
Sign up to save your podcasts
Or


Filip Najdovski co-founded EaseAccess24, a Swedish AI startup for digital accessibility, at 18, backed by T-Meet, a 25-year-old Swedish accessibility company. EaseAccess24 audits websites against WCAG standards and uses AI to fix what it finds, alongside a broader product line that includes AI-powered assistive tech for blind and deaf users.
This is the first episode of The AI Product, a show about how AI features actually fit into real products. I talk to founders, product people, and engineers who've actually done this and have experience with it. Every episode digs into the trials, the mistakes, the misconceptions, the unexpected, the actual product thinking and strategy behind the decisions, and how cost factors into building these features. The goal is to clear the mystique off "AI features" and figure out where they actually belong in software.
In this one, Filip and I get into:
Follow Filip!
Filip Najdovski: https://www.linkedin.com/in/filipnajdovski/
EaseAccess24: https://www.easeaccess24.com/
Follow me and my newsletter!
My Linkedin: https://www.linkedin.com/in/gocestojcevski/
Full breakdown and takeaways: https://gocestojchevski.substack.com/
Timestamps:
(00:35) Intro to EaseAccess24 and the scale of digital accessibility
(02:01) How the product works: audits, AI fixes, and the widget
(03:31) The EU Accessibility Act and who has to comply
(05:44) Co-founding at 18 and the Swedish parent company T-Meet
(09:03) Why it's not "just a compliance company"
(10:28) User research with people with disabilities, and the taxi story
(14:03) "AI slop" and building AI for the sake of it
(16:31) The AI feature stack, from audits to an unreleased product for deaf users
(18:33) Going backwards from LLMs to rule-based fixes
(22:32) Pricing AI when marginal cost isn't zero
(26:04) Why they moved to credits
(28:23) Deciding what's worth building, and the Miro example
(32:05) Designing for non-deterministic AI outputs
(34:59) Proactive vs. reactive AI, and the monitoring tool
(37:13) Onboarding B2B clients vs. designing for B2C users
(40:32) The hardest and most surprising parts of building with AI
(44:48) Testing without a dedicated QA team
(47:02) How AI changed sprint planning, and building internal tools overnight
(51:09) Bringing AI into legacy businesses
(52:36) Finding talent and hiring in a small market like Macedonia
(56:09) Why young people thrive in startups
(59:15) Age, timing, and a founder's daily rhythm
(62:00) Signal vs. noise, delegation, and ego
(68:25) Being blunt instead of sugarcoating
From the publisher's feed
I'm obsessed with how AI actually gets implemented in products. So I started talking to the people who've actually done it: founders, product people, and engineers, to get an inside look at how it…
What actually happened. The results, the outcomes, the problems they ran into, the unexpected stuff nobody plans for, and what they learned from it, so people building AI features don't make the same mistakes.