AI in the Workplace: What Actually Changes in the First Year

Implementing AI in a company with 10–200 employees isn’t a quick fix. It’s more like a year where the first quarter is a cost, the second is a break-even point, and profits don’t appear until the third. Guides show the end goal and skip over the journey, so here’s the journey itself.

Broken down quarter by quarter, including what usually goes wrong.

First quarter: You pay and see no results

The first month is spent on things that don’t look like AI. Someone documents the process, someone organizes the documents, someone figures out where the data is even located. This is often the most time-consuming part of the entire project and takes everyone by surprise.

The second month is for development and the first limited rollout—usually on a single channel and a single case type—so it can be rolled back without consequences. The system works. It gets it right 70 percent of the time.

The third month is for calibration. This is where most projects fail, because someone has to review the system’s output every day, flag nonsensical responses, and refine the rules—and this work wasn’t planned because it wasn’t included in the proposal.

If there’s one thing to take away from this quarter: schedule someone’s time to review the results in the third month. You can’t skip this step.

Second quarter: You break even, and resistance begins

Around the fourth month, the system gets it right often enough that it stops being a nuisance. This isn’t a profit yet. It’s a return to square one—except after paying for implementation and after three months of someone else’s time spent correcting responses.

That’s when resistance usually sets in—and not from the people who’ve lost their tasks, but from the people who’ve lost control over how the task is being performed.

The symptom is always the same. Someone keeps their own parallel spreadsheet—not out of spite, but simply because they don’t trust the system and prefer to have their own version of the truth at hand. Don’t forbid it. Ask what’s missing.

Third Quarter: First Real Profit

It’s only at this point that a figure emerges that can be presented.

Rarely is it a reduction in headcount. More often, it’s response time, timeliness, or the fact that things have stopped falling through the cracks between people. These are soft outcomes, but they’re fully measurable—provided someone measured the baseline before implementation.

If you didn’t measure it, you won’t be able to calculate the benefit, and you’ll have trouble justifying the next step. This is the most common mistake we see, and it costs more than the implementation itself.

Fourth Quarter: The Scaling Decision

At the end of the year, you’re faced with one question: expand or stop.

The reason to expand isn’t simply that the first process is working. It’s that people have stopped talking about it. A system that no one discusses is a system that has been successfully implemented.

The reason to stop is a situation where the process works solely because one specific person monitors it daily and quietly fixes it. In that case, scaling up will multiply the dependency, not the effect.

What’s Different in the Second Year

Almost no one writes about the second year, yet it’s what determines the bill.

The development cost disappears, leaving only maintenance, so the monthly bill goes down. At the same time, language model usage increases as the company grows, because this cost is tied to the number of cases, not the number of employees.

A cost also emerges that wasn’t there before. We call it drift. The offering changes, procedures change, and the system responds based on outdated knowledge. Someone has to update it, and this is a recurring task, not a one-time effort.

A company that planned for this reaps the benefits in the second year. A company that didn’t plan for it discovers in the second year that the system has quietly become obsolete.

What happens when you reach out to us

We start with what can be measured today so that we’ll have something to compare it to a year from now. That’s the first step—not the last.

The owner reaches out, the conversation lasts 45 minutes, and ends with a realistic timeline for your specific case, along with a price range. We also tell you which month will be tough—because it will be.

The first scope covers a single process. The team subscription starts at 3,875 PLN per month, and you can calculate the cost using our calculator before the call. We stick around after implementation because keeping things on track in the second year is our job, not yours. The guarantee covers fixes at no extra charge.