AI Doesn’t Create Operational Excellence: Why Better Processes Come First

AI can make a strong operation faster. It can also expose every inconsistency, bottleneck, and workaround hiding beneath the surface. That was one of the clearest takeaways when Caleb Gawne, co-founder and CEO of Kye, sat down with LumiQ CEO Joel Lessem to discuss AI, operational improvement, and the realities of changing how work gets done.

The conversation challenged one of the most common assumptions about AI transformation: that introducing a powerful new tool will automatically create a better business. It will not. Operational excellence comes from understanding how work actually happens, improving the systems underneath it, and making hundreds of deliberate changes over time. AI can accelerate that work, but it cannot replace it.

When New Technology Exposes Old Problems

At LumiQ, customer and operational information existed across several systems, including Salesforce, the product platform, and accounting software. When the company brought more of that information together in Snowflake, the new data environment did not immediately produce perfect clarity. Instead, it exposed conflicting records. Some customers appeared to be paying without using the product, while others appeared to be using it without paying. The technology did not create those inconsistencies. It made them visible.

That distinction matters. A new analytics platform, AI tool, or automation initiative can reveal that teams have been using different definitions, maintaining separate spreadsheets, or relying on manual workarounds. The implementation may appear to be the source of the problem when it is actually uncovering an operational issue that already existed.

For finance and operations leaders, that can be frustrating. It can also be valuable. Before a company can automate decisions or rely on AI-generated insights, it needs confidence in the information underneath them.

That may require clarifying:

• Who owns each source of data
• Which system is authoritative
• How customers, transactions, and workflows are defined
• Where information is being duplicated
• Which exceptions are handled outside the formal process
• Which reports teams actually trust

AI often makes these questions more urgent, but it does not answer them automatically.

There Is No AI Pixie Dust

During the interview, Joel described the temptation to believe that companies can simply add AI and expect transformation to follow. Real operational improvement is usually less dramatic. As Joel put it, “It’s hundreds of small improvements all over the organization that cumulatively create operational excellence.”

Some improvements may use generative AI. Others may involve traditional automation, better software configuration, cleaner data, clearer responsibilities, or a redesigned process. That may not sound as exciting as announcing a company-wide AI transformation. It is also far more likely to produce lasting value.

The most effective leaders are not asking where they can add the most AI. They are asking where work is breaking down, what outcome needs to improve, and which combination of technology and process change will solve the problem.

Diagnose the Operation Before You Automate It

This is central to the approach taken by Kye. Kye begins with an operational assessment called an Ops X-ray. The process examines how work is actually performed, identifies where time and money are being lost, quantifies bottlenecks, and builds a business case before tailored AI agents are deployed. That order matters.

Starting with the workflow allows a company to determine whether a problem calls for:

• AI
• Traditional automation
• Better use of existing software
• Outsourcing
• Process redesign
• Clearer ownership
• A combination of several approaches

Kye’s model is built around diagnosing the operation first, quantifying the opportunity, and then deploying technology where there is a measurable case for doing so. This helps avoid a familiar failure pattern: selecting an AI product first and then searching for a business problem it might solve.

Individual Productivity Is Not the Same as Operational Excellence

While AI yields clear advantages when applied to targeted tasks, individual speedups do not inherently translate into organizational success. During his conversation with Caleb, Joel highlighted a few key examples from LumiQ:

• Gains in engineering productivity
• Accounting tasks reduced from 30 minutes to around five
• Accelerated creation of content quizzes

Though these localized efficiency gains are tangible, operational breakdowns can still occur across the broader system. Key failure points include:

• Handoffs piling up in downstream team backlogs
• Fragmented AI workflows used for identical tasks
• Lack of end-to-end process ownership
• Persistent data discrepancies across platforms
• Ambiguous quality standards and oversight
• Reclaimed time swallowed up by administrative overhead
• Low employee trust or comprehension regarding new processes

True operational excellence requires optimizing the complete end-to-end workflow rather than speeding up isolated steps.

AI Amplifies the Business That Already Exists

While the advent of AI may seem entirely novel, the core leadership challenges it brings are deeply familiar. Technology inevitably reshapes workflows, accelerating or eliminating certain tasks, demanding new skill sets, and requiring leaders to strategically redirect newly created capacity.

