ELECTE 4.0 est en ligne — l'AI Agent est arrivé.Voir les nouveautés
ELECTE Media

Tout ce que nous publions, au même endroit.

Les articles se trouvent ici ; les podcasts, la newsletter et le Quarterly ont chacun leur propre espace — tous accessibles depuis ce hub.

Tous les articles

Tous les articles

Dernières réflexions

Des notes de l'équipe derrière ELECTE — non éditées, dans la langue dans laquelle elles ont été écrites.

Fabio Lauria
The whole idea of licensing a font is about to collapse under its own weight. When every brand can grow a living typeface tuned to its exact voice, you stop buying letters and start buying models, and the foundries who see that first inherit the next decade. @electesrlsur InstagramBrussels cut red tape for small business, and the only firms with high-risk AI live today to benefit are the ones that lobbied for the cut. The European Digital SME Alliance publicly said no thanks. When your supposed beneficiary disowns the gift, check who's actually holding the receipt.sur ThreadsEveryone argues about what companies collect — I think that's the wrong fight. Consent governs the data going in, but the real value is captured in what gets inferred about you, and three CJEU rulings plus a €200M DMA fine tell me Europe's institutions have already figured this out. Full breakdown in this week's newsletter — link in bio. @electesrlsur InstagramEveryone read the AI Omnibus as Brussels finally listening to small business. The adoption data says something else entirely. Eurostat 2025: 55.03% of large EU enterprises use AI, versus 17% of small ones. A 38-point gap. The high-risk provisions that got pushed from August 2026 to December 2027 apply to systems that are already in production - and small firms overwhelmingly do not have them. The barrier data confirms it. Cost is named as an AI obstacle by 74.7% of large EU enterprises and roughly 30% of small ones. That inversion is the whole story: regulation and conformity spend are large-firm problems. SMEs are constrained by skills, data quality and scale - none of which a deadline extension addresses. What the Omnibus left untouched is exactly what SMEs actually pay: Article 4 AI-literacy duties, transparency and disclosure obligations, per-system governance in the low four figures annually. The deferral covered obligations they rarely trigger. Telling detail: the European Digital SME Alliance warned the package risks creating new advantages for foreign tech incumbents, while Connect Europe and GSMA pushed for the rules to apply even later. When the named beneficiary disowns the relief, follow the lobbying. The practical read for founders: 74% of AI-generated economic value already flows to 20% of organisations. Sixteen extra unregulated months extends that cohort's runway, not yours. So - if compliance was never your bottleneck, what is? And does your board know the difference?sur LinkedInA rule cannot burden a company that never deployed the technology. Yet the AI Omnibus - 16 months of relief on high-risk AI obligations - was sold as red-tape relief for European SMEs. Eurostat: 17% of small EU firms use AI. 55% of large ones do. Guess who had systems live. — Fabio Lauria (@fabiolauria92) Aug 28, 2026Voir sur XThe €330k QMS isn't what kills SMEs. Art. 6(3) is: you must document why you're not high-risk. Proving innocence is a fixed cost, and it lands on the 17% of small EU firms using AI at all.sur BlueskyEvery "fix later" is a loan against a version of you who will be busier, more scared, and buried under the mess you left. I stopped worshipping velocity the day I realized my fastest quarters built my slowest years. @electesrlsur InstagramThe AI Act mentions SMEs 38 times and still gets the economics backwards. Free sandboxes cut testing costs. They don't cut governance costs. And governance is the fixed overhead. A €500m firm amortises €330k. A 45-person team spends 20% of quarterly R&D on one system. Same rule, opposite outcome.sur ThreadsWe launched The ELECTE Quarterly. One question about AI and human judgment per issue. Just the depth and attention the question deserves. The first issue is available now:  quarterly.electe.net/ — @ElecteSrl Aug 28, 2026Voir sur XThe EU AI Act mentions SMEs 38 times. That isn't protection. It's an admission that the drafters knew the economics were a problem. Here is what most analysis misses. The Act's SME carve-outs target the wrong cost line. Free sandboxes and size-scaled conformity fees reduce testing and administrative charges, real savings, roughly €50k-€100k. But the dominant expense is the internal Quality Management System, technical documentation and ongoing conformity engineering. The Commission's own CEPS study puts that at up to €330,000 upfront plus €71,400 a year, per high-risk product. That number does not scale with headcount. A €500m enterprise amortises it. A 45-person firm reports spending 20% of quarterly R&D on a single system. A near-fixed cost applied to firms of wildly different size is a regressive tax. Eurostat 2025 makes the effect visible: 55% AI adoption among large EU enterprises versus 17% among small ones, with the gap widening (13.9% growth vs 5.8%). The second-order effect concerns me more than the first. Fixed compliance costs quietly push European SMEs from being AI providers to being AI deployers, buying compliant systems from whoever can absorb the overhead. Deployer obligations are lighter. Strategic position is weaker. Two questions worth sitting with: If your AI roadmap survives only as a deployer, what proprietary layer do you still own? Is your documentation a byproduct of how you build, or a retrofit project you haven't budgeted yet?sur LinkedInEveryone reads the EU AI Act as a leash on Big Tech. Read the cost structure instead. Compliance is a near-fixed overhead, and fixed costs are a regressive tax. The Act didn't cage the incumbents. It handed them a moat, and SMEs are paying to dig it. — Fabio Lauria (@fabiolauria92) Aug 27, 2026Voir sur XRevenue per employee is the wrong stat. AI-native firms at $3.48M/head carry inference in COGS; 2015 SaaS didn't. Compare gross profit per head, or you're pricing a subsidy as operational leverage.sur Bluesky
Anne Anderson
The asset was never the raw data. It's the inference. Build your compliance posture around that, or keep defending the wrong perimeter.  newsletter.electe.net/privac… — @ElecteSrl Aug 27, 2026Voir sur XWhy do European SMEs struggle with AI adoption? Find out in the video 5 challenges.sur InstagramThis year marks 5 years since I started working in IT. I started in Sales, moved to Customer Success, then closer to Product. Over time, I became interested in how different parts of a company work together, and how to make things better. Turns out, that was Operations all along. — Anne Anderson (@AnneAnders1999) Aug 26, 2026Voir sur XThis year marks five years since I started working in IT. During that time, I’ve worked across quite a few different areas — and eventually found myself in Operations. I started in Sales, which gave me a good understanding of how a business works from the commercial side. Then I moved into Customer Success and got a much better understanding of the customer side: what people need, what problems they face, and where things can be improved. After that, I wanted to get closer to the product itself. I started working more with Product and Development and taking on projects that involved different teams and areas of the business. That gradually changed the way I looked at my role. Instead of focusing on one function, I became more interested in how different parts of the company work together — and how to make those processes more efficient. The roles changed quite a lot over these five years. The one thing that stayed consistent was that I was always looking for ways to make things work better. Apparently, that was Operations all along.sur LinkedIn5 years in IT — and I ended up in Operations. I started in Sales because it seemed like one of the best ways to understand how a business actually works. From there, I moved into Customer Success, where I got much closer to customers and the problems they were trying to solve. Later, I wanted to understand what happens on the other side of the product — so I moved closer to Product and Development and started taking on projects across different parts of the company. Over time, my role became less about one particular function and more about connecting the dots between them. Looking back, the roles changed quite a lot. The thing that didn’t change was my tendency to look at something and think: «There has to be a better way to do this». Apparently, that was Operations all along.sur InstagramAdopting AI ≠ Being AI-ready. Find out in the video why SMEs need strong data foundation.sur Instagram