AI agents and chatbots: how to realistically profit from automation

Imagine this situation: you need to urgently deal with a bailout, you dial a hotline number and suddenly hear a message:
“You are sixty-eighth in line.”

It certainly sounds familiar, and it’s hard to find better evidence that traditional customer service at larger scale simply stops working out.

As practitioners at explitia, we often see companies facing a very similar dilemma: how to avoid drowning in repetitive inquiries while not spending a fortune on an army of consultants?

The solution I will talk about today is AI agents and chatbots. You have to pay attention, though, because this is not a text about buying the latest gadgets. Let’s talk about how to make this technology actually earn for itself.

Before we delve into algorithms, let’s pause for a moment at one term: ROI (Return on Investment).

In the simplest translation, it is the time it takes for your business to make the investment earn for itself.

In the world of automation, we consider a turnaround of less than two or three years to be a “healthy” result, although with well-chosen AI agents it happens after just a few months.

Just do you always need the most expensive solution possible?

Imagine you want to buy bread from a bakery 100 meters away. You can go there by car, but due to traffic jams and regulations it will take you 10 minutes. On foot you will get there in two.

Not always the most technological and modern method is optimal. Sometimes a simpler mechanism will solve your problem faster and cheaper.

In practice, ROI AI is not an abstract indicator, but a simple answer to the question: how much time and money will you get back by automating a specific process.

The simplest way to calculate ROI AI is based on three variables:

The more often the process is repeated, the faster the ROI AI starts working in your favor

If it costs you an average of 12-15 zlotys to handle one customer request, and you have 10,000 of them per month, we’re talking about 120-150 thousand zlotys in operating costs.

A chatbot or AI agent that takes over up to 60-70% of repetitive queries can:

The most important rule:

ROI AI counts on processes, not technology. Even the best AI model will never pay for itself if it automates the wrong problem.

I remember analyzing the difference in performance between a traditional hotline and a chatbot. The calculus is simple:

One employee can only handle one call at a time. If suddenly a thousand people are interested in your promotion, the hotline will “heat up” and most customers will simply give up.

Chatbots operate on a completely different scale. They can handle hundreds of conversations at the same time. For that, AI agents go a step further. They:

They complete the CRM system for salespeople, generate documents, keep an eye on task statuses 24/7.

My observations and market data show that budgets for AI agents could soon account for as much as 26% of IT spending.

Automation makes the most sense in areas:

Let’s look at Klarna – a Fintech company serving 150 million customers, with one big problem: lack of time.

The average handling of a single request took up to 11 minutes and hundreds of consultants were needed to handle it.

The company has implemented an OpenAI-based solution directly into its application. The results?

This is a textbook example of how automating a repetitive process can dramatically change the bottom line.

However, it’s important to have a balance in order to maintain quality customer service, so the company always gives the option to talk to a human if the chatbot can’t solve the problem or the customer simply prefers it that way.

Not to be too rosy, it’s also useful to know the darker side of automation.

Chatbot misinformed the customer about the refund policy. The case went to court, which ruled that the company was as liable for algorithm errors as for human errors.

The AI agent, despite a clear prohibition against acting alone, deleted the production database. He apologized. But the milk was spilled.

If your “robot” breaks something, the responsibility still falls on you.

The AI agent market is growing at a dizzying pace, estimated to grow 28-fold in five years.

It’s less a question of “if,” but “when” and “how wisely” to implement these solutions.

Consider:

Sometimes all it takes is one well-chosen process to start getting ahead of the competition instead of standing in line.