Atomic

The ugliest workflows are often the best wedges. A wedge is your get-in-the-door product that earns you the right to expand. It should be easy to deploy, fast to value, and good enough to make the market trust you with the next problem. The mistake a lot of founders make is choosing a wedge based on what looks clean in a demo. Clean workflows are usually already mapped: dashboards, fields, checklists, reports, approvals. Ugly workflows are where the real work lives: Voicemails, angry emails, PDFs, fax, attachments, half-filled forms, undocumented handoffs, conversations that never make it into the system. Simple workflows make nice software features. Ugly workflows make great AI wedges. Why? Because ugly workflows have 4 things founders should care about: -High labor intensity -Broken or messy intake -Clear downstream action -Weak incumbent coverage If you can ingest the chaos, structure it, and trigger action, you can own the workflow before the incumbent even knows the work exists. That’s why some of the best wedges in the AI era don’t start as systems of record. They start as systems of intake, triage, and orchestration. Take Owner.com for example. They didn’t try to beat Toast by going deeper on operations. They shifted the basis of competition to revenue and front-of-house economics. Different job. Different buyer urgency. In the AI era, the best wedge is often not the cleanest workflow. It’s the messiest one with real labor, real urgency, and real economic consequence. The workflows everyone hates are usually the ones nobody has properly owned. That’s where the wedge is.

What most founders get wrong about AI: A lot of founders still assume AI will follow the old SaaS playbook: Build software Sell seats Minimize services Win on speed and scale That works in low-stakes workflows. But in high-consequence industries, that mental model breaks. Take power development as an example. If you’re software is helping businesses decide where to build infrastructure, “mostly right” is useless. 99% accuracy isn’t impressive when the 1% error can cost millions, kill a project, or destroy trust. That’s what makes companies like Euclid Power interesting. Nic Poulos and I just interview their Founder & CEO Jacob Sandry . Euclid isn't just selling an AI copilot... They’re building systems that combine software, human expertise, and workflow ownership to operate in an environment where precision matters more than novelty. That’s the part many founders miss: The best AI companies may not look like pure software businesses. They may look more like: Productized services Vertically integrated workflows Trust-heavy systems Outcome-driven businesses Not because that’s less scalable. Because that’s what the market actually pays for. In a lot of industries, customers don’t want “AI.” They want fewer mistakes, faster decisions, and someone to own the result. That’s why vertical AI is so compelling. The opportunity isn’t just to add intelligence to software. It’s to redesign the entire workflow around reliability. And in markets where errors are expensive, the winners won’t be the teams with the flashiest model. They’ll be the ones that understand the domain deeply enough to make AI actually usable. We go much deeper in the episode. Check it out in the comments.

If you're building in vertical software or vertical AI, we're going to co-build your GTM playbook live on next week's workshop... Raj Bhaskar (Founder & CEO at Tight ) and I are doing a live workshop with the CRO whisperer, Martin Roth (Former CRO at Levelset, a Procore Company ), who built and scaled their GTM to a massive exit to Procore Technologies . Martin has spent his 10,000 hours thinking deeply about what actually works in vertical GTM, and this session will get into the practical side of it: A few topics we plan to get into: -How to build a repeatable sales motion around the realities of your market instead of forcing a generic one onto it -Where most teams get stuck as they try to scale from founder-led sales into a real GTM engine If you’re a founder, operator, or investor thinking about how vertical companies actually win distribution, this should be a good one. Registration link: https://lnkd.in/eTnD_ETm Limited to the first 25 founders.

Here is a good litmus test to identify how strong your moat is... If a smart team copied your product and had access to the same frontier models tomorrow… What wouldn’t they be able to replicate quickly? If the answer is “not much,” you probably don’t have a moat. If the answer is: -Services tie-in -Regulatory know-how -Deep Integrations -A compounding feedback loop tied to outcomes Now we’re talking. Vertical AI winners will be master moat builders, constantly ensuring that Anthropic's next release won't take them out. Something all founders should probably be thinking about quite a bit these days...

Five years ago, VC's told him adding hardware would kill his valuation (& potentially his company). They were wrong... Alexander Jekowsky , Founder & CEO of Cents , just proved something investors are finally starting to understand: in the AI era, hardware is defensibility. The stats: • 4,500+ laundromat locations powered • $1B+ in annual payments processed • 1 in 6 laundromats in the US • Contrarian bet that everyone said was wrong Three playbooks that scaled: 1️⃣ Build Hardware In-House – Don't outsource to contract manufacturers. Hire recent grads from top hardware companies. Give them equity. Let them build. Alex found his head of hardware (an Apple grad) through a network of senior engineers he befriended. No bidding wars. Just hungry talent. 2️⃣ Partner With the Channel, Don't Disrupt It – Most founders see distributors as legacy middlemen to beat. Alex made them his GTM engine. His team attended 60 distributor trade shows in 2 months. They proved they were committed for the long haul. Today, that network is one of Cents's primary growth engines. 3️⃣ Do the Math for Investors – Skip negotiation. Instead, build a sum-of-the-parts valuation model projecting to 2031. Include all the market comps and their multiples. Build a share price calculator. Walk in and say: "Here's the formula. Here's the price. We're looking for the right partner." The ones who got it moved fast. The bigger picture: Software is infinitely replicable. Hardware is not. Even if competitors build identical software, they can't easily steal your customers. Switching requires ripping out terminals from every machine, installing new hardware, retraining staff, and managing downtime. That's friction that software-only products don't have. This is vertical SaaS done right... Unsexy market. Contrarian bets. Hardware moat. $140M Series C just announced. Full episode in first comment.

