# Szymon Paluch — AI consultant, implementation partner & keynote speaker

> Szymon Paluch helps companies put AI agents into production — through hands-on AI implementations, AI consulting (agent security audits, Fractional CAIO, Private AI infrastructure) and keynote talks. Ex-CTO with a startup exit; 10+ AI agents and systems shipped to production. Bilingual site: English at the root, Polish under /pl/.

This file is a map of the site for AI agents and assistants. The content is free to read, cite and learn from. Every page below also exists in Polish at the same path prefixed with /pl/.

## Start here
- [Home](https://szymonpaluch.com/): AI agents that don't fall over. I'm an ex-CTO building AI strategy and agents that keep working after the demo — 10+ in production for startups and enterprises.
- [About Szymon](https://szymonpaluch.com/about): Ex-CTO with an exit, Claude Certified Architect and Anthropic partner for Claude Code. I build teams and AI systems that actually run in production.

## Services
- [AI Implementations (Wdrożenia AI)](https://szymonpaluch.com/implementation): We build agents, RAG and automations that run in production. Team augmentation or end-to-end delivery — one point of accountability.
- [AI Consulting](https://szymonpaluch.com/consulting): AI advisory from an ex-CTO with three companies built and an exit behind him: agent security audits, Fractional CAIO, Private AI infrastructure.
- [Enterprise Agentic AI Masterclass (2-day workshop)](https://szymonpaluch.com/workshops): A 2-day intensive for technical teams: how to use AI agents in day-to-day coding work and how to build your own agents so they survive past the demo. Run by an ex-CTO and Claude Certified Architect. For teams that write code, up to 12 people. A room that came to listen — any audience, any size — is https://szymonpaluch.com/speaking instead.
- [AI Readiness Canvas](https://szymonpaluch.com/framework): AI Readiness Canvas - a hands-on method for scoring how ready your organization is to ship AI. 7 dimensions, concrete metrics, an action plan.

## Proof & track record
- [Case studies](https://szymonpaluch.com/case-studies): Real AI builds: a brand-message compliance agent, a digital-forensics studio, tender-spreadsheet normalization, a sales agent, and spatial ML in football. Concrete numbers from the repos, clients anonymized.
- [Speaking & keynotes](https://szymonpaluch.com/speaking): A practitioner, not a theorist: founder of Diagnostyka Digital Hub, ex-CTO of uPacjenta, ex-Head of Strategy at Golem. AI keynotes, panels and sessions. (31 verified engagements since 2020: talks, panels, podcasts and department sessions — any audience, technical or not, any room size, including a series of dates across a season. The two-day hands-on workshop at laptops is https://szymonpaluch.com/workshops instead.)

## Tools & research
- [AI readiness self-assessment](https://szymonpaluch.com/assessment): Where does your team really stand, from prompting to production? 12 questions, ~5 minutes. Get your position on the Agentic Engineering Radar and one next move.
- [Agent Engineering Radar](https://szymonpaluch.com/radar): 28 field-tested agentic techniques across 4 sectors and maturity rings. Deterministic control over non-deterministic AI — from the trenches, not the slides.
- [Agentic Engineering Radar — full report](https://szymonpaluch.com/raport): 28 practices for building LLM systems across 4 sectors and 4 maturity rings, on two axes: who holds the control flow and whether a loop verifies the output.

## Writing
- [Blog](https://szymonpaluch.com/blog/): A digital garden of patterns, build logs and essays on building AI agents that run in production. A growing, practical knowledge base on agentic AI.
- [AI Futures: OpenAI asks who holds power when AI stops needing people](https://szymonpaluch.com/blog/posts/openai-ai-futures): A good argument, the same one Anthropic published seven months earlier, and one citation that leads back to OpenAI.
- [The AI GTM stack: five stages you can buy, four handoffs you define](https://szymonpaluch.com/blog/posts/ai-gtm-stack): The four handoffs between funnel stages that no vendor sells, what an agent can run alone, and where a person still has to stand.
- [A public report written with AI: nine errors you can see in the file](https://szymonpaluch.com/blog/posts/ai-report-verification): A number that contradicts its own footnote, one methodology in three versions, and nine checks that stop all of it.
- [Mutation testing: can you trust AI-generated tests?](https://szymonpaluch.com/blog/posts/mutation-testing-ai-generated-tests): Coverage measures execution, mutation score measures assertion. How mutation works, what it found in production code, and how to run it in CI.
- [AI training for a non-technical team: how to run it](https://szymonpaluch.com/blog/posts/ai-training-non-technical-teams): One task instead of twenty tools, the sceptic against the blocker, and six failures from outside the programme. Built on a session for the TVP Television Academy.
- [What an AI agent costs and what actually drives the quote](https://szymonpaluch.com/blog/posts/ile-kosztuje-agent-ai): Six line items a quote is made of, two of which only show up after launch. How to count a deployment in days of work instead of licences.
- [Prompt injection: an instruction hidden in content the agent fetches itself](https://szymonpaluch.com/blog/posts/prompt-injection-agenci-ai): How prompt injection works, five risks specific to agents, and a checklist your team can run before the agent reaches production.
- [AI agents on a team: what training gives you and who is ready](https://szymonpaluch.com/blog/posts/agenci-ai-szkolenie): Agents are one of the last steps of AI adoption, not the first. What a team learns in two days, what it will not learn, and who should skip it.
- [Workflow Studio: observe and design Claude Code workflows](https://szymonpaluch.com/blog/posts/introducing-workflow-studio): A local dashboard that reads Claude Code's own run artifacts, plus a 12-block canvas that compiles to a runnable Workflow script.
- [Claude Certified Architect certification: the complete guide](https://szymonpaluch.com/blog/posts/claude-certified-architect-certification-guide): What the CCA Foundations exam tests, the five domains and weights, how it works, and how I prepared and passed 833/1000.
- [Claude Certified Architect practice exam: 60 free questions](https://szymonpaluch.com/blog/posts/claude-certified-architect-practice-questions): 60 original questions I wrote while preparing for CCA Foundations. Study mode, exam mode and per-domain scoring.
- [I passed Claude Certified Architect: decisions, not definitions](https://szymonpaluch.com/blog/posts/claude-certified-architect-exam): The proctoring step by step, why the exam tests decisions not definitions, and how to prepare.
- [12 rules for building agents](https://szymonpaluch.com/blog/posts/agentic-design-rules): Notes from prepping for the Anthropic certification. The rules that decide whether an agent survives production.
- [Claude Partner Badge — Claude Code: what it is](https://szymonpaluch.com/blog/posts/claude-partner-badge-claude-code): Anthropic partner badge: 8 courses, a capstone, and real rollout from demo to production.
- [awesome-ai-gtm: a map of the autonomous GTM funnel](https://szymonpaluch.com/blog/posts/awesome-ai-gtm): 138 AI tools arranged into 7 funnel stages. Not a link list, but a blueprint.
- [Model self-evaluation: the agent that grades itself](https://szymonpaluch.com/blog/posts/agent-self-evaluation-trap): The most common agentic engineering pitfall — and the fix.
