Use AI where it actually makes sense

Don't start by buying an AI tool.

Start by looking at where your company is losing time. What are employees doing repeatedly? What takes longer than it should? Where are people manually moving information? What reports take hours to prepare? What information is difficult to find?

Those are the places I would investigate first. Once we identify the problem, we can determine whether AI, traditional automation, a process change, or a combination of them makes the most sense.

I look for tasks that are repetitive, predictable, and consume more employee time than they should.

But frequency alone isn't enough. We also need to consider how much judgment the task requires, what happens if something goes wrong, what information is involved, and how much time automation would actually save.

The best first automation usually isn't the most impressive one. It's the one that solves a real problem with reasonable effort and risk.

It can, but I wouldn't promise savings simply because AI is involved.

The savings have to come from somewhere measurable. Maybe employees spend fewer hours preparing reports. Maybe information can be processed faster. Maybe managers get information earlier and can make decisions faster.

That's why I want to understand the current process first. If we don't know how much time or money the problem is costing today, it's difficult to know whether the solution is actually an improvement.

Maybe. First I would ask why it's manual.

Sometimes a manual process is a great automation candidate. Other times the company is performing unnecessary work that should simply be eliminated.

I don't want to spend time and money automating something the business shouldn't be doing in the first place.

That's not how I approach AI implementation.

I'm much more interested in identifying work that prevents employees from spending time on responsibilities that actually require their experience.

If a project manager is spending hours manually assembling information for a report, I would rather find a way to reduce that administrative work so the project manager can spend more time managing the project. The objective is to use people's time better.

You should at least know how they're being used.

Employees may use AI for perfectly reasonable purposes, but management should understand what company, customer, or confidential information could be entered into external systems. There should also be expectations around accuracy and human review — AI can produce information that sounds convincing and is still wrong.

The question isn't necessarily whether employees should be allowed to use AI. It's whether the company understands how it is being used and has reasonable controls around that use.

I would compare the expected value with the total effort and cost.

How much time will it realistically save? How often will employees use it? Does it replace something you're already paying for? Does it require significant setup? Who will maintain it? Will employees actually adopt it?

If a tool costs more to implement and manage than the problem it solves, I don't consider that a successful automation.

Maybe not.

One of the first things I would look at is what you already have. Sometimes companies buy new software without realizing their existing platforms already have capabilities they're not using.

Other times the real opportunity is connecting existing systems or improving the process between them. Another subscription shouldn't automatically be the answer.

Potentially, and reporting is one area I would look at closely if employees spend significant time manually gathering and organizing information.

But first we need to ask whether the report itself is useful. There's no reason to automate a twelve-page report nobody reads.

I would first determine what information management actually needs. Then we can look at how that information is collected, prepared, and delivered.

Yes, particularly with some of the administrative and information-heavy parts of project management.

It may help organize information, summarize updates, support reporting, work with documentation, or assist with other repetitive tasks.

But AI doesn't understand your project the way an experienced project manager does. I see it as a tool that can reduce some of the administrative burden so project managers can spend more time managing risks, people, decisions, and outcomes.

Potentially. Internal knowledge is one of the areas where AI can be useful because companies often have years of valuable information that employees struggle to find.

But access needs to be designed carefully. Not every employee should necessarily have access to every document simply because an AI system can search it.

Information structure, permissions, confidentiality, and accuracy all need to be considered as part of the solution.

Automation generally follows defined rules: when something happens, the system performs a particular action.

AI can work with information in ways that aren't always based on a simple fixed rule — for example, summarizing information, classifying documents, analyzing text, or identifying patterns.

A business process may use one, the other, or both. I don't think companies need to worry too much about which label sounds more advanced. They need to know whether the solution works.

No.

Some processes require judgment, flexibility, or human interaction. Other processes don't happen frequently enough to justify the effort required to automate them. And sometimes maintaining an automation becomes more work than performing the original task.

The goal isn't maximum automation. It's useful automation.

Start small.

Find one recurring problem that consumes meaningful time and determine whether it can be improved. You don't need to transform the entire company at once.

A small automation that saves several hours every week can be more valuable than an expensive AI initiative that nobody actually uses. Once something works and produces measurable value, then you can look for the next opportunity.

I would establish what the current process costs first.

How often does the task happen? How long does it take? Who performs it? What do mistakes or delays cost? Then compare that with the implementation cost, ongoing cost, and expected improvement.

Saving five minutes doesn't sound significant until you discover the same task happens hundreds of times every month. That's why I prefer using the actual process and data instead of assuming an automation will create value.

Then we haven't solved the problem.

A technically impressive solution that nobody uses isn't a successful implementation. I would want to understand why adoption is low. Does the tool make the process harder? Was the problem misunderstood? Were employees involved? Does it require extra steps? Does it actually help them?

Technology has to fit into the way people work, or there needs to be a very good reason for changing the way they work.

You may need a strategy, but I wouldn't start by creating a large AI plan just because AI is currently a priority.

I would first understand where AI could realistically support the company's business goals. Once there are several meaningful opportunities, then it makes sense to prioritize them, establish guidelines, consider risks, and create a roadmap.

The strategy should come from the business needs. The business shouldn't be reorganized around the technology.

I start with your business, not an AI product.

We look at the processes creating the most manual work, delays, errors, or frustration and understand how they're currently being performed. Then we determine what should be eliminated, simplified, or standardized before considering automation.

Once we have a good process, we can identify where AI or automation could create measurable value and evaluate the effort, cost, and risk involved. Sometimes the answer is AI. Sometimes it's traditional automation. And sometimes the best solution is simply fixing the process.