How a mini app built with AI improved call center metrics at a multi-billion dollar company

← AI Profit Garden

Companies large and small struggle with the same quiet problem. Internal systems that do not keep pace with the work, and staff who are required to hit ambitious goals without adequate tools to get there.

A large part of my work as a fractional CFO means stepping inside companies, large and small, to evaluate their internal infrastructure, determine what is truly effective, and strengthen the areas that are lagging behind to help improve the company's bottom line. That kind of evaluation is rarely accurate from the outside, and in this case, it took putting myself into the role, hands on, to see it clearly.

The company

I had the opportunity to evaluate the effectiveness of the customer service channels of a Fortune 500 financial technology company, a multi-billion dollar business, and the metrics they used to track key performance indicators on every call, including how well their customers were being served.

I sat on the team that fielded incoming calls from some of their most valued customers.

The challenge

Leadership tracked multiple metrics on every customer service agent, and used them to determine how well customers were being served. Some of the metrics scrutinized most closely by the management team were average call handling time, the productivity of each call, transfer rates, and how quickly issues got resolved.

In working this role, I quickly realized there were ambitious goals that needed to be met, but no effective way of accomplishing them. At the same time, meeting those goals was simply part of the job for every agent.

I found that the systems the company had in place lacked the key features that would have made the work easier, and that were driving up after-call handling time. Proper case documentation, along with the other documentation different departments required, was taking longer than it should have.

Even with a lifetime spent serving customers, I still needed extra time after every call to document it properly, so the company had the historical record it needed.

After-call handling time was one of the metrics the company tracked closely, and they wanted it as low as possible. I quickly realized there was a real gap between the goals the company had set and the means available to its customer service agents to achieve them, so I set out to find the best solution.

What I built

I designed and built a lightweight tool, a small application made for one purpose. It gave every field a home. Dictation instead of typing, wherever dictation was faster. One click to copy a section instead of retyping it into the various fields it needed to live. Multiple fields that were pre-filled from what had already been entered once. Formatting handled automatically, every time, the same way, as well as customized notes to guide the agent on what data to collect.

It was not complicated. It did not need to be. It needed to exist, and it did not, so I built it.

I am not a software developer or a coder. What I created was a software tool built with the help of AI, using an approved large language model to turn what I needed into something real.

The results

I used it on my own calls first. Before the tool, after-call documentation was taking me roughly 35 minutes across an eight hour day. With the tool, it dropped to 5. That is thirty minutes returned every day, a reduction of 86 percent, and it happened inside the first week. Not eventually. Immediately.

Across a working year that is more than one hundred hours handed back to a single agent, and the division has many.

Company wide today

The tool was adopted across the rest of the customer support division and is still in use today, company wide, continuing to give customer service agents improved metrics, as it did for me.

I own the application. The company uses it with my permission.

The work continues

Through my experience helping this multi-billion dollar company solve one of its greatest challenges, I embarked on an endeavor to use AI solutions as an effective way to solve company problems, for businesses of every size.

AI does not replace the judgment a business still needs. What it replaces is the manual work standing between a good idea and a working solution. A tool like the one I built once required a developer, a project timeline, and a budget most companies could not justify for an internal fix. Today, the same result can be built in days, not months, at a fraction of the cost.

That shift changes what is possible inside a company. Work that once took a person hours, retyping the same information into three different systems, can be handled in seconds. Time that was lost to repetitive documentation gets returned to the people doing the actual work of serving customers, closing deals, or solving problems that require a human mind. And because the tool itself costs so little to build, the return on that investment shows up almost immediately, rather than something a company has to wait years to see.

This is the endeavor I have built my practice around. Not AI for its own sake, but AI aimed squarely at the internal friction that quietly drains a company's time, labor, and money, replaced with something built specifically for the problem at hand.

What it proves

One small application, built without a development team and without a long timeline, cut documentation time dramatically and gave an entire department a faster, more accurate way to do their jobs. That is the same opportunity sitting inside most companies still relying on manual processes to do what a well built tool could do in a fraction of the time. Fewer hours lost to repetitive data entry. Fewer errors from retyping the same information more than once. More time for the work that actually requires a person.

AI is not here to replace people. It clears the ground so people can do what they were gifted to do.

Questions people ask about this

Do you have to be a developer to build an internal tool like this?

No. I am not a software developer or a coder. I built this tool with the help of an approved large language model, which turned what I needed into working software. The skill that mattered was knowing the work well enough to specify exactly what the tool had to do.

How long does a tool like this take to build?

Days rather than months. A tool like this one used to require a developer, a project timeline, and a budget most companies could not justify for an internal fix. That is no longer the constraint.

What kind of problem is this approach best suited to?

Repetitive documentation and retyping. Any place where a person is entering the same information into more than one system, formatting the same thing by hand every time, or spending time after the real work is finished just to record that it happened.

Does this only work at large companies?

No. The example here happens to be a Fortune 500 company, but the problem is not a big company problem. Small businesses run on the same manual processes, usually with fewer people to absorb them, which makes the hours returned worth even more.

What was actually measured, and by how much did it change?

After-call handling time, which is the documentation work a customer service agent completes once a call has ended. It went from roughly 35 minutes across an eight hour day to 5 minutes, a reduction of 86 percent. It was one of several metrics leadership tracked on every agent, alongside average call handling time, transfer rates, and how quickly issues were resolved.

Who owns the tool?

I do. I designed and built it, and I own the application outright. The company uses it with my permission.