- [Multi-agent consensus isn’t verification](https://szymonpaluch.com/blog/posts/multi-agent-consensus-verification): Three models agree and fail the same way. What to do instead of voting.
- [RATCHET: iterating against a frozen yardstick](https://szymonpaluch.com/blog/posts/agentic-ratchet-immutable-yardstick): Keep-if-better, else revert — a pattern you can’t game.
- [Silent failures in agentic systems](https://szymonpaluch.com/blog/posts/silent-failures-agentic-systems): Silent retry and fallback return “OK” and hide the errors.
- [Position bias: the first answer wins on order](https://szymonpaluch.com/blog/posts/llm-judge-position-bias): An adversarial panel with rotation instead of a single judge.
- [A confidence score isn’t correctness](https://szymonpaluch.com/blog/posts/llm-confidence-score-trap): Confidence is a tone of voice. Validate it from the outside.
- [Bounded Autonomy: How much freedom to give an agent?](https://szymonpaluch.com/blog/posts/bounded-autonomy): A design pattern for safe AI agent autonomy.
- [5 questions before deploying AI](https://szymonpaluch.com/blog/posts/5-pytan-przed-wdrozeniem-ai): A checklist that will save you months and thousands.
- [AI Governance: How not to lose a million](https://szymonpaluch.com/blog/posts/ai-governance-jak-nie-stracic-miliona): A governance framework that works in practice.
- [Multi-Agent Systems: When one agent isn't enough](https://szymonpaluch.com/blog/posts/multi-agent-systems): Orchestrating multiple agents and communication patterns.
- [Hybrid AI: Combining models for better results](https://szymonpaluch.com/blog/posts/hybrid-ai): When and how to combine different AI models in one system.
- [AI in logistics: A deployment case study](https://szymonpaluch.com/blog/posts/ai-w-logistyce): How we built an agent for route optimization.
- [From pilot to scale: Lessons learned](https://szymonpaluch.com/blog/posts/od-pilota-do-skali): What changes when an agent hits production.
- [AI Stack for your company: My tool choices](https://szymonpaluch.com/blog/posts/ai-stack-dla-firmy): The stack I use in every agent project.
- [Agentic AI in 2026: From experiments to production](https://szymonpaluch.com/blog/posts/agentic-ai-2026): 62% of companies are experimenting with AI agents. Only 11% have them in production.
- [Why AI strategies end up in a drawer](https://szymonpaluch.com/blog/posts/dlaczego-strategie-ai-koncza-w-szufladzie): Strategy documents vs. real change in the organization.
- [ROI from AI: How to measure it without fooling yourself](https://szymonpaluch.com/blog/posts/roi-z-ai-jak-mierzyc): The metrics that actually show the value of a deployment.
- [Legacy Systems vs AI: How to bridge old and new](https://szymonpaluch.com/blog/posts/legacy-systems-vs-ai): Integrating AI agents with existing infrastructure.
- [6 roles of a CTO in the AI era](https://szymonpaluch.com/blog/posts/6-rol-cto-w-erze-ai): How the role of the tech leader is changing.
- [AI Ethics Lead: A new role on the team](https://szymonpaluch.com/blog/posts/ai-ethics-lead): Why every AI team needs someone owning ethics.
- [Workforce 2026: AI and the future of work](https://szymonpaluch.com/blog/posts/workforce-2026): How to prepare your team to work alongside AI agents.

## Contact & booking
- Email: contact@szymonpaluch.com
- Phone: +48 500 012 445
- Book a call: https://calendar.app.google/g4JFWBua3gCprNbc6
- LinkedIn: https://www.linkedin.com/in/szymonpaluch/
- X (Twitter): https://x.com/szymonpalucheth
- YouTube: https://www.youtube.com/@szymonpaluch
- GitHub: https://github.com/hculap

## Full version
- Expanded profile for ingestion: https://szymonpaluch.com/llms-full.txt

## Machine interfaces (for agents)
- Structured profile (JSON): https://szymonpaluch.com/api/profile.json
- Services + indicative pricing (JSON): https://szymonpaluch.com/api/services.json
- OpenAPI 3.1 description: https://szymonpaluch.com/api/openapi.json
- API catalog (RFC 9727): https://szymonpaluch.com/.well-known/api-catalog
- Agent skills index: https://szymonpaluch.com/.well-known/agent-skills/index.json
- Markdown of any page: send `Accept: text/markdown`
- DNS-AID discovery: _index._agents.szymonpaluch.com (SVCB, DNSSEC-signed)