Every company has procedures, documentation, and reports. But do processes really look the way they’re described on paper? From our experience – not quite.
In this episode of AI Automation Dojo, Andrzej Kinastowski talks with Michał Kozubski (Office Samurai) about productivity mining and process discovery. With tools like KYP AI, organizations can finally see how work is actually being done – and use that data to improve efficiency, balance workload, or identify automation opportunities.
In the conversation we cover:
🔸 Why companies invest in productivity mining – and what benefits they expect
🔸 The gap between “documented processes” and “real-life processes”
🔸 What productivity mining uncovers (hidden apps, endless copy-paste, overloaded teams)
🔸 Using data to redistribute work and improve team efficiency
🔸 How to address employee concerns and show it’s not about surveillance
🔸 Challenges around people, processes, technology, and data security
🔸 How AI supports analysis and points to potential improvements
Samurai & Friends Conference is not a public event. The website and agenda exist, but access is limited to invited guests only, you won’t find them on social mediaor on our official site. It’s an invitation-only meeting, where managers, directors and C-level executives openly share what AI, automation and process improvement really look like in practice.
The third edition brought together around 160 people and, as always, the focus was not on generic visions or inspirational slides, but on real projects: what works, wherebarriers appear, and how to measure actual ROI in automation.
In this episode, we recap the key takeaways from both days of the conference:
🔸 Why an invitation-only formula creates a very different atmosphere
🔸 Agentic automation and meta-prompting – today’s real boundaries of AI
🔸 Email automation – where classical approaches end and AI begins
🔸 How a simple idea can translate into dozens of FTE savings
🔸 End-to-end process transformation and why “lean” can’t stop halfway
🔸 Task/mining and interaction analysis – surprising insights worth paying attention to
🔸 How RPA platforms are evolving and where AI fits in
🔸 AI automation playbook – a phased framework for introducing AI (based on MIT & McKinsey reports)
This was an exchange of experiences about what already works, what doesn’t, and what to improve before committing to new AI projects.
Ever wondered how work really gets done in your company, beyond the training manuals and manager's assumptions? 🧐 These “guides” say one thing. Tuesday at 11:40 says another. Someone drags data from a PDF to Excel, retypes it into an ERP, fixes a typo, then starts again. No one calls it a process, because it’s not a process. Rather a lack of ideas on how to automate that.So we decided to watch the work and make it “work”.A tiny desktop agent captured clicks, keys, and which apps were in play. Personal stuff stayed out of scope. Data was anonymized, masked where needed, and locked down. Then AI did the boring part we never have time for. It stitched patterns together and pointed at the loops, the detours, the places where time leaks.This is called Task Mining, a powerful technology that uses software to observe and understand how people interact with their work devices to do their jobs. Forget "big brother" - this is about gaining a crystal-clear, data-driven picture of your operations, all while prioritizing security and privacy through anonymization and granular control.We break down the core concept, explaining how a small software agent captures mouse clicks, keyboard entries, and application usage, then uses AI and machine learning to identify patterns, repetitive tasks, variations, and bottlenecks.
Tune in to discover how to transform vague complaints into data-driven arguments that leadership can't ignore, leading to a workforce that's engaged, productive, and less frustrated by broken processes and terrible software.This episode was created in a partnership with KYP.ai.
🎙️ What is the real role of process improvement when more and more work is done by AI?In this episode, Andrzej Kinastowski is joined by Zuzanna Pamuła – Lean Six Sigma Black Belt and Process Transformation Manager at Office Samurai – to explore how process excellence is practiced today in large organizations, and why it still matters.
We talk about what has changed since the early days of lean projects: what tools we now use, how automation reshaped the work of business analysts, and how much of the “old school” thinking is still relevant – sometimes more than ever.Zuzanna shares examples from her experience across UBS, Heineken, Lundbeck, and now Office Samurai. We look at how tools like productivity mining can support large-scale analysis without manual effort, and how Microsoft Copilot, Teams, and Outlook are being used to speed up repetitive work – without removing the need for expertise or good judgment.There’s also a broader question in the background: what happens when organizations go too fast into automation, without first standardizing their processes? Many teams are now coming back to continuous improvement methods after realizing that automation alone doesn’t solve the deeper issues. We discuss why that’s happening, and what the right balance looks like.For anyone building a career in business analysis or process transformation, Zuzanna also shares practical advice on where to start, how to learn by doing, and what skills are worth developing – even in a world shaped by AI.🔸 What’s changed (and what hasn’t) in process excellence🔸 How productivity mining and GenAI tools are being used today🔸 When automation makes sense and when it doesn’t🔸 Why more companies return to lean, CI and harmonization🔸 Practical advice for new process analysts
Is your company smarter than an ant colony? 🐜
In this episode of AI Automation Dojo, we dive into Continuous Improvement — no corporate fluff, just real stories you won’t forget.
You’ll discover three ways to build a culture of constant improvement:
Kaizen – small daily anarchy that can lead to big results.
Lean Action Workouts – three-month “change sprints” for entire teams.
Six Sigma – tools only for the true “anarchists at heart.”
Expect samurai walking into offices, ant anarchists finding better paths, and the moment someone discovered that years of printing invoices were… completely unnecessary.
🎙️ If you want your organization to be full of “big fat dead cows” (in the best possible way), this episode is for you.
Is Intelligent Document Processing (IDP) a real solution or just another overhyped technology?
In this episode, we take a practical and honest look at where IDP actually delivers results, where it struggles, and what to consider before rolling it out.
Our guest is Tomasz Wierzbicki, Senior Automation Consultant at Office Samurai, who shares hands-on experience from IDP implementations across finance, HR, and compliance-heavy industries. Together with host Andrzej Kinastowski, they go beyond theory to explore how IDP works in practice and what still requires a lot of human effort.
This episode is aimed at business and automation leaders who want to understand whether IDP is worth the investment and how to avoid common mistakes.
If you're working in operations, automation, or digital transformation and want a grounded understanding of how IDP can support your business processes – this episode isfor you.
We cover not only how the technology works, but how to evaluate ROI, manage exceptions, and decide when automation is not the right answer.
This episode was recorded in collaboration with UiPath.
We can imagine your casual day in the office starts (and quite often ends) with never-ending PDF invoices, databases out of sync, and a disorganized shared drive.
It’s the moment in which brilliant minds with potential are occupied with copy-paste routines. Every new document format is an error and automation dreams go to waste.
But what if your overworked team could finally throw it into AI to do the heavy lifting – not just reading documents, but understanding them?
In this episode of our podcast, Andrzej Kinastowski, our Head of Delivery & Managing Partner, puts Intelligent Document Processing (IDP) in the spotlight.
IDP is a technology that teaches computers to "read" documents the same way humans do, but faster and without complaining.
We are talking about comprehension and semantic understanding, where the simple recognition of printed text is not the end goal, but rather unstructured and semi-structured information is made accessible to automated workflows and systems.
What topics are covered in this episode?
🔸 What makes IDP such a powerful tool?
🔸 What are the seven key stages of a typical IDP workflow?
🔸 How do we still remember about "human in the loop”?
🔸 What are the real and measurable benefits of implementing IDP into your workflow?
🔸 In which industries IDP works best?
Finally, we take a leap into the future and explore how IDP is becoming a key component of hyperautomation, the rise of multimodal IDPs, and the transformative impact of Generative AI and Large Language Models (LLMs), shifting capabilities from mere data extraction to knowledge discovery and creation through techniques like Retrieval Augmented Generation (RAG).
If you're ready to fight your company's "paper monster" and transform "unstructured chaos into structured data," stick around – it's going to be less painful than your last expense report.
This episode was created in partnership with UiPath.
RPA not enough? How LLMs + UiPath Agents could unlock better automations in your processes.
We built actual working flows using UiPath Agents and Maestro and here’s what really happened.
What started as “let’s see what this does” turned into a working prototype that screens resumes, extracts data from HR docs, talks to the orchestrator and even escalates queries to humans.
Our devs came in skeptical. They left saying: “I spent weeks building an agent in Python. With UiPath? I just pasted the prompt and it worked.”
That’s... not nothing.
In this episode of AI Automation Dojo, we explore where UiPath’s Agent Builder and Maestro actually help, where theystill fall short, and how they could open up real use cases for teams without deep AI skills (or time).
Yes, this is early tech.
Yes, some parts are still rough.
But for once, it feels like “hyperautomation” might actually live up to its name.
This episode was recorded in collaboration with UiPath.
