New Episodes Every Wednesday @ 📺 Youtube | 🎧 Spotify | 📖 Substack
We launched Verticals to give Vertical AI the space it deserves: a weekly, first‑principles show co‑hosted by Luke Sophinos (founder) and Euclid’s own Nic Poulos. Each episode runs in three parts:
Vertical Market Pulse (what mattered this week)
Vertical Titan (a deep-dive founder interview)
Vertical Playbook (a repeatable strategy you can apply)
This week’s guest: Ilir Sela, founder & CEO of Slice—the vertical platform powering tens of thousands of independent pizzerias nationwide. Ilir has done the thing most VCs used to say can’t be done: turn a hyper‑fragmented “small TAM” SMB market into a durable, scaled billion-dollar platform.
Full episode guide below, including: a round-up of our Market Pulse, highlights from our Titan, and a breakdown of our Vertical Playbook for founders.
I) This Week’s Vertical Market Pulse
1) EliseAI’s raise, valuation, and the voice‑AI moat
EliseAI announced a fresh round led by a16z at a reported ~$2B valuation, doubling down on housing while expanding in healthcare. The headline raised a familiar question: is voice AI really vertical—or will horizontal platforms overrun the category? Our view echoes what we’ve written before: vertical voice wins on distribution, integrations, and workflow fit—not just model quality. Data (and the right data exhaust) is the moat; GTM and service design are the shield.
2) Epic enters ambient scribing; Abridge answers with prior auth
Epic’s move to ship its own scribe validates the category—and turns a partner into a competitor for AI scribes that scaled on top of Epic’s rails. With ~42% hospital market share, Epic’s distribution is the story; the question is whether ambient notes are the end state or merely the wedge into downstream clinical ops. Abridge’s immediate counter—codesigning real‑time prior auth with Highmark—is the right instinct: go deeper into workflow where value accrues. For founders, this is a live case study in whether to partner with the system of record or become one.
3) What the next model wave actually unlocks
We discuss Elad Gil’s recent framing on how successive model releases create new company‑formation surfaces. Whether or not you buy any specific benchmark delta, the app‑layer opportunity keeps shifting rightward—from copy and search → to agents in real workflows → to systems that own the authoring layer. More about the authoring layer in our prior essay “Emerging Playbooks in Vertical AI.”
II) Vertical Titan: Ilir Sela (Founder / CEO @ Slice)
Ilir’s story is an incredible lesson in starting small + the power of vertical compounding:
Start where the money moves. Slice’s first wedge was dead simple: move phone orders online for independent shops. That single channel shift increased AOV, reduced errors, and created a direct line to the customer for reorders.
Bootstrap to proof. Before raising meaningful outside capital, the company had already scaled to material GOV with healthy unit economics—then selectively brought in operators and investors who added functional muscle (sales, finance, operations).
Build deliberately from wedge to OS. The evolution: independent websites → the Slice app (reorder engine) → in‑store OS (POS & kitchen routing) → phones answered → supply chain (boxes, bags) → membership. ACV per shop rose an order of magnitude as Slice moved from “online orders” to “how the shop runs.”
Partner until you must own. When your wedge sits at the edge of the system of record, you can rent distribution—until the core platform wants your margin. Plan for both futures from day one.
Two memorable lines from Ilir you’ll hear in the episode:
“Independent, not alone.” Slice is a reverse franchise, offering the shared stack and economies of scale of a Domino’s—while allowing owners to maintain their own brand. It’s instructive positioning for many synthetic rollup startup models.
“Quality beats novelty.” For Slice, AI voice serving drives a 90% completion rate on transactions. But humans still drive 98%+. That’s an extremely meaningful difference for low-margin pizza shops. His take: until AI clears that bar, humans stay in the loop.
III) Vertical Playbook: Unbundling the Franchise
Why it works
Franchises win on standardized ops, buying power, and shared data—but impose uniformity that doesn’t fit hyper‑local markets. The reverse‑franchise play takes the capabilities without the brand lock‑in.
How to run it (the Slice blueprint):
Wedge in by moving analog to digital. Start at the highest‑frequency, highest‑ROI workflow (e.g., ordering). Price for adoption (simple, transaction‑based).
Capture the authoring layer. Become where the work starts (menuing, orders, tickets). This is the kernel of your data model and future system of intelligence.
Integrate “left & right.” Left into marketing & demand (reorder engine, CRM); right into ops (kitchen routing, phones, staff scheduling).
Verticalize supply. Add SKUs the operator already buys (boxes, bags, disposables) to increase ACV and defensibility.
Graduate to “membership.” Bundle services so the operator “runs on you.” The company’s perception shifts from app → platform → OS → indispensable partner in everything related to the business.
Founder litmus tests
Is the incumbent system of record a fast mover? If yes, design to coexist and harvest data; if not, plan to replace it over time.
Is the status quo market fragmented (#2–#10 players hold good market share)? Partnerships and distribution come easier, and leverage is easier to gain.
Can your first product win with near‑zero integrations? If yes, you can scale while you build the hard plumbing. But you should think about the end-game early.
What we debated on‑air
Voice AI defensibility. Luke leans “data + incumbent distribution win,” Nic stresses workflow specialization and GTM complexity by vertical. We agree: horizontal platforms will either specialize / verticalize or get outflanked by those who do.
Partner vs. platform. You can partner with the system of record—until you can’t. The Epic / Abridge dynamic is the canonical case right now. If your wedge is peripheral, you likely need to plan a path to core.
What the next wave enables. Model improvements matter, but founders should orient around data ownership, authoring layers, and closed‑loop outcomes—where value and margins live.
Next week on Verticals
Jeremy Yamaguchi @ Cabana joins us to unpack the AI-First Roll‑Up model.
Thanks to our partner
Verticals is made possible by Parafin—embedded capital for your merchants. If you’re building a vertical platform and haven’t explored how financing can drive retention and revenue, start here.
If you’re building a vertical AI company and want a sounding board, reply to this post or ping us on LinkedIn. And if you enjoyed this episode, consider subscribing to Verticals here, in addition to Euclid Insights for weekly analysis on Vertical AI.
