About

Architects first, AI second.

LogicStream is a transformation-consulting practice built around one person's 25+ year record of turning executive intent into delivered architecture — now applying that exact method to AI adoption.

Pavel Kolář

Transformation consultant, dual MSc & Executive MBA

A 25+ year record of turning executive intent into delivered architecture — on programmes such as ČEZ's group-wide SAP S/4HANA swap and Erste Group's core investment-banking replacement. Dual MSc Computer Science and Executive MBA.

Now applying that method to AI adoption: leading discovery workshops with senior stakeholders, mapping and simplifying business processes, selecting technology, building proofs of concept, and defining the target agentic architecture for the organisation.

Domain expertise

  • Investment banking
  • Corporate & consumer lending
  • Leasing
  • Energy & utilities
  • Telecommunications
  • Manufacturing ERP
  • Public-sector legacy systems

Certifications

Credentials behind the practice

AWS AI Practitioner (eight-course path); AWS Cloud Practitioner Essentials

IT Business Analysis with GenAI and LLM

SAP Certified Associate — SAP Activate Project Manager; Prince2 Foundation & Practitioner

Sun Certified Enterprise Architect for Java EE

In progress: Anthropic AI Architect track; hands-on with Claude Code

Engagements

Organisations we've worked with

  • ČEZ a.s.
  • Erste Group / Česká spořitelna a.s.
  • Vodafone Czech Republic a.s.
  • Aurora Cannabis Inc.
  • NESS Czech s.r.o.
  • ČSOB Leasing a.s.

Our story

We built AI-driven products before "AI" was the pitch

In 2012, LogicStream co-founded the fintech startup CloudTeq.eu, building a cloud e-invoicing (EBPP) consolidator and an OCR-based automatic invoice extractor — using machine learning for proprietary automated invoice processing, years before these techniques became mainstream.

That generation of machine learning was genuinely ahead of its time — and also genuinely limited: narrow models, hand-tuned pipelines, and accuracy that needed constant human correction. We say this plainly because it's exactly the contrast that makes today's agentic AI practice credible rather than theoretical: the same invoice-extraction capability could be rebuilt today, on LLM-based, agentic foundations, faster to build, cheaper to run, and at near-100% accuracy. We didn't read about that shift — we lived through the "before."

Work with us

Tell us where you are today

Contact →