<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>FDE Műhely — Blog</title><description>Forward deployed engineering in practice: process discovery, evals, exception handling and getting AI into production.</description><link>https://fdemuhely.hu/</link><language>en-GB</language><item><title>Why most AI pilots fail</title><link>https://fdemuhely.hu/en/blog/why-most-ai-pilots-fail/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/why-most-ai-pilots-fail/</guid><description>Pilots don&apos;t fail on the model. They fail because nobody mapped where it was supposed to go. Five recurring causes, and what to do about each of them.</description><pubDate>Tue, 14 Jul 2026 00:00:00 GMT</pubDate><category>For leaders</category><category>AI deployment</category><category>pilots</category><category>change management</category><category>ROI</category><author>hello@fdemuhely.hu</author></item><item><title>The documented process and the real one</title><link>https://fdemuhely.hu/en/blog/documented-vs-real-process/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/documented-vs-real-process/</guid><description>Every company has two processes: the one written down, and the one that runs. AI deployments build on the first — here&apos;s how to capture the second.</description><pubDate>Tue, 30 Jun 2026 00:00:00 GMT</pubDate><category>Discovery</category><category>process discovery</category><category>discovery</category><category>exception handling</category><author>hello@fdemuhely.hu</author></item><item><title>How an AI discovery actually works, step by step</title><link>https://fdemuhely.hu/en/blog/how-an-ai-discovery-works/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/how-an-ai-discovery-works/</guid><description>What an FDE does during the two to four weeks of discovery, who they talk to, what they pull from your systems, and exactly what you get at the end.</description><pubDate>Tue, 16 Jun 2026 00:00:00 GMT</pubDate><category>Discovery</category><category>discovery</category><category>process mapping</category><category>method</category><author>hello@fdemuhely.hu</author></item><item><title>The ROI matrix: what&apos;s worth automating, and what isn&apos;t</title><link>https://fdemuhely.hu/en/blog/roi-matrix-what-to-automate/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/roi-matrix-what-to-automate/</guid><description>Not every process deserves attention. How we rank workflows on volume, recoverable time, risk and feasibility — and why the first project isn&apos;t the biggest one.</description><pubDate>Tue, 02 Jun 2026 00:00:00 GMT</pubDate><category>Discovery</category><category>ROI</category><category>prioritisation</category><category>discovery</category><category>business value</category><author>hello@fdemuhely.hu</author></item><item><title>Evals: how to measure whether it actually works</title><link>https://fdemuhely.hu/en/blog/evals-how-to-measure/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/evals-how-to-measure/</guid><description>“Looks good” isn&apos;t a result. How a non-deterministic system becomes a measurable number, and how you find out where the errors really come from.</description><pubDate>Tue, 19 May 2026 00:00:00 GMT</pubDate><category>Evals</category><category>evals</category><category>measurement</category><category>quality</category><author>hello@fdemuhely.hu</author></item><item><title>Building a golden dataset in practice</title><link>https://fdemuhely.hu/en/blog/building-a-golden-dataset/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/building-a-golden-dataset/</guid><description>How to assemble a measurement set from your own historical cases that genuinely tells you whether the system works — and the mistakes to avoid.</description><pubDate>Tue, 05 May 2026 00:00:00 GMT</pubDate><category>Evals</category><category>evals</category><category>dataset</category><category>measurement</category><category>method</category><author>hello@fdemuhely.hu</author></item><item><title>How to evaluate a task with no single right answer</title><link>https://fdemuhely.hu/en/blog/evaluating-open-ended-tasks/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/evaluating-open-ended-tasks/</guid><description>A summary or a deck can be done well a thousand ways. It&apos;s still measurable: break it into criteria and build in human feedback from the start.</description><pubDate>Tue, 21 Apr 2026 00:00:00 GMT</pubDate><category>Evals</category><category>evals</category><category>human-in-the-loop</category><category>measurement</category><author>hello@fdemuhely.hu</author></item><item><title>Human in the loop: where approval should stay</title><link>https://fdemuhely.hu/en/blog/human-in-the-loop/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/human-in-the-loop/</guid><description>Human-in-the-loop isn&apos;t a compromise, it&apos;s a design decision. Where to keep a person in the process, and how to stop approval degrading into rubber-stamping.</description><pubDate>Tue, 07 Apr 2026 00:00:00 GMT</pubDate><category>Deployment</category><category>human-in-the-loop</category><category>risk</category><category>design</category><author>hello@fdemuhely.hu</author></item><item><title>Unhappy paths: why everything rides on exception handling</title><link>https://fdemuhely.hu/en/blog/unhappy-paths-exception-handling/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/unhappy-paths-exception-handling/</guid><description>A process goes right one way and wrong several hundred ways. Build only for the happy path and you&apos;ve built a demo, not a system.