How to Introduce AI in Your Company in 30 Days

I know two types of companies. The first ones ignore AI completely. Management says "it's just hype, we'll wait," while their competition automates reports, speeds up sales, and saves dozens of hours each week. The second type gets overly excited, buys Enterprise licenses for everything at once, and three months later discovers that nobody actually uses the tools. Both approaches are wrong.

This is a proposed four-week pilot, not a promise of an outcome or an account of a specific client. For each process, define an owner, permitted data, baseline time, and output review before starting. Only a before-and-after comparison can show whether it is worth continuing.

Week 1: Audit and pilot selection

The first week is not about technology. It is about understanding where the pain is.

Mapping the processes that eat time

Sit down with team leads and ask one question: "Where does your team repeatedly do something that is boring, predictable, and takes more than an hour per week?" Write it all down. You are looking for things like:

• Writing sales proposals from templates
• Summarizing meeting notes
• Answering repetitive questions from clients or internal colleagues
• Creating recurring reports from data
• Transcribing and categorizing documents
• Initial screening of resumes or applications

The goal is not to find the company's biggest problem. The goal is to find a specific, repeatable task where AI can help quickly and measurably.

Picking 3 candidates for the pilot

From the full list, pick a maximum of 3 processes. No more. Three things decide it.

Frequency comes first. A task done several times a week is worth the effort. A task done once a quarter costs you a month of setup to save an hour a year. Then structure. The task needs a clear input and a clear output, because that is what AI can actually work with. When a team lead tells you "sometimes this way, sometimes that way, and it depends on gut feeling," park it. Last is measurability. If you cannot say today how many hours the work takes, you will not be able to say in a month whether AI helped. Measure it before the pilot starts.

Week 2: Tool setup and pilot launch

Now you know what you want to solve. Time to pick the tools.

Choosing the right tool

There is no point buying a tool and then looking for a use case. Start from the process, not the technology. Which mostly means starting from wherever your people already work.

If the company lives in Outlook, Teams, and SharePoint, Microsoft Copilot for M365 is the fastest path. It summarizes emails, generates presentations, drafts in Word. Because it sits inside the apps people have open all day anyway, adoption costs you almost nothing. The license runs about 30 EUR per user per month.

ChatGPT Enterprise or Business is the more universal option, and it suits teams generating text, analyzing documents, and working with data outside the Microsoft world. The Business tier starts at 25 EUR per user per month.

Notion AI is most useful when Notion is already your knowledge base or project tool. Before enabling it, verify current settings, permissions, and data processing. Notion says its AI uses subprocessors and that Enterprise workspaces use zero retention with LLM providers by default. Notion AI security.

Teams on Jira and Confluence get Atlassian Intelligence and Rovo, which search across the entire Atlassian ecosystem. At hundreds of Confluence pages, that is the difference between "we wrote that down somewhere" and "here it is."

And Slack AI will summarize the channel you have not had time to read for two days, or dig out the answer buried in a thread from six months ago.

Setting up access and rules

Do not buy licenses for the entire company. Start with 5-15 people who will run the pilot. Agree on who has access to which tool, where outputs get saved, and who reviews them.

Be strict about one thing in particular. Personal data, sensitive financial data, and trade secrets do not go into AI, and that rule has to exist before anyone gets their first license. You do not need a 20-page policy. One page does it.

Week 3: Training and adoption

You have the tool, you have access set up. Now comes the hardest part. Convincing people to actually use it.

AI champions in teams

In every team that is part of the pilot, identify one person who is enthusiastic about technology. It does not have to be someone from IT. It can be a marketing manager, a sales rep, or an account manager. What matters is that this person wants to try new things and is not afraid to ask questions.

These people are your AI champions. Give them an extra day or two of training, and they then help their colleagues in their own team. This works 10x better than centralized training for 50 people in a conference room, because the colleague sits two desks away and asking them costs nothing.

Framework: Task, Context, Output, Check

Teach people four steps that hold for every prompt.

It starts with the task, and the task has to be specific. Not "write something about the product," but "write 3 versions of an introductory email for a potential logistics client." Then context. Attach the data, the samples, the previous outputs. A model knows nothing you do not tell it, and most bad outputs are really just missing context. Then the output itself. Say upfront whether you want bullets, a table, or prose, and how long, or you will be doing a second round.

And finally the check. Nothing an AI produces goes any further until a person has read it. That rule has no exceptions.

What to measure

Watch three numbers. How many people in the pilot group actually open the tool at least 3 times a week. How many minutes or hours a week a given process saves, which you find out by asking people directly and having them log it. And whether the outputs are comparable to or better than before, judged by the people who work with them downstream.

What is NOT a metric

Prompts per day tells you nothing about value. The "wow" at the first demo fades within a week. And a count of installed tools measures purchasing, not usefulness.

Week 4: Evaluation and scaling

The fourth week is about hard data and decisions.

Measuring pilot results

Collect data from the pilot group. For each of the 3 pilot processes, answer:

• How much time was actually saved?
• What was the output quality compared to manual work?
• What resistance was there, and what caused it?
• Would people keep using it even without someone checking on them?

That last question is the one that matters. If the answer is no, you finished a pilot but you did not introduce anything.

Scale, adjust, or stop

For each pilot, make one of three decisions.

Scale it if it works, people use it, and it measurably helps. Roll it out to more teams.

Adjust it if you can feel the potential but something is grinding. Change the tool, change the process, or add training, and give it another 2 weeks.

Stop it if it does not work. AI does not make sense for this particular process. That is not a failure. You just saved yourself a year of pushing something that was never going to stick.

Internal AI rules

Based on the pilot, write a simple internal policy:

• What we use AI for and what we do not
• What data must not be entered into external AI tools
• Who is responsible for reviewing AI outputs
• How to handle situations where an AI output contains errors

Budget and costs

Calculate the real costs. Licenses are the smaller line item. The bigger one is the time people spend on training and administration, and that is the one budgets routinely forget. Compare it against the time saved. The ROI is usually obvious. If it is not, you picked the wrong process to automate.

Checklist: 30-day AI plan

• Map time-consuming processes in every team
• Select 3 pilot processes (frequency, structure, measurability)
• Choose a tool based on existing infrastructure
• Purchase licenses for the pilot group (5-15 people)
• Write a one-page document with AI usage rules
• Identify an AI champion in each pilot team
• Train AI champions (framework: Task, Context, Output, Check)
• Launch the pilot and measure active users, time saved, and output quality
• After 2 weeks, collect feedback from the pilot group
• Evaluate results for each pilot process
• Decide whether to scale, adjust, or stop
• Write an internal AI usage policy
• Calculate ROI and prepare budget for scaling

Rise.sk case study

We ran an AI Skills Sprint for a mid-size company in Slovakia. When we started, their relationship with AI was "a few people tried ChatGPT for writing emails." No structure, no measurement, no rules.

An illustrative pilot can start with proposal preparation, call summaries, and first drafts of technical documentation. Compare time and quality with a recorded baseline before deciding whether to expand.

The key was not the tools. It was the structured approach to identifying where AI actually makes sense and where it does not.

What is next?

List three repeated processes, their owners, and the time they take each month. On a first call, we can select one pilot by value, risk, and data readiness.