be in the future. You can continue to hone your costs and improve services alongside existing competitors, but someone else may come along and say, “Hey, we can do it different and better” without the baggage of legacy processes.

It’s important to step back and say, “If I were coming into this industry new and fresh, would I even try to automate this existing process?” If the answer is no, then somebody else is going to reach that same conclusion and become your latest competitor.


Related Articles:
Column

Process mining 101

Want to analyze and troubleshoot core business processes? There’s an app for that.
Flow State

Solving the productivity paradox

The digital revolution was supposed to drive prosperity by making us all more productive. What happened?
Column

4 ways AI can power IT asset management

AI can help reduce IT costs through better demand forecasting, asset retirement and other capabilities