Today Matia landed on Israel's Top 50 Most Promising Companies list Most of you already know, but for those who may not: This list is 15+ years old. You can't apply. You can't pay to play. Its curated by global VCs and insiders. Focused on private companies with off-the-charts momentum. Every year, all eyes on it. Past names in their early days include JFrog ($6B), Riskified ($3B), and Innoviz Technologies ($2B)... This year's cohort features some of the strongest AI and infrastructure companies operating in Israel, and Matia is leading the Data category. Huge congrats to the teams who were highlighted!

Founders in “unsexy” industries make the same recruiting mistake over and over: They apologize for the market. They say things like: “We’re not as exciting as OpenAI, but…” “We know this isn’t the sexiest category, but…” That positioning kills you. Because top talent doesn’t just want status, they want to truly work on something that matters. And a lot of the so-called unsexy markets are actually way more meaningful than the flashy ones. Supply chain. Construction. Waste. Insurance. Manufacturing. Healthcare ops. Trades. Logistics. These industries move the real economy. They determine whether goods arrive. Whether claims get paid. Whether buildings get built. Whether families get care. Whether entire sectors function. The best founders know how to tell that story. Not fake story. Not polished brand story. True story. The room you’re sitting in right now? Almost everything in it was shipped, installed, insured, financed, repaired, or maintained by an industry most software people ignore. That’s the pitch. You’re not building another toy for people already drowning in software. You’re modernizing industries the world actually runs on. And ironically, that mission often attracts stronger people over time. Because once someone has worked on a “hot” company with fuzzy impact, the appeal of solving a real, painful, economically important problem gets a lot bigger. The takeaway for founders: Don’t downplay the category. Elevate the mission. Make the significance obvious. Some of the biggest opportunities sit inside the least glamorous places.

Sales people aren’t even really using the CRM anymore. They’re talking in Slack. Updating deals in chat. Asking an agent for pipeline changes after a call. And the CRM is just quietly updating in the background. That’s a huge shift. For 20+ years, enterprise software trained us to think the app interface was the product. Open the dashboard. Fill in the form. Click the workflow. Log the activity. Now we’re moving into a world where the UI matters less and less. The interface is the conversation. The workflow is the prompt. The software is becoming headless. That’s why Salesforce’s latest moves matter. If your users can update records, pull insights, trigger workflows, and collaborate around customers from Slack, then the CRM homepage is no longer the center of gravity. And honestly? That was always a salesperson's dream. They HATE updating the CRM. The best software doesn’t ask busy people to do more admin. It disappears into the way they already work. That’s the real promise of headless software. Not prettier UX. Less UX. If the interface becomes ambient, conversational, and agent-driven… What does “product design” even mean in a few years?

Here is a list of the ChatGPT-era Vertical AI winners (thus far): ⚖️ Harvey — AI for legal research, drafting, review, and legal workflows Serves: law firms + corporate legal teams ⚖️ Legora — AI legal platform for research, drafting, and review Serves: law firms + in-house legal teams ⚖️ EvenUp — AI for personal injury case prep, demand letters, and workflow automation Serves: personal injury law firms ⚖️ Eve — AI platform for plaintiff-side casework across the full case lifecycle Serves: plaintiff law firms ⚖️ GCAI — AI platform for general counsels across the full case lifecycle Serves: General Counsels inside businesses 🏥 OpenEvidence — AI evidence assistant built for doctors Serves: physicians + clinical workflows 🏥 Abridge — ambient AI and clinical documentation Serves: clinicians + health systems 🏥 Hippocratic AI — healthcare AI agents for patient and operational workflows Serves: providers, payors + pharma 💰 Hebbia — AI workflow / research platform for complex document-heavy work Serves: finance teams, investors + knowledge workers 🏦 Rogo — AI platform for research, analysis, and deal workflow Serves: investment banks, PE firms + finance teams 📚 MagicSchool AI — AI tools for lesson planning, differentiation, assessment, and classroom workflows Serves: teachers, schools + districts A few takeaways: Legal is one of the clearest breakout categories in AI Healthcare is right behind it The biggest winners are becoming real systems of work. Where does the next breakout cluster comes from? 🛡️ insurance 🏗️ construction 🏭 manufacturing 🧾 accounting / tax 🏢 real estate