Insights

Writing on complex product design.

Articles on product design, UX architecture, design systems, and complex enterprise systems - organized by persona and topic, drawn from real projects.

13 articles

Menora Insurance

FoundersAI Strategy in the Product

Where AI Belongs in a Field-Heavy Product

Founders keep asking for an AI story. In a field-heavy product like Menora's agent zone — many fields, many roles, many conditional rules — the real opportunity isn't a homepage chatbot. It's using AI to narrow a complex space: surface the right fields for this role and context, suggest the filter path, flag a combination that doesn't add up. In regulated, high-stakes domains, augment the expert and keep them able to verify. Don't replace judgment with opaque output.

Read

Movement

FoundersAI Strategy in the Product

Where AI Belongs in a Checkup App, and Where It Doesn't

Founders want an AI story. Here's the precise version for Movement (Ichilov's checkup app): users aren't sick, they're checking in — it has to feel like a benefit, not a diagnostic engine. That rules out the flashy temptation: an AI that interprets results and plays doctor. In a clinical-adjacent product, an overconfident explanation isn't a feature, it's a liability. Where AI could earn its place is the experiential layer — the self-assessment questionnaires, where completion rate is everything. Adaptive, conversational flow that lifts completion without overstepping scope. Don't aim AI at your riskiest surface to look modern. Aim it at the metric you live or die by, and keep clinical truth with the people qualified to deliver it.

Read

Enterprise Mobile

FoundersAI Strategy in the Product

AI in a Frontline App Should Remove Taps, Not Add a Chatbot

Founders keep asking me to add AI to their B2E app. My first question: does it remove taps from the frontline worker, or just look good in the deck? The best AI in enterprise mobile is invisible — read a photo to auto-fill a report, turn a scanned form into payroll codes, flag a timesheet error before it bounces. Not a chatbot. Two rules: it must degrade gracefully into offline mode, and keep a human confirm step anywhere it touches pay. The best AI here is the AI nobody notices, because it just deleted three taps.

Read

Ministry of Defense

FoundersAI Strategy in the Product

Why a Trust-First Portal Says No to Flashy AI

Everyone wants AI somewhere visible in the product. On the Ministry of Defense public portal, the smarter strategy was to keep it out of the citizen-facing surface entirely. When a product's whole job is institutional trust, an AI speaking on a ministry's behalf isn't a feature - it's a liability. Use it behind the curtain to make the work; keep the public surface deliberate and accountable. Strategy is as much what you withhold as what you ship.

Read

Marketing & Campaign Analytics

FoundersAI Strategy in the Product

A Grounded AI Strategy for Analytics Products

Founders, 'add AI' is not an analytics strategy. AI to do what is. The best place for AI in a data product is the interpretation layer — turning raw numbers into prioritized, explained insights, with the numbers kept underneath as verifiable ground truth. And don't stop at narration. Attach an action to every AI insight, so the model routes to a decision, not just a paragraph. An insight that produces no action is one to question. Augment the judgment, don't hide it.

Read

Hacker U

FoundersAI Strategy in the Product

Where AI Belongs on a Conversion Site — and Where It Doesn't

Every founder wants to know where AI fits. My honest take for the HackerU site: less visibly than you'd hope, more usefully than you'd expect. It's a conversion marketing site — no LMS to bolt an 'AI tutor' onto. The real strategy lives behind the experience: sharpening which audience sees which path, refining the personalized entry layer over time. Restraint is the strategy. AI should make the existing decision quieter and sharper — not become a feature users have to notice and operate.

Read

Zammit

FoundersAI Strategy in the Product

The Honest AI Question for a Made-to-Order Platform

Founders always ask where AI goes in the product. The disciplined answer beats the hopeful one. Zammit ran on a configurator engine where value came from deterministic, fabricable output — a contractor sets dimensions and edge profiles, and a laser cutter has to make it, every time. So the bar for any AI is simple: does it respect the manufacturing constraints? The opening isn't replacing the configurator. It's reducing the cognitive load of letting non-experts configure complex made-to-order products without CAD and without a salesperson. Anchor AI to the human job you removed — and let the engine's hard constraints, not the hype, set the boundary.

Read

Shibolet & Co.

FoundersAI Strategy in the Product

Why I Didn't Reach for AI Here

Founder confession: on a project everyone expects an AI story for, my strategy was restraint. Shibolet's hub needed warmth and reasons to show up — not an algorithm guessing who you should befriend. In a law firm, that reads as the opposite of warm. The disciplined version of an AI strategy is knowing where intelligence serves the user and where it erodes the trust the product runs on. Earn engagement with human-centered design first. Let AI earn its way in later.

Read

Shomrat Hazorea

FoundersAI Strategy in the Product

Where AI Earns Its Place in a Furniture Buying Journey

The lazy AI strategy was obvious: bolt a chatbot on the furniture site, announce 'AI,' move on. I pushed the other way. The real job to be done: people stall on high-ticket furniture because choice is overwhelming and stakes feel high. So our assistant, Asaf, points intelligence at exactly that moment — conversational visual cards and a live results feed replacing an intimidating filter wall. Founders, the trap is treating AI as a category to check off instead of a tool aimed at one painful job. Do one thing that visibly moves someone from paralysis to a shortlist. That beats sprinkling 'AI' everywhere as novelty.

Read

BlackSwan

FoundersAI Strategy in the Product

AI That Earns Trust Before It Earns Autonomy

Founders, in regulated products AI has to earn trust before it earns autonomy. On BlackSwan, analysts and investigators are personally accountable for every screening call. An AI that quietly decides for them is worse than none — because a result you can't explain is one you can't defend to a regulator. So I designed toward AI as an accelerant inside an auditable system: surface what matters, suggest what to investigate, keep the human deciding. Autonomy demos well. Legibility is what actually ships in regtech.

Read

Government & Public Sector

FoundersAI Strategy in the Product

AI in Public Services: Assist, Don't Decide

Founders pitching AI into public services start with the wrong question: 'where can the model decide?' The right one is 'where can it assist while a human or a clear rule still owns the decision?' In govtech, trust is the whole game. An opaque model making calls on someone's benefits is a headline waiting to happen. So I use AI where it lowers effort without raising stakes — plain-language explanations, finding the right form, visible pre-fill the user can correct, faster triage to a person. Never where the user can't understand or appeal it. The mature AI strategy isn't the flashiest. It's the one a regulator, a journalist, and a stressed citizen can all live with.

Read

Beehive

FoundersAI Strategy in the Product

Where AI Belongs in Beehive — and Where It Doesn't

Every founder is asked when AI goes in the product. With Beehive, the real question wasn't "can we add a chatbot?" It was: where would intelligence quietly make this feel more personal — without undermining a brand whose entire value is human expertise? For a network of 100 financial experts, AI belongs in service of personalization, never as a substitute for the experts. Good AI strategy starts with the brand promise and protects it.

Read