AI is everywhere these days—employees are using it to write emails, summarize meetings, create spreadsheets, and draft presentations.
As an operations expert, I am all about individuals saving time.
However, as a business consultant, I am much more interested in using AI to make entire organizations more efficient.
If your team spends hours waiting for approvals, chasing missing information, fixing errors, or redoing work that someone else already did, AI might make you faster—but it also might create a whole new set of problems at a much faster rate.
True opportunities for increased productivity and outcomes come through:
Figuring out how work actually gets done,
Identifying what is slowing it down, and
Redesigning the process so people and technology can work together more effectively.
The Problem With Automating a Mess
Imagine a company where the CEO approves every purchase over $100. Every request goes through email, the CEO asks follow-up questions, the employee provides more information, and then everyone waits for approval.
Someone introduces AI to help employees write better purchase requests.
Now, the requests are beautifully written, but they are also still sitting in the CEO’s inbox.
The problem wasn't the writing—it was the approval process.
Alternatively, consider a growing real estate company that uses AI to respond to tenant maintenance requests. The responses are quick and professional, but nobody has clarified who assigns the work to contractors, who confirms the repair is complete, or who follows up when the contractor doesn't show up.
Now, tenants can receive faster updates about repairs, but the repairs still aren't getting done.
AI didn't cause these problems, but it also didn't solve them.
Operational efficiency depends on the entire workflow, not just the individual tasks within it.
Research highlighted in Harvard Business Review's article, For Real Performance Gains with AI, Start by Mapping Your Workflows, reinforces this point. In research involving 515 companies, organizations exposed to examples of how other businesses redesigned work around AI identified more use cases and achieved better business outcomes.
In other words, the biggest gains come from rethinking how work happens, not simply adding a new tool to the process.
Step 1: Map the Work Before You Automate It
Most leaders know what their teams are supposed to accomplish, but fewer know exactly how work moves from beginning to end. In high growth organizations, processes often develop organically—someone creates a spreadsheet, another person starts sending emails, and a third person builds a workaround because the original system doesn't quite work.
Eventually, the workaround becomes the process. Nobody remembers why it exists, and nobody wants to change it.
Before investing in AI tools or automation, you should take time to understand what actually happens.
Start here: choose one important workflow—examples include hiring an employee, enrolling a student, responding to a customer inquiry, or completing a property repair.
Then, document:
The starting point: What triggers the work?
The steps: What happens between the beginning and the end?
The owners: Who is responsible for each step?
The handoffs: Where does work move from one person or department to another?
The decisions: Who approves, reviews, or determines what happens next?
The delays: Where does work sit waiting for someone or something?
The rework: Where do mistakes, missing information, or unclear expectations cause people to start over?
The outcome: How do you know the work is actually complete?
Don't document the process as leadership thinks it works. Ask the people doing the work to walk you through it.
You may discover that the official process and the real process are two entirely different things. This is not necessarily a failure of your employees—it is often a sign that the organization has outgrown its original systems.
Action item: Ask one employee to walk you through a recurring task from beginning to end. Write down every step, including the emails, spreadsheets, approvals, and workarounds. Look for the places where time and effort stack up.
Step 2: Find the Bottleneck Before Buying Another Tool
Once you've mapped the workflow, look for the biggest constraint.
A bottleneck is the point where work slows down, piles up, or can't move forward without something else happening first.
Common examples include:
One executive who must approve every significant decision.
Employees who can't proceed because they don't have access to the right information.
Multiple systems that don't communicate with one another.
Unclear roles where two people do the same work.
Manual data entry that creates errors and requires additional review.
Meetings that exist primarily to figure out what happened since the last meeting.
A lack of documented standard operating procedures so employees must reinvent the wheel every time.
AI can help with some of these problems—for example, an AI tool can prepare a report in seconds. However, if the leader takes three weeks to review that report, the report-writing step was never the primary constraint. Often, improving processes and outcomes require changes to decision-making authority, staffing, communication, or process design.
A better solution might involved delegating approval authority, establishing clear decision thresholds, or eliminating an unnecessary review.
In other words, sometimes the best technology investment is a better operating model.
Action item: For your chosen workflow, ask these three questions:
Where does work wait the longest?
What causes employees to redo work?
What single change would allow the rest of the process to move faster?
Start there, and remember you don't need to automate everything at once.
Step 3: Redesign the Process Around People and AI
Now that you understand your process and its constraints, you can evaluate where AI can actually help.
But remember—not every task can be improved by AI. Some tasks need a checklist. Some need clearer instructions. Some need a software integration. Some need a human being to make a judgment call.
Your goal should be to match the work to the right solution.
Here are a few examples:
AI may be useful for:
Drafting routine communications.
Summarizing lengthy documents or meetings.
Organizing information from multiple sources.
Identifying patterns in large amounts of data.
Preparing an initial analysis for human review.
Creating a first draft of a report or standard operating procedure.
Process improvements may be more useful for:
Clarifying who owns a task.
Eliminating duplicate approvals.
Establishing consistent intake requirements.
Setting deadlines and escalation procedures.
Reducing unnecessary handoffs.
Creating a single source of truth for important information.
Human judgment remains essential for:
Decisions involving significant financial, legal, or operational risk.
Sensitive employee or customer situations.
Exceptions that require context and experience.
Decisions that require accountability, empathy, or relationship management.
