Overview
Holmes was founded in 2026 by Robin Praet, Robbrecht Delrue, and Sofie Buyse, three Belgian operators each with a prior exit. The team builds an autonomous software-testing platform aimed at engineering teams shipping AI-generated code at high velocity. Rather than asking developers to hand-write and maintain test suites, Holmes observes how a product is actually used, infers the end-to-end journeys that matter, and turns those into self-updating tests inside the CI/CD pipeline.
The platform deploys a set of specialised AI agents covering happy paths, edge cases, responsive layouts, accessibility, and error recovery. The founders position Holmes around a structural shift: traditional QA workflows cannot keep pace with AI-assisted development, so testing itself has to become autonomous. At launch the company was working with 30 design partners, with broader access rolling out in 2026.
The founder credentials anchor the pitch. Praet and Delrue previously co-founded Smartendr, a hospitality ordering platform that grew to 600+ venues and 400,000+ monthly users before being acquired by OrderBilly in September 2025. Delrue had earlier scaled go-to-market at Teamleader. Buyse was the first hire at Henchman, the Ghent legal AI tool acquired by LexisNexis for EUR 136.1M in June 2024.
In May 2026 Holmes announced a EUR 1.1M pre-Seed round led by Syndicate One, with participation from Roeland Delrue and Willem Delbare (co-founders of Aikido Security), Louis Jonckheere (co-founder of Showpad), serial entrepreneur Thomas Van Overbeke, and funds NewSchool.vc, RDY Capital, and 100IN. The advisory bench includes engineering leaders from Collibra, Luzmo, Upsellplus, and Lighthouse.
Funding history
- Pre-Seed: EUR 1.1M (May 2026), led by Syndicate One, with NewSchool.vc, RDY Capital, 100IN, and angels Roeland Delrue, Willem Delbare (Aikido Security), Louis Jonckheere (Showpad), Thomas Van Overbeke.
Key people
In the news
- In her latest blog, Clarice M Holmes, RN and VP of Clinical Practice at Sagility, explores the four rights of better care and why a more thoughtful approach to clinical operations can make a meaningful difference for health plans and their members: https://lnkd.in/ez6323yW #ClinicalServices #UtilizationManagement #PriorAuthorization
- Pourquoi la fraudeuse Elizabeth Holmes cartonne à nouveau aux États-Unis, cette fois sur grand écranOnze ans après le début de sa chute, Elizabeth Holmes, qui prétendait avoir révolutionné les tests sanguins, refait les gros titres suite à la parution d'un film immersif et troublant, qui fait un carton dans les salles obscures.
L'Echo Entreprises · 20 Oct 2026
- Imagine walking into work each day not knowing what is in store for you. (Experienced QAs won't need to imagine this, but just play along, okay?) While a stressful idea for some, for the right people this is a gold mine. Such variety in tasks corresponds to vast learning across many different areas. You might start the day analyzing trends, jump on a supplier call alongside procurement, and get called in to consult on a recurring deviation before getting a jog on to make it to the training session you're delivering on the other side
- Alexey Kryukov gave Holmes 30 minutes on SauceDemo. No setup. No codebase. No integrations. 10 pages explored. 12 findings. Every one reproducible. That's the value without connecting anything. Now imagine connecting your tools and putting Holmes in your workflow. That's where the real magic starts. But findings alone aren't the point. Anyone can wire up an AI agent that reports bugs. The problem: you can never fully trust it. Is it real? Can someone reproduce it? Did it test what you think it tested? That's where Holmes makes
- I gave an AI QA agent 30 minutes to test SauceDemo. It came back with 12 findings. Sounds impressive. But can someone else reproduce them? After a chat with Robbrecht Delrue, I tried Holmes with no codebase context and no custom Bug Hunt instructions. It explored 10 pages on its own. Then I manually checked all 12 findings. A few examples: - Checkout completed with an empty cart. I even generated a $0.00 PDF receipt. - Reset App State cleared the cart counter, but the item remained visible until refresh. - Three spaces were accepted
- One summer Friday night, we shipped a new feature at Smartendr, our previous company. It broke the checkout for our hospitality customers. Half an hour later, the phone was on fire. That stress never went away. Today every team feels it. AI writes more code than ever, but someone still has to check that it works. That's why we built Holmes. Every time you ship, Holmes tests your product like a real customer. It logs in, it pays, and it finds what's broken before your customers do. On 21 October I'm pitching Holmes as a finalist at
- We’re building the founding engineering team at Holmes. 🚀 Looking for exceptional Engineers and AI Engineers who want to join early, take real ownership and build with us from Ghent. Know someone? Send them our way.
- We're growing the engineering team at Holmes. 🚀 We're looking for exceptional Engineers and AI Engineers to join us early and build Holmes with us. A few months ago, Holmes was pre-revenue. Today: → €1.15M raised pre-revenue → Backed by Syndicate One, 100IN, NewSchool.vc, RDY.VC, Aikido Security, ... → Launched our product in May → 25+ paying customers → More companies testing Holmes every week We're moving fast. Demand is growing. Holmes is an AI-native QA platform that autonomously tests software like a real user. You'll work on
Related profiles
Something wrong or missing? Send an update. Fixed within 24 hours.