Advice on a specific task
We review the current workflow, compare suitable options and choose a sensible first step. You receive a clear recommendation on what to test, what should remain with people and how to judge whether the approach works.
Practical AI training · consulting · prototypes
Make AI work for you.
Built around your documents, tasks and decisions.
You practise how to brief AI, check the result and recognise when a person needs to decide. You leave with a method you can use straight away.
Track record
- +117 %monthly sales at firms in the programme I designed
- +66 %monthly profit at the same firms
- €530cost per participant, recovered within one month
- 240+ hoursof training delivered to firms and institutions in 2025
The impact figures come from the randomised World Bank trial published in June 2025. Training hours are evidenced by invoices.
Services
Ways we can start
Depending on the situation, I can provide focused advice, practical team training or a small prototype. The scope reflects how big the problem is, how many people the change touches and what source material exists.
Practical training for your team
The training is built around documents and tasks the participants already handle. They practise briefing AI, checking its output and recognising when it should not be used, then leave with methods they can apply in their work.
A working prototype
Using a limited sample, I build a working demonstration of the new approach. You can see the quality of the output, common failure points, the checks required and the evidence needed to decide whether to continue.
Projects
Selected projects
World Bank: practical AI for small and medium-sized businesses
An enterprise support programme across three Czech regions needed generative AI turned from slideware into something a small business owner uses on Monday morning.
- What I did
- I designed the programme structure and the entire educational content, selected the methods and tools, and wrote exercises built on documents the participants actually work with. I delivered the introductory courses and the AI-in-marketing sessions myself.
- Scale
- Cohorts of entrepreneurs in the Ústí and Karlovy Vary regions, five connected workshops, train-the-trainer sessions and follow-up materials for independent practice.
- Measured result
- Monthly sales +117 %, monthly profit +66 %, intensity of AI use +26 % against the control group. ROI: the €530 cost per participant was recovered within one month. Measured across 173 firms in the randomised trial the World Bank published in June 2025.
Inovační centrum Ústeckého kraje: from idea to decision
Businesses and public bodies in the region knew they wanted to do something with AI. What they lacked was someone to tell them what their own people could actually run, and what would survive internal approval.
- What I did
- I ran consultations on concrete briefs: proposals and marketing copy, sorting and searching documents, preparing material for decisions. For each one we walked through the current process, tested an AI version on their own data and settled where the decision has to stay with a person.
- Scale
- Individual consultations and group sessions for companies and public-sector representatives in the Ústí region.
- Measured result
- Each consultation ended with a specific method and a line of accountability, not a list of tools. Several firms left with a decision not to invest in AI yet, because the numbers did not work for them.
Libraries in Czechia and Slovakia: AI for information work
Library professionals write literature searches, annual reports and grant applications. A generic AI course does not help them. They need a method for their own documents and confidence that nothing invented survives into the text.
- What I did
- I built and delivered multi-day courses on their real work: searches across their own collections, drafts of annual reports, grant application copy. Every task came with a checking routine: how to spot a fabricated source and what must never be pasted into a tool.
- Scale
- Vědecká knihovna v Olomouci, Knihovna Ústeckého kraje and Slovenská národná knižnica. Multi-day courses, not one-off talks.
- Measured result
- Participants left with working templates for their own workload and a checklist for verifying output. Average instructor rating 4.64 out of 5.
Selected collaborations
Participant testimonials
Testimonials
Average course rating 4.64 out of 5.
“What we learned in the course will help not only us, but many of our clients as well. We can reduce the administrative burden and devote much more time to direct work with clients.”
Jaroslava M.Fundraiser
“I am no newcomer to AI, yet I still learned a great deal. The course was well structured and professionally delivered. I appreciated the strong examples drawn from real life and practice.”
Jan R.Integrations & AI Automations Solutions Engineer
“He tailored the course to each participant’s professional focus, answered every question and demonstrated different AI models. He came across as a true professional and a leading expert in his field.”
Lenka Ch.Librarian
Accountability
What a person answers for
Every engagement settles who answers for the output, what gets checked, and how anyone will know it worked. Here is what that looks like in practice.
Review points
In the library courses every task carried a checklist: what gets verified every time, what gets spot-checked, and what must never be pasted into a tool. For literature searches that meant tracing every cited source, because a fabricated citation in an annual report is the library's problem, not the model's.
Regulation and internal rules
A public institution needs a different level of documentation than a five-person print shop. With institutions I therefore start from what the AI Act and their own data rules require, and what that means for the specific task. Settling it upfront is cheaper than settling it after an audit.
Evidence that people actually use it
The World Bank programme ran as a randomised trial with a control group, so the change could actually be measured. Smaller engagements get a simpler yardstick but the same principle: we agree upfront how we will know, three months later, whether it was worth it.
Checking the output
Quality is not judged by impression. I use scoring criteria, side-by-side comparison and error analysis, and participants run it on their own documents so they can see where the model is reliable and where it is not.
Process
How the engagement works
Every step has a concrete output and a point where you can say it is not worth continuing.
We mapBrief
We walk through the current workflow, the result you need from the change and the people it touches. Right at the start we agree how we will know it works.
We testApproach
On a limited sample I assess output quality, common failure points and the checks it will need. The findings show whether to proceed, revise the brief or stop.
We priceScope
Based on the findings you get a proposal for a consultation, training, prototype or a combination, including the fee and what will be finished at the end.
We deliverHandover
We test the solution with the people who will use it. They leave with the method, templates and checking rules they need to carry it on without me.
Lukáš Rejchrt
About me
I help businesses and institutions find where AI actually earns or saves money, and hand their people a method they can run and review without me watching.
- I designed the training programme for the World Bank that went through a randomised trial with a control group. Firms in it raised monthly sales by 117 % and the cost per participant was recovered within a month.
- I build courses on the documents people really work with: proposals, literature searches, annual reports, grant applications. In 2025 that came to over 240 hours of teaching.
- With institutions I also handle what they can get approved internally: lines of accountability, AI Act documentation and output checks. When a solution does not pay off, I say so.
Lukáš Rejchrt · AI consulting, training and prototypes
Contact
Working on a specific project or planning a broader AI rollout?
On an initial call, we will discuss the current situation, the result you need and the people affected by the change. I will then recommend a sensible first step and explain what is worth preparing.
The email link and Book an intro call button open your email app. The website itself does not send or store any information. How I handle personal data.