Before You Add AI, Ask Why

I began my working life before computers became part of the everyday workplace.

We had paper. We had mimeograph machines and teletypes. We had calculators and thirteen-column pads. The work got done, but much of it was done manually.

Then computers began to arrive.

Some of the earliest business applications were in finance. That made sense. Finance was already governed by established rules, accounting practices and reporting requirements. The work was structured. In many cases, computerization meant taking a well-defined manual process and making it faster and more reliable.

Material requirements planning (MRP) followed a similar path. Manufacturers already had to determine what materials were needed, how many were needed and when they had to be available. Once again, computers provided a better way to manage a process that was already reasonably well defined.

As computers moved deeper into business operations, however, something interesting happened.

The usual approach was to document everything people were doing manually. We collected the forms. We mapped the workflows. We recorded how information moved from one person or department to another.

Then we created computer systems that did the same things.

At the time, that seemed perfectly reasonable. If people were performing a task manually, the obvious opportunity was to have a computer perform it. We could move the information faster, reduce some of the paperwork and perhaps eliminate a few errors.

But some of us began raising a different question:

Why are we doing this in the first place?

Instead of beginning with the existing process, we wanted to begin with the purpose of the operation and the capabilities of the new technology.

What was the organization actually trying to accomplish?

What could this new resource do?

Given these new capabilities, was there a better way to accomplish the objective?

Those questions were not always welcomed.

Processes become familiar. Responsibilities grow around them. Departments, jobs and systems are built to support them. Eventually, the fact that “this is how we do it” becomes its own justification.

It took time for organizations to recognize that computerizing an old process was not the same as improving it.

When the Pendulum Swung

Then the pendulum began to swing in the other direction.

As large enterprise systems became common—and especially as organizations prepared for Y2K, now a full generation back—the software increasingly began to dictate the process.

Instead of programming a system to reproduce everything the organization had always done, organizations frequently changed the way they operated to match the system they purchased.

That solved some problems, but it created others.

First, we made technology conform to every inherited business practice. Then we made the business conform to the technology.

Neither approach necessarily began with the most important question:

What are we trying to accomplish?

The answer does not come from blindly preserving the old process. It does not come from blindly accepting the process built into a new system.

It comes from understanding the objective, examining the entire operation and then deciding how the technology can best help us achieve the desired result.

We Are Repeating the Pattern With AI

Today, I see organizations beginning to travel down the same road with artificial intelligence.

They are adding AI to existing tasks. AI writes the document, summarizes the meeting, answers the question, reviews the information or completes another step in an established workflow.

There is value in this experimentation. People need opportunities to use AI before they can understand what it might mean for their work. Organizations also need to begin somewhere.

But adding AI to an existing process is only the beginning. It is not transformation.

The first question being asked is often:

How can AI do this task faster?

A better set of questions would be:

Why does this task exist?

What outcome are we trying to produce?

Does the process still make sense?

What could we do differently now that this capability is available?

Could the work be simplified, redesigned or perhaps eliminated?

If we fail to ask those questions, we may use one of the most powerful technologies we have ever created to preserve work that no longer needs to be done.

We may perform the wrong task faster.

We may produce information no one needs in a fraction of the time.

We may automate one department’s work while creating more problems for another department.

We may make an individual part of the process appear more efficient while making the total process worse.

Someone Must See the Whole Process

The challenge is that most people naturally see the part of the process closest to them.

A department understands its own work, its own problems and its own objectives. A software provider understands the capabilities of its product. An AI specialist may understand the technology. Organizational leaders may understand the results they want to achieve.

But who is looking at the entire process?

In my experience, that responsibility often belonged to engineering. Engineers were trained to examine how work moved through the complete operation. They looked for constraints, duplicated effort, unnecessary steps and problems that had simply become accepted over time.

The objective was not merely to make one task more efficient. It was to improve the performance of the total system.

That responsibility may be organized differently today. It may belong to engineering, operations, systems engineering, continuous improvement, business process management or a transformation group. It may require a cross-functional team rather than one department.

The name is not important.

The perspective is.

Someone must be trained and authorized to step back from the current procedures, examine the total flow of work and challenge the assumptions on which that work is based.

That person or group must apply the same standard: a local improvement is not automatically an organizational one.

If one department uses AI to produce information ten times faster, but another department must still review every item manually, what has been improved?

If AI eliminates a task but leaves all the approvals and delays surrounding that task in place, has the process really changed?

If a company automates an outdated report, should it celebrate the time saved—or ask why the report still exists?

Without someone examining the whole system, AI adoption can become a collection of disconnected improvements. Each one may look successful on its own while the organization as a whole becomes more complicated.

Start With the Objective

This does not mean organizations should wait until they have designed the perfect AI strategy.

The technology is developing too quickly for that.

Ideas that seemed advanced only a short time ago can quickly become unnecessary. Techniques that once required specialized knowledge are becoming easier to use. New capabilities continue to appear, and people continue to discover applications that were not obvious before.

Organizations need to experiment. They need to learn what AI can do. They need to allow employees to explore possibilities.

But experimentation should not be confused with direction.

A pilot project may help us understand a capability. It should not automatically become a permanent process.

An AI tool may provide a faster way to complete a task. That does not mean the task should remain unchanged.

A successful implementation today may need to be reconsidered as the technology develops.

This is one of the most exciting things about AI. Unlike traditional systems that could hard-code a process into an organization for years, AI gives us the opportunity to remain flexible. We can test an idea, learn from it and change direction.

But that flexibility will not happen automatically.

Organizations can turn an AI experiment into a rigid procedure just as easily as they did with earlier technologies. Once responsibilities, policies and performance measures grow around a new process, it can become difficult to challenge—even when the technology has already moved beyond it.

Someone must keep asking whether this is still the right process, the right tool and the right problem to be solving—or whether a new capability has quietly created a better option.

The View From the Tree

There is an old observation about leadership. While everyone else is busy cutting a path through the forest, the leader climbs a tree, looks around and says, “We are headed in the wrong direction.”

AI will help organizations move through the forest faster than ever.

That makes the view from the tree even more important.

The organizations that benefit most from AI will not necessarily be the ones that adopt the most tools or automate the largest number of tasks.

They will be the ones that remain clear about what they are trying to accomplish.

They will examine the total process rather than optimize isolated tasks.

They will bring together people who understand the operation, the technology and the broader system.

They will be willing to question processes that have been accepted for years.

And they will use AI to reconsider the work, not simply accelerate it.

The question is not merely what AI can do.

The question is what we should do now that AI is here.

Dan Holland has been in management consulting since 1982, including twenty years leading his own independent practice. Today he works directly with owners and leadership teams at mid-sized companies on the decisions, plans, and organizational problems that deserve more than an internal gut check. No sales calls, no funnels — just a direct reply from Dan at [email protected].