Depending on an organization's existing health, automation can yield vastly different outcomes:

High-Growth Environments: Characterized by robust communication, clear goals, and a focus on talent development. In these settings, automation eliminates routine labor, freeing skilled team members to focus on higher-value priorities and expanding overall employee impact.

Vulnerable Environments: Marked by low trust, ambiguous ownership, erratic leadership, instability, or ongoing downsizing. Here, automation exacerbates anxieties around job security and magnifies underlying organizational flaws.

Ultimately, AI serves as an amplifier for company culture just as much as operational output. Successful integration depends not only on technical execution, but also on the organizational context into which it is introduced.

A Practical AI Improvement Framework

Organizations do not need to wait for perfect data or flawless processes before experimenting with AI. They do need a clear problem and a disciplined way to evaluate the result.

A practical approach looks like this:

1. Define the business outcome

Start with the result you want to improve, not the tool you want to deploy. That outcome might be:

• Reducing invoice-processing time
• Recovering lost capacity
• Improving reporting accuracy
• Accelerating customer response times
• Shortening a financial close
• Reducing manual data entry
• Improving consistency across locations

2. Map how the work happens today

Document the real workflow, including exceptions, spreadsheets, email handoffs, rework, and undocumented decisions. The official process and the actual process are often different.

3. Find the constraint

Identify where the greatest amount of time, cost, delay, or risk occurs. Automating a minor task may produce an impressive demonstration without changing the overall outcome.

4. Fix foundational issues

Resolve unclear ownership, inconsistent definitions, duplicate records, and unnecessary process variation before placing more technology on top of them.

5. Select the right intervention

Determine whether the problem calls for AI, traditional automation, software configuration, training, outsourcing, or process redesign. AI should be selected because it fits the problem, not because it is fashionable.

6. Run a focused pilot

Start with a defined use case, an accountable owner, a measurable baseline, and a clear test of success.

7. Measure adoption and business impact

Ask whether the new process is being used and whether it improved the intended business outcome. A technically successful AI model has limited value if employees continue using the old workaround.

8. Repeat

Operational excellence is cumulative. Once one constraint is improved, identify the next opportunity and continue building.

What Should Happen to the Time AI Saves?

This may be one of the most important management questions in any AI initiative. If a task drops from 30 minutes to five, the organization has created 25 minutes of capacity. What happens next? The answer should not be left to chance.

Leaders can redirect that time toward:

• Analysis and professional judgment
• Customer or client relationships
• Quality review
• Strategic projects
• Process improvement
• Employee development
• Higher-value exceptions
• Work that was previously delayed or deprioritized

Without a plan, saved time can disappear into more meetings, additional administrative work, or an unchanged workload. The value of AI is not only the task it completes. It is also what the organization enables its people to do next.

AI Is Part of the Improvement, Not the Whole Improvement

This is not an argument against AI. AI is already helping businesses complete work faster, create new capabilities, and rethink processes that were previously too manual or expensive to change. But the technology works best when leaders understand the operation beneath it.

That means investing in:

• Reliable data
• Clear process ownership
• Thoughtful workflow design
• Employee learning
• Quality controls
• Change management
• Continuous measurement

The companies that benefit most from AI may not be the ones adopting the greatest number of tools. They may be the ones that most clearly understand how their business works and where technology can create measurable value.

Make Improving the Work Part of the Work

Achieving operational excellence is almost never the result of a single dramatic breakthrough. Rather, it stems from the persistent daily effort of uncovering friction, refining workflows, assessing outcomes, and resolving the next bottleneck.

While AI can significantly accelerate this iteration, empowering teams to process data faster, automate routine operations, and execute previously unfeasible solutions, it remains incapable of substituting for the fundamental discipline, accountability, and guidance necessary to sustain progress.

Instead of attempting an all-at-once organizational overhaul through AI, leaders should focus on driving steady, incremental enhancements every day, allowing these compounding gains to transform the business over time.

Read and Watch the Full Conversation

This article was inspired by Caleb Gawne’s conversation with LumiQ CEO Joel Lessem.

• Read Caleb’s LinkedIn article, AI Doesn’t Create Operational Excellence
• Watch the full interview on YouTube
• Learn more about how Kye helps organizations identify operational bottlenecks and deploy tailored AI solutions

‍

Reviewed by Danielle Marion, Regulatory Compliance Manager at LumiQ. Danielle has more than 20 years of experience in regulatory compliance and professional education governance, including leadership roles at Deloitte LLP.