Jakub Linhart
Solution architect
Czechia

Solution Architecture, Project Management

Originally, I graduated from a military university as a commander of chemical units. However, my career path took a turn, and I transitioned to working in a company specializing in the development, production, and supply of extensive radiation control systems, particularly for nuclear power plants. To effectively manage a substantial portfolio of ongoing projects, I took it upon myself to program a multi-project management system. This system proved to be highly successful and remained in use for nearly a decade. Seeking a deeper understanding of business processes, I furthered my education by earning a degree from the Faculty of Economics. Subsequently, I ventured into the realm of ERP systems, working as a consultant with hands-on experience in production, development, and project management. Over the years, I've had the opportunity to lead numerous substantial implementation projects as a solution architect. In this role, I've professionally overseen entire teams and collaborated closely with project managers to ensure successful outcomes.

Managing AI Orchestration Systems – Where We Are Now and Where It Is Heading: How Long Will We Keep Hold of the Wheel?
communication
copilot
future-of-work
ai-agents
ai-orchestration
human-in-the-loop
continuous-improvement
society

AI has moved beyond chat. Orchestration systems and agents now work as a central brain: they draw on almost all of humanity's recorded knowledge, and they are connected to the systems around them – they read their data and they control them. As an ERP consultant and solution architect, I work with such a system every day. Today I still hold the wheel. But the machine gets faster and more capable every month. I see AI development in three orders – in the sense of first-, second- and third-order effects. First-order work built the models. Second-order work turned them into agents and orchestration systems. Third-order work is ours – the people who now teach these systems to do real work. We see both in our own company: our technical lead builds the agents and implements them technically – that is second-order work. We consultants teach them how to behave in real work – that is third-order work. And this third-order work, I believe, is where it is decided what AI will actually do and how well. This session follows that journey in three parts. First, discovering the central brain: what changes when AI stops answering questions and starts doing real work across e-mail, Teams, documents, DevOps, ERP, HelpDesk (and hundreds of other systems in the future). Second, third-order work – how to talk to it and teach it using instructions, skills and plugins: why context, reasons, doubts and even emotions are data, not politeness – and why people who talk to AI "like a machine" get worse results. Including a side-by-side experiment: the same task, written two ways. And how one person's corrections become shared rules for the whole team – from features through user stories to code review. Third, the question none of us can avoid: we will most likely let go of the wheel voluntarily, for efficiency and comfort. What does that mean for our skills, our work and our society? No vendor pitch and no prompt templates – principles and real experience. Key takeaways: – The three orders of AI development – and why third-order work, people teaching AI to work, is decisive right now – Why context and the "why" matter more than prompt tricks – The principle of continuous improvement: mistake → correction with a reason → rule → shared with the team – Where will AI that can run everything at lightning speed take us – and why are the worlds of Skynet and I, Robot closer today than we think? (A philosophical reflection)

Read more

The Consultant's Multiplier: Running ERP Projects with an AI Orchestration System
knowledge-management
business-central
consulting
copilot
ai-agents
ai-orchestration
context-engineering

AI orchestration systems now read our e-mails, meetings, documents, DevOps and ERP, and they act in them. The real question for consultants is how to make one reliable in day-to-day project work. Using it more often is not the answer. This session presents a working model I use daily as an ERP consultant and solution architect on Business Central projects. First, the anatomy: the orchestrator, specialised agents, tools connected to company systems, instructions and skills, and memory. The orchestrator plans the steps, then runs, checks and fixes each one, and puts the result together. It is driven by feedback, a principle from cybernetics rather than prompt engineering. Second, the formula: Multiplier = Tools × (Memory × Instructions)^k. The exponent k is the skill of communicating: stating the goal, supplying the inputs and specifying the procedure. When k is high, a well-built context multiplies the output. When k is low, the same context only wastes time. Every task for AI is a detective case: with complete inputs, a clear goal, a way to verify it, a precise procedure and the right tools, AI can hardly go wrong. When it does fail, one of these five is missing. Third, the results from real projects. A full review of a live project was built from 47 apps, 782 tickets and 826 DevOps items in 6 hours instead of an estimated 3 months. On an urgent ticket in a system nobody remembered, it took 8 minutes from connecting to finding the root cause. Producing a vision and a solution architecture diagram now takes hours or minutes, not days. The know-how behind this moves out of consultants' heads and into shared skills and memory. Fourth, the traps: a person stays accountable for every output, AI can be wrong very convincingly, models keep drifting back to their trained answer style, and people risk losing their skills when they only approve. Key takeaways – A clear model of what an AI orchestration system consists of and how it works step by step – The multiplier formula: why the skill of communicating decides whether AI multiplies your work or wastes your time – The five conditions of a flawless result, and how to find which one failed – How consultant know-how becomes shared instructions, skills and memory – Real numbers from Business Central projects, and the traps that come with them.

Read more

Jakub can deliver sessions in
English
Czech
Relevant industries
Business Services
Connect with Jakub

Report speaker profile

Reason for reporting this profile (multiple options possible)


Please select at least one option.

Please select at least one option.

Please select at least one option.

Please select at least one option.

Please select at least one option.
Please complete this required field.
Please complete this required field.

Thank you for reporting this profile, we are going to review it as soon as possible.