Of course, these categories overlap. AI can assist with complex work, and humans can make mistakes on routine work. The important thing is to define what the technology is allowed to do, when a person must review the result, and who remains accountable.
A useful rule of thumb: Do not automate a step until you understand why it exists and what would happen if it went wrong.
You might find that a task should be eliminated entirely, simplified, standardized, or reassigned before you introduce AI.
Step 4: Measure Results, Not Just Activity
One of the easiest mistakes in an AI initiative is measuring how much technology people use instead of whether the organization performs better.
Depending on the process, useful metrics might include:
Business outcomes: Are you increasing revenue, improving retention, reducing risk, or advancing your mission?
Customer experience: Are customers, clients, students, or tenants receiving better service?
Employee capacity: Are employees spending more time on higher-value work?
Cycle time: How long does it take to complete the work?
Cost per transaction: How much time and money does each completed task require?
Error rate: How often does the work need to be corrected?
Rework: How much effort is spent fixing or repeating tasks?
Let's say AI reduces the time required to prepare a weekly report from four hours to one. That is a meaningful improvement, but only if the report is useful and the saved time is used productively.
If the employee simply spends those three hours producing more reports nobody reads, you have increased report production but haven't improved organizational performance.
Action item: Choose one or two measures for your pilot project. Record current performance, test the new process, and compare the results. Ask whether the change improved the outcome—not just whether it made one step faster.
Step 5: Expect the Next Bottleneck
Another reason to look at the entire workflow is because fixing one constraint often reveals or produces another.
Suppose your team takes two weeks to process new customer inquiries. Then, you introduce AI to categorize and route incoming requests, cutting the response time to one day.
Fantastic!
However, now the sales team has more qualified inquiries than it can handle—the bottleneck just moved from intake to sales capacity.
Should you abandon the improvement? Absolutely not. However, you need to keep managing the entire system rather than focusing on making one step faster in a vacuum.
Organizations are interconnected—a change in one department affects the work of another. A new technology can eliminate one problem while exposing a different constraint that was previously hidden.
Workflow improvement should not be a one-time project—it should be part of how you manage operations for a growing business.
As your organization grows, revisit the processes that matter most. Look at where work is accumulating, whether responsibilities still make sense, and whether the systems supporting your team are keeping up.
Action item: After implementing a change, check the downstream effects. Did the improvement reduce total cycle time, or did it simply move the waiting period somewhere else?
Step 6: Give Employees Room to Experiment—and Share What Works
Leaders don't have to identify every AI opportunity themselves.
The employees closest to the work often know exactly where the inefficiencies are. They may already be experimenting with AI to simplify tasks, improve communication, or organize information.
To turn isolated experiments into repeatable improvements, create a simple way for employees to suggest and test ideas. Encourage them to identify a problem, propose a solution, and measure the result.
For example:
An employee identifies a recurring task that consumes several hours each week.
The employee proposes using AI to reduce the manual work.
The team checks privacy, security, accuracy, and any applicable legal or regulatory requirements.
The employee tests the approach on a small scale.
The team compares results against the existing process.
If the test succeeds, the organization documents the new process and shares it with others.
Ensure your employees know that you do not expect every experiment to work. The point of experimentation is to learn quickly without introducing unnecessary risk.
In addition to turning to ground-level employees, leaders should look beyond their own organizations for ideas. Another business, nonprofit, or school may have solved a similar problem in a way your team hasn't considered. You don't need to invent every solution from scratch.
By encouraging and rewarding experimentation and collaboration, you can increase the chance that successful experiments become part of the organization's operating practices rather than remaining dependent on one employee who knows a clever trick.
The Bigger Lesson: AI Is an Operations Challenge
It's tempting to treat AI adoption as an IT project. Choose a tool, train the employees, establish a few rules, and move on to the next priority.
In reality, meaningful AI-driven productivity gains require more than technology—they require leaders to understand how work gets done, where resources are being wasted, how decisions are made, and what the organization needs to accomplish.
That is an operations challenge.
Solving this challenge is particularly important for growing organizations, where informal processes and founder-dependent decisions can become increasingly expensive.
A small business might get away with having the owner approve everything. A school network might rely on one person who knows how every system works. A nonprofit might have employees maintaining five spreadsheets because nobody has time to build a better process.
AI can help, but it won't automatically clarify decision rights, create accountability, improve communication, or build scalable systems.
Those things require operational leadership.
The good news is that you don't need a massive transformation initiative to get started—pick one workflow. Understand it. Fix the biggest constraint. Test a practical improvement. Measure the results. Then repeat.
You might discover that AI is the right answer. You might discover that a simpler process would solve the problem. Or, you might discover that you need both.
Make sure you are solving the right problem before investing in the solution.
Ready to Make Your Operations Work Better?
If your team is working harder but not getting better results, adding another tool may not be the answer. You may need a clearer process, better-defined roles, more effective decision-making, or a practical plan for using AI where it will make a difference.
A fractional COO can help you identify operational bottlenecks, redesign workflows, improve accountability, and build systems that support sustainable growth—without adding a full-time executive to your payroll.
At Chou Consulting, we help growing organizations turn operational chaos into scalable systems.
Want to figure out where your organization is losing time, money, or capacity? Set up a free consultation, and let's identify the changes that will make the biggest difference for your company.
