Inovai / Workshops and analysis
Do you know where AI will really save you money?
We go through processes and data and rank candidates by benefit. The output is a plan.
AI candidates by benefit and effort
Recommendation: start with the first two, the third after a data inventory.
Problem
AI gets bought on demos, not on your processes.
What we’ll do
Analysis of processes and data, candidates, architecture and priorities.
Result
A plan with a benefit estimate, and workshops for management, IT and teams.
What exactly we do
We go through your processes, find where AI saves time and propose how to do it.
How work really gets doneWhere data gets retyped, where people wait and where checks are done by eye. With the people who do the work, not just management.
Numbers behind every candidateVolume, staff time, error rate and the impact of an error. And where AI won’t help and fixing the process is enough.
A plan with a first stepSolution design, priorities and benefit estimate in one document. Usable without us, too.
A half-day workshop for directors and owners.
Where AI makes senseUsing examples from your industry: extraction, workflows, assistants, BI. What works in production and what only works in presentations.
Where it doesn’tTasks where AI costs more than a person, data is missing or an error costs more than the saving. We’ll tell you straight.
Risks and obligationsWhat the Czech Cybersecurity Act (Act No. 264/2025 Coll.) and the AI Act mean for your deployment, without scaremongering.
Attacks, defence and evidence in practice, with participants at the keyboard.
Attacks first-handPrompt leak, jailbreak in Czech, injection via a document, agent manipulation. Participants attack a practice chatbot themselves.
Layered defenceSystem prompt, filters, permissions, tool sandbox, judge model. We show how an attack bypasses each layer.
Evidence for the auditorHow to record a finding so a developer can reproduce it and an auditor accepts it. Led by Radek Palla, ISO/IEC 27001 Lead Implementer.
Safe use of AI and Copilot under Article 4 of the AI Act.
What AI can do and where it gets things wrongHow answers are generated, why the model makes things up and how to verify an answer. Using examples from your work.
What doesn’t belong in AIPersonal data, contracts, passwords, customer data. Where data goes in Copilot and where it goes in a public chatbot.
Copilot in practice, with proof of trainingShort exercises in Teams, Word and Outlook. Attendance sheet, content and a test as evidence for Article 4.
How the analysis works
Four steps from processes to a prioritised plan.
People
Step 1Processes and peopleWhere data gets retyped, where people wait and where checks are done by eye. With the people who do the work.ERPquality okCRMduplicatesWarehouseno owner
Step 2Data and systemsWhat is where, in what quality and who owns it.Step 3Where AI helpsVolume, staff time, error rate, impact of an error. And where fixing the process is enough.
deviation
We label benefit estimates as estimates. The pilot brings the real numbers.
Where do we start if we don’t know what we want?
With the analysis - it reveals where time is really lost. A management workshop makes sense if you first want a shared language.
Is Article 4 training mandatory?
Article 4 of the AI Act addresses the AI literacy of people who work with AI. The scope of the obligation is being adjusted under the Digital Omnibus, but the form is always up to you. Our course provides the content and proof that it took place.
Do you run workshops online too?
For management and teams, yes. We prefer to run the IT workshop in person, because participants attack a practice environment and work out together why the attack got through.