Kratikaitse
EN

Practical AI · Agent Watch

Practical AI. Clear responsibility.

I build working AI with your team, with clear rules for data and responsibility. And I test whether the chatbot that talks to your customers can be tricked.

Get in touchSee my work
From data to decision

Approved data

Agreed rules for use

AI processing

Human review

Approval where needed

What I do

Practical AI

Working AI in your team's everyday work.

  • We pick one or two tasks where AI saves real time.
  • I build the solution with you and put it into use.
  • We agree which data it may use, who owns it and where a person approves.

I follow ISO/IEC 42001 principles for managing AI and the OWASP Top 10 for LLM applications for security. This is not a certification.

The AI register and staff rules can also live in my tool, AiAiAI. ai3.ee

Get in touch

Agent Watch

Security testing for AI chatbots. Can yours be tricked?

  • Pre-check: ten harmless questions to your public chatbot.
  • Check: 47 attacks based on the OWASP Top 10 for LLM applications, with reproducible examples and fixes.
  • Watch: the same test every month. I tell you when something gets worse.

For companies with an AI chatbot answering customers.

How I test

Experience and selected work

Mart Liivand

Mart Liivand

AI implementation and information security

Mart on LinkedIn

Education and training

  • MSc in IT Management (2021)
  • TalTech, Information Security Manager programme (2024)
  • TalTech, Cyberops Associate

I have worked in IT since 2009, including 12 years in support and programme management at Skype and Microsoft, followed by roles in regulatory technology and manufacturing analytics. In recent years, I have combined information security work with building AI systems. I work as part of a team that holds ISO 27001 certification. I am also working towards ISO/IEC 42001 Lead Implementer certification.

AI for manufacturing analytics

The language-model solutions I built run daily at a manufacturing analytics company: customer support quality review, knowledge capture from support tickets, natural-language queries about a device fleet, and a website assistant that creates CRM contacts with conversation transcripts.

Laboratory monitoring, from sensor to application

I built a monitoring platform for an Estonian battery research company, replacing manual measurement records. It covers automatic data collection, alerts, controlled access and tested recovery from backup. The platform is still in use and under my care. The client remains unnamed by agreement.

Governance in everyday use

For each AI solution, I document the model choice, where data is processed, what is logged and what each run costs. I define where human approval is needed and how output quality is assessed. I also apply ISO/IEC 42001 principles to my own systems.

I see AI risk as a developer, a daily user and the person responsible for the rules. That experience helps me turn requirements into working solutions.

Contact

What could work better?

Tell me a little about your company and what you would like to improve. We will find a time to talk.

mart@kratikaitse.ee