Imagine starting your day without scrolling through dozens of emails. Instead, you get a short summary of everything important right on your phone. Or instead of spending hours building reports, you just ask a simple question in plain language—and get the result in minutes.
For some forward-thinking companies, this is already reality—thanks to AI agents who are becoming indispensable digital colleagues.
A Real-World Story: When an AI Agent Runs Data Processes
“I don’t read emails anymore. Every morning I get a digest of everything that landed in my inbox delivered to Telegram. And if something important comes in, the agent flags it and alerts me directly,” says Jiří Maňas, CTO at Keboola. His vision is clear: “AI agents will take over all the unproductive tasks in a company so people can focus on what actually creates value.”
This shift isn’t some distant future. Keboola, the Czech data platform that raised over 700 million CZK in 2023 and is trusted by companies such as Rohlík, Česká spořitelna, and Publicis Groupe, is already showing where the future of data work is heading.
What Exactly Are AI Agents—and Why Is Everyone Talking About Them?
AI agents aren’t just another tech buzzword. They’re autonomous assistants powered by artificial intelligence that can take care of tasks on their own—from sorting emails and analyzing data to building complex data pipelines.
Globally, the results speak for themselves:
- At JPMorgan, an AI agent helps bankers find information up to 95% faster.
- Startup Talent Inc. cut content creation time by 78% using AI agents.
- Research platform Causaly reduced literature review times by 90%.
The Czech Path to AI: MCP Server as a Breakthrough
In June, Keboola implemented the so-called MCP Server (Model Context Protocol), which acts like a USB-C port for AI. Developed by Anthropic (the company behind Claude AI) at the start of this year, it’s a new standard for connecting AI assistants with your tools and data.
What does that mean in practice?
A manager at an e-commerce company can ask their AI assistant: “What were the profit margins of our 10 best-selling products in Prague last month?” and get an answer right away—without waiting days for a manual analysis or knowing SQL or Python.
Similarly, you can ask it to set up a live connection to an external tool—for example, pulling weather forecasts regularly to enrich customer behavior predictions. Again, no programming languages or API documentation required.

MCP Server allows AI assistants like Claude, Cursor, Windsurf, or Make to build secure, production-ready data pipelines from natural language. In other words, even a non-technical user can build fairly advanced data infrastructure simply by talking to the system—something that used to take engineers weeks.
Practical Use Cases for Startups
- Automated Reporting
Instead of manually compiling weekly reports, you can simply ask: “Create me a weekly sales overview by product categories and compare it with last month.” The agent pulls the data, runs the analysis, and prepares a visualization. - Customer Analytics
“Find all customers who bought more than three times last month but didn’t place any order last week.” The agent scans your database and instantly identifies retention campaign opportunities. - Data Integration
Connecting data from your e-shop, accounting system, and marketing tools is no longer a nightmare. An AI agent can sync the data automatically and give you a unified business view.
How to Get Started with AI Agents in Your Startup
- Start with a Small Pilot
Pick one specific use case with fast value—like auto-generated recurring reports or sales data analysis. - Think Security First
Tools like Keboola come with strong security and clearly defined access rights so AI only sees verified data. Every action is fully auditable. - Involve the Team and Communicate
AI won’t replace people. The best results happen when humans and AI work together. Be transparent—employees should know the goal is to make their work easier, not replace them. - Leverage Czech Know-How
Local solutions understand the specifics of the Czech market and offer support in Czech. And today’s large language models (GPT-4, Claude, etc.) already handle Czech well, so language isn’t a barrier. - Take It Step by Step
Don’t dump the entire task on AI at once. Start by connecting a data source, check that the data loads correctly, then filter it, then aggregate it. AI makes mistakes, but checking each step makes errors much easier to spot.
Risks to Keep in Mind
It’s definitely a good idea to create and follow best practices, because while AI agents deliver significant advantages, they also come with certain limitations.
- AI Hallucinations: Even advanced models can occasionally produce wrong or made-up answers. Always verify critical outputs.
- Shadow AI: More than 80% of companies have experienced employees using some AI tool without IT knowing. It’s safer to provide an official, secure option.
- Governance Needed: Without clear rules and processes, AI can cause more harm than good.
The Future Is Already Here
“Everyone will have an AI agent, and everyone will have an MCP server. We’re about to see an explosion of these tools,” predicts Jiří Maňas from Keboola. “Companies that can use AI agents safely and smartly will gain a huge advantage in the coming years.”
For Czech startups, this means a unique opportunity. With AI agents, they can compete with much larger players—the advantage no longer comes from headcount, but from how smartly you use technology.
Concrete Next Steps for Your Startup
If you want to start implementing AI in your startup, you need a clear plan. Here’s a simple roadmap to help you integrate AI effectively and maximize the value:

As Keboola’s CEO Pavel Doležal says: “Working with your own data the right way is the best path to growth and innovation.” With AI agents, even the smallest startups can now take that path.
—
The authors of the article work at Keboola, a Czech data platform that helps companies automate data operations and deploy AI solutions. Keboola is a member of the Czech Association for Artificial Intelligence, and its technology is used by leading Czech and global companies.
References
1E. (n.d.). Embracing AI in IT: Augmentation, not replacement. https://www.1e.com/blogs/embracing-ai-in-it-augmentation-not-replacement/
GoFast.ai. (2025). Agent hallucinations in the real world: When AI tools go wrong. https://www.gofast.ai/blog/ai-bias-fairness-agent-hallucinations-validation-drift-2025
Holzmannová, A. (2025, September 29). Keboola šetří čas a dává lidem svobodu pracovat s daty. Jak s platformou pracovat, abyste se vyhnuli začátečnickým chybám? IT Trendy. https://ittrendy.cz/clanky/keboola-setri-cas-a-dava-lidem-svobodu-pracovat-s-daty-jak-s-platformou-pracovat-abyste-se-vyhnuli-zacatecnickym-chybam.
LangChain. (n.d.). State of AI agents report. https://www.langchain.com/stateofaiagents
McKinsey & Company. (2025). AI in the workplace: A report for 2025. https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/superagency-in-the-workplace-empowering-people-to-unlock-ais-full-potential-at-work
Multimodal. (n.d.). 17 useful AI agent case studies. https://www.multimodal.dev/post/useful-ai-agent-case-studies
Nielsen Norman Group. (n.d.). AI improves employee productivity by 66%. https://www.nngroup.com/articles/ai-tools-productivity-gains/
Reuters. (2025, May 5). JPMorgan says AI helped boost sales, add clients in market turmoil. https://www.reuters.com/business/finance/jpmorgan-says-ai-helped-boost-sales-add-clients-market-turmoil-2025-05-05/
Salesforce. (n.d.). 5 ways to ensure AI accountability in your AI agents. https://www.salesforce.com/blog/ai-accountability/
Slížek, D. (2025, July 21). Jiří Maňas (Keboola): AI agenti budou dělat všechny neužitečné činnosti ve firmě. Lupa.cz. https://www.lupa.cz/clanky/jiri-manas-keboola-ai-agenti-budou-delat-vsechny-neuzitecne-cinnosti-ve-firme/
Springs. (n.d.). Tackling regulatory compliance challenges with AI agents: Proven solutions to compliance problems. https://springsapps.com/knowledge/top-compliance-problems-solved-by-ai-agents-real-world-case-studies