Sharpen your ears and unleash your curiosity – we’re diving headfirst into Agentic Automation ⚔️Putting agents into your existing automations can add a layer of smart decision-making – without starting from scratch. And no, GenAI won’t replace your team tomorrow. But it can waste a lot of money if you don’t know what you're doing.Andrzej Kinastowski, our Head of Delivery & Managing Partner, shares his first-hand experience on UiPath AI Agents and Maestro orchestration after Office Samurai was selected as an early partner for testing these tools. We explore what makes them a game-changer, what they really do (and don’t), and how they fit into complex enterprise automation scenarios – from anti-phishing bots to recruitment copilots.Naturally, we sent our top samurai to investigate. Their mission: learn, break, and report back.What we found? Genuinely surprising: 🔸 How agents use tools, orchestrator processes, and even each other🔸 What “context” really means in AI Agents🔸 Why escalation logic is crucial for trust🔸 Why humans won’t fall out of the loop🔸 How testing agents works – and how “LLM as judge” changes QA🔸 Where Maestro wins over long-running workflows🔸 A real-life example: modeling a recruitment process with GenAI, CV parsers, action center and layered agents UiPath’s agents + Maestro orchestration have only been in the wild for a few weeks.We’re just getting started.But even with the wrapping paper still on, we see that this combo, put over your good old RPA setup, unlocks the doors we thought were sealed shut (and probably guarded by a grumpy dragon).There comes a hero with superpowers. Hyperautomation. And we see that right now it might actually come out of the fantasy tale and finally be real.About damn time.This episode was recorded in collaboration with UiPath.
GenAI won't replace your team tomorrow – but it can waste millions today.
In this episode, we cut through the AI hype (again).
Remember when RPA was supposed to eliminate 70% of accounting jobs? Didn’t happen.
Now GenAI is making the same promises – louder, faster, and with even bigger budgets.
Andrzej Kinastowski is joined by Dominik Jaskulski, co-founder of Office Samurai, for a brutally honest look at:
🔸 Why most enterprise AI projects still fail
🔸 How to avoid burning money on licenses no oneuses
🔸 What real AI assistants and agents can (andcan't) do in 2025
🔸 How RPA hype cycles help us predict whereGenAI is really going
You'll also hear about Clara, CoPilot, corporate FOMO, and what happens when “just buy 5,000 licenses” becomes your AI strategy (spoiler: it's not great).
Alright, AI adventurers and office warriors — it’s time for round two.
In this episode of the AI Automation Dojo by Office Samurai, Andrzej dives into the wild world of AI Agents. You’ve seen the hype: “They’ll save us!” “They’ll ruin us!” “They’ll take your jobs!” “They’ll create new ones!”
This episode isn’t about fantasy tech. It’s about what’s real today — and what’s dangerously overhyped. Like the “agents” that turned out to be teams of underpaid freelancers clicking buttons behind the scenes. We break down what AI assistants can actually do right now (spoiler: it’s a lot), and what makes an agent truly autonomous — not just a fancy FAQ bot in disguise.
Meet Cyber Ola – our HR alter ego living inside Microsoft Teams, answering questions, creating docs, and sometimes just saying: “I don’t know, talk to real Ola.” That last bit? Way harder than it sounds.
From real RAG pipelines to Jotform APIs and PDF workflows – this isn’t a vision deck. It’s how we’re doing it today.
Spoiler: The real threat isn’t that AI will take your job. It’s that people will sell you snake oil while pretending it already did.
Alright, gather 'round, knowledge seekers and corporate warriors!
In the thrilling inaugural episode of the Office Samurai Podcast, the one and only Andrzej unveils his groundbreaking "Kinastowski Principle."
Now, this isn't your grandma's dusty old business theory. Oh no. This is a razor-sharp tool for slicing through the tech hype and figuring out what's actually useful for your business.
Picture this: two axes of decision-making doom! On one side, you've got Usefulness – you know, how much actual work this shiny new tech does and whether it'll make the bean counters smile with a hefty ROI. On the other, the ever-elusive, yet undeniably powerful, Coolness Factor. Yes, "coolness." Because let's be honest, sometimes a technology's ability to generate buzz and make your company look like it's living in the 22nd century is half the battle in getting anyone to sign off on it. It's like trying to convince a kid to eat vegetables – sometimes you gotta make it look like a dinosaur.
The episode then wades through examples of each, essentially making the case that if you're not betting on "Lifesavers" like Communication Mining or those shiny new GenAI Agents, you're probably still trying to figure out how to unjam the photocopier. These champs, according to Andrzej, are where the smart money is, thanks to their potent cocktail of actually being useful and being cool enough to get adopted without too much corporate whining.
Which may make you want to do a little dance.
Maybe even the Macarena.
We're not judging.