</description><pubDate>Tue, 24 Mar 2026 00:00:00 GMT</pubDate><category>Deployment</category><category>exception handling</category><category>reliability</category><category>design</category><author>hello@fdemuhely.hu</author></item><item><title>If you can&apos;t see what it did, you won&apos;t trust it</title><link>https://fdemuhely.hu/en/blog/audit-trail-and-trust/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/audit-trail-and-trust/</guid><description>An audit trail isn&apos;t a nice-to-have, it&apos;s a precondition for going live. What to log in an agent, in what form, and what people actually do with it.</description><pubDate>Tue, 10 Mar 2026 00:00:00 GMT</pubDate><category>Deployment</category><category>logging</category><category>reliability</category><category>transparency</category><category>compliance</category><author>hello@fdemuhely.hu</author></item><item><title>Don&apos;t replace your ERP. Build on it.</title><link>https://fdemuhely.hu/en/blog/dont-replace-your-erp/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/dont-replace-your-erp/</guid><description>A company that spent two years migrating doesn&apos;t want another migration. Why building on existing systems is the only viable route, and how we do it.</description><pubDate>Tue, 24 Feb 2026 00:00:00 GMT</pubDate><category>Deployment</category><category>ERP</category><category>integration</category><category>change management</category><author>hello@fdemuhely.hu</author></item><item><title>Shadow mode: going live without anyone carrying the risk</title><link>https://fdemuhely.hu/en/blog/shadow-mode/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/shadow-mode/</guid><description>The system runs but doesn&apos;t decide. Why this is the most useful six weeks of a deployment, what to measure during it, and when you can move on.</description><pubDate>Tue, 10 Feb 2026 00:00:00 GMT</pubDate><category>Deployment</category><category>deployment</category><category>shadow mode</category><category>risk</category><category>measurement</category><author>hello@fdemuhely.hu</author></item><item><title>Which model should you use? Wrong question.</title><link>https://fdemuhely.hu/en/blog/which-model-should-you-use/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/which-model-should-you-use/</guid><description>Model choice is a measurement, not a belief. How to build so the model stays swappable, and when switching is actually worth the effort.</description><pubDate>Tue, 27 Jan 2026 00:00:00 GMT</pubDate><category>Fundamentals</category><category>model selection</category><category>architecture</category><category>cost</category><author>hello@fdemuhely.hu</author></item><item><title>Token maxing: how to burn a year&apos;s budget in three months</title><link>https://fdemuhely.hu/en/blog/token-maxing/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/token-maxing/</guid><description>Handing everything to the model is more expensive, slower and less reliable than code would have been. Where the money goes, and what to do instead.</description><pubDate>Tue, 13 Jan 2026 00:00:00 GMT</pubDate><category>For leaders</category><category>cost</category><category>architecture</category><category>FinOps</category><author>hello@fdemuhely.hu</author></item><item><title>How to measure AI ROI so finance accepts it</title><link>https://fdemuhely.hu/en/blog/measuring-ai-roi/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/measuring-ai-roi/</guid><description>Three groups of metrics matter: cost savings, risk reduction and revenue impact. How to turn each into defensible numbers without cheating.</description><pubDate>Tue, 16 Dec 2025 00:00:00 GMT</pubDate><category>For leaders</category><category>ROI</category><category>business value</category><category>reporting</category><author>hello@fdemuhely.hu</author></item><item><title>A 30-day plan to become a forward deployed engineer</title><link>https://fdemuhely.hu/en/blog/30-day-plan-become-an-fde/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/30-day-plan-become-an-fde/</guid><description>Four weeks, week by week: build a real agent, make it survive reality, measure it, and learn to defend it in business language. With concrete checkpoints.</description><pubDate>Tue, 18 Nov 2025 00:00:00 GMT</pubDate><category>Career</category><category>career</category><category>FDE</category><category>learning</category><category>agents</category><author>hello@fdemuhely.hu</author></item><item><title>When not to use AI — six cases from practice</title><link>https://fdemuhely.hu/en/blog/when-not-to-use-ai/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/when-not-to-use-ai/</guid><description>One of the most valuable sentences an FDE can say is “no model belongs here”. Six situations where AI is a worse answer than a rule or a better form.</description><pubDate>Tue, 28 Oct 2025 00:00:00 GMT</pubDate><category>Fundamentals</category><category>architecture</category><category>decision</category><category>cost</category><author>hello@fdemuhely.hu</author></item><item><title>The two sides an FDE has to master</title><link>https://fdemuhely.hu/en/blog/the-two-sides-of-an-fde/</link><guid isPermaLink="true">https://fdemuhely.hu/en/blog/the-two-sides-of-an-fde/</guid><description>An FDE isn&apos;t the average of a consultant and a developer — it&apos;s the best of both. What that means in skills, and how to build the weaker side.</description><pubDate>Tue, 14 Oct 2025 00:00:00 GMT</pubDate><category>Career</category><category>career</category><category>skills</category><category>FDE</category><author>hello@fdemuhely.hu</author></item></channel></rss>