Skip to content
Aakash GuptaAakash Gupta

Complete Course: AI Agent Products (with Warp.dev CEO Zach Lloyd)

Zach Lloyd, CEO of Warp ($1M ARR growth every 10 days), reveals how to build AI agents that developers actually pay for. He breaks down the exact frameworks for profitable agent development, and shares his controversial take on why most AI products are built the wrong way. ----- Full Writeup: https://www.news.aakashg.com/p/zach-lloyd-podcast Transcript: https://www.aakashg.com/the-ai-pms-guide-to-building-profitable-agents/ ---- Timestamps: 00:00 Intro 02:00 How big is Warp 08:27 How he made Warp 14:53 Ads 15:42 Why Most AI Agents Fail at Launch 16:37: UX process on an AI agent 19:35 Live Demo: Building AI Agent with Warp 29:32 Ads 31:05 Systems That Drive Adoption 38:25 Workflow to build AI Agents 46:15 How to choose right metrics for your Agent 53:00 How to Actually make Money with AI Agents 59:24 Why Traditional SaaS Pricing Breaks 1:06:00 AI Agents will change the way you work 1:11:25 Outcome-Based Pricing Strategies 1:17:50 Roadmap to Build Ai Agents 1:10:55 Outro ---- Thanks to our sponsors: 1. Vanta: Automate compliance, manage risk, and prove trust - http://vanta.com/aakash 2. Kameleoon: Leading AI experimentation platform - http://www.kameleoon.com/ 3. Amplitude: The market-leader in product analytics - https://amplitude.com/session-replay?utm_campaign=session-replay-launch-2025&utm_source=linkedin&utm_medium=organic-social&utm_content=productgrowthpodcast 4. The AI Evals Course for PMs: Get $1155 off with code ‘ag-evals’ - https://maven.com/parlance-labs/evals?promoCode=ag-evlas ---- Key takeaways: 1. Find Where People Hate Rules: Look for workflows where users write algorithms, formulas, or complex syntax. Zach discovered developers were already telling computers what to do through terminal commands - they just needed to do it in English. 2. Apply the $20 Intern Test: Ask what tasks you'd give a smart college intern. Focus on time-consuming work requiring intelligence. If you wouldn't pay someone $20/hour for it, don't automate it. 3. Check If It Really Hurts: Test against four criteria: frequency (weekly use), expertise barrier (requires learning), Google dependency (users search "how to"), and time cost (saves hours). Most failed AI features only hit one criterion. 4. Make Old Things Smarter: Don't add chat panels. Make existing interfaces understand natural language. Users already express intent through formulas and commands - make those conversational instead of teaching new behaviors. 5. Help When People Get Stuck: Surface agent suggestions during error states, not randomly. When users hit errors, auto-suggest "Let agent fix this" with one-click activation. 6. Start Small, Grow Trust: Begin with simple, safe capabilities and add tools as users get comfortable. Week 1: basic requests. Month 3: complex workflows with approval gates. 7.Don't Let Power Users Bankrupt You: Fixed subscriptions make power users unprofitable. A user with 2,000 monthly interactions costs $80 in API fees while paying $50 subscription. 8. Price Like Cell Phone Plans: Use base subscription plus overages. Give predictable costs for normal usage but protect margins on heavy use. Users understand and prefer this model. 9. Charge for Results, Not Usage: When possible, price based on outcomes like resolved tickets or completed tasks rather than conversations. Aligns your success with customer value creation. 10. Catch the Next Wave: Three phases exist - autocomplete (done), interactive agents (current opportunity), full automation (future). Most industries are still in phase one, creating first-mover advantages. ---- Where to find Zach: LinkedIn: https://www.linkedin.com/in/zachlloyd/ X: https://x.com/zachlloydtweets?lang=en Warp: https://www.warp.dev/ ---- Where to find Aakash: Twitter: twitter.com/aakashg0 LinkedIn: linkedin.com/in/aagupta/ Newsletter: news.aakashg.com #aiagents #productmanagement #artificialintelligence ---- About Product Growth: The world's largest podcast focused solely on product + growth, with over 187K listeners. Hosted by Aakash Gupta, who spent 16 years in PM, rising to VP of product, this 2x/week show covers product and growth topics in depth. Subscribe and turn on notifications to get more videos like this.

Aakash GuptahostZach Lloydguest
Sep 26, 20251h 11mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

Warp CEO shares playbook for launching profitable AI agents

  1. Warp’s growth inflection came from shifting from “better terminal” to a native agentic development environment where users can issue English intents directly in the core input, not a side chat panel.
  2. Most AI agents fail at launch because they bolt chat onto old workflows instead of embedding intelligence into the product’s native UI and catching users at moments of real need.
  3. Warp drives adoption through in-context assistance (an “agentic autocomplete” moment) that suggests next actions during errors, creating immediate aha moments and deeper engagement.
  4. Reliable agent products require rigorous evaluation: public benchmarks plus internal task-based eval harnesses, coupled with real-world failure-pattern mining from anonymized user interactions.
  5. Monetization is constrained by variable inference costs: traditional per-seat SaaS breaks, pushing toward hybrid subscription + overages or outcome-based pricing where outcomes are measurable (e.g., customer support resolution).

IDEAS WORTH REMEMBERING

5 ideas

Start with a real user problem, not an AI feature.

Lloyd emphasizes classic product fundamentals: identify a deep pain point first, then hypothesize where inserting “intelligence” into an existing workflow measurably reduces time or friction.

Avoid “put chat in my app” as your core integration strategy.

A generic chat panel is a thin moat and often feels non-native; durable agentic UX comes from integrating intent expression directly into the product’s primary interaction surface.

Design for the product’s native abstraction, then let English replace complexity.

Warp’s unlock was treating the command line as an execution interface and letting users express the same intent in English—replacing memorized commands and complex rule systems with higher-level intent.

Give agents tools, then make tool-use reviewable.

Warp’s agent can run commands, edit/read files, and more, but the UX includes visibility into diffs and actions so the user can review, iterate, and safely accept changes like a code review.

Ship fast to validate demand, then invest heavily in evals.

Warp launched agent features quickly without evals to confirm users wanted the workflow, then built benchmark-driven and internal evaluation systems once failures became the primary limiter.

WORDS WORTH SAVING

5 quotes

We’re adding… over a million dollars in ARR every 10 days, and that’s accelerating.

— Zach Lloyd

There’s this one very, very powerful new primitive, which is intelligence.

— Zach Lloyd

What I would think is… not necessarily the right path is… ‘let me put chat in my app.’

— Zach Lloyd

What’s the version of autocomplete for agentic work?

— Zach Lloyd

The typical SaaS pricing mechanism of a fixed price per seat subscription… doesn’t work that well with agents.

— Zach Lloyd

Warp growth metrics and inflection pointNative agent UX vs chat bolt-onsAgent tools: terminal commands, file edits, web/file readsContext, memory, and “Rules” for agent behaviorOnboarding via in-the-moment suggestionsMeasuring engagement, retention, and conversion signalsEvals: benchmarks, internal harnesses, and production feedback loopsPricing, margins, and value capture across the AI stackCompetitive positioning: IDE forks vs pure CLI vs Warp’s ADE90-day roadmap for PMs to skill up (hands-on building)

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome