Why Agents Break PLG (And How to Rebuild It)

Notorious: the best founders, startups, strategies, metrics and community.

Current subscribers: 11,567, +195 since last post

Share the love: please forward to colleagues and friends! 🙏

Why Agents Break PLG (And How to Rebuild It)

Product-led growth (PLG) is built on a simple premise: let users use your product, get hooked on its value, and drive adoption organically. But what happens when users stop using and start delegating? The rise of AI agents, systems that perform tasks for users rather than empowering users to do them, is flipping the PLG playbook on its head. Instead of clicking around your interface, users now hand off an intention and watch the agent work. This shift challenges the very tactics PLG teams have honed for years, from onboarding flows to UI optimizations. It’s a sharp wake-up call: users aren’t “users” in the old sense anymore, and growth strategies must adapt.

When Users Delegate, Classic PLG Falls Apart

Traditional PLG assumes a hands-on user journey. A new user signs up, navigates UI, explores features, hits an “aha” moment after gradually discovering value, and moves themselves down the funnel. Onboarding checklists, tooltips, and gradual feature discovery are all designed to guide humans through using the product. But in an agentic UX, the paradigm flips: you’re not executing the task, you are managing the team who execute it. In other words, the user’s role shifts from operator to delegator. They express intent (“what” they want), and the agent handles the “how.”

This flip comes with serious side effects for PLG tactics:

• Onboarding flows break: An AI agent doesn’t need your cute interactive tour. If a user can simply tell the agent their goal and let it handle the steps, those guided walkthroughs and tooltip tutorials go unused. The user isn’t learning the interface; they’re outsourcing the work.

• UI optimization loses impact: We’ve spent years polishing UX, minimizing clicks, highlighting buttons, all assuming a human is clicking. But an agent might be using your product via API or script, or skipping the UI entirely.

• Feature discovery dwindles: In a direct-use world, users wander through menus, trial new features, and stumble on value. In an agent-driven world, nothing is stumbled upon, the agent executes exactly what the user requested, no more, no less. The user might never see half your features, even if the agent uses them in the background.

• “Aha” moments and usage-based activation weaken: PLG teams love to track the moment a user first realizes the product’s value (the aha moment) and the usage milestones that correlate with retention (e.g. “user uploaded 10 files”). With agents, value can be delivered in a single, delegated leap. The user’s “aha” might simply be the task got done. They didn’t click 10 times or slowly build up usage; the agent just did the thing.

In short, AI agents break the linear, user-driven progression that PLG optimizes for. Instead of a journey from novice to power-user, we get users who issue a command and expect instant outcomes. The old growth levers – optimize funnel steps, drive feature adoption, nail the UX – feel wobbly when the user isn’t actually in the driver’s seat.

Agents vs Toolkits: Examples in the Wild

This isn’t theoretical, it’s happening now. Let’s look at a few examples of agentic products and how differently they operate compared to traditional tool-like products:

• Perplexity’s Copilot: If Google is a search engine, Perplexity is an answer engine. Its AI “copilot” acts like a genius intern that does research for you: it performs the searches, reads through the links, summarizes what matters, and provides answers with references. Instead of the user manually querying and clicking, the agent handles the workflow. For PLG, onboarding and UI don’t matter much, the user’s loyalty hinges on result quality, not interface mastery.

• Windsurf: Windsurf is an AI-native IDE designed for a world where AI writes 90% of the code. Developers describe features and review AI-generated changes, rather than manually building everything. The product is structured around delegation, not exploration. Traditional activation metrics like “first 5 edits” or “time to Hello World” don’t apply when the agent handles the heavy lifting.

• Granola: Granola is an AI notepad that joins meetings, transcribes conversations, and handles follow-ups like drafting recap emails and creating summaries. Users barely touch the product. All the engagement happens under the hood, but value is still delivered. Notes written, emails sent, summaries created, all by the agent. PLG metrics like feature usage or editing time are irrelevant here.

Each of these products shows the same shift: agents treat products as tool, while users just expect results. The agent doesn’t need to explore your features or appreciate your UI. And the user doesn’t care how the result got done – only that it did.

Rebuilding PLG for an Agent-Native World

If AI agents change how users experience value, then PLG needs to evolve. Here’s how to rebuild it:

1. Optimize for intent capture, not feature engagement

Your “aha moment” is no longer clicking feature X, it’s expressing an intent and getting what you want. Focus onboarding around helping users articulate their goal clearly. Build interfaces (or APIs) that make it dead simple to say: this is what I need. Your new activation event is a successful delegation.

2. Make outcomes measurable and shareable

Since users won’t click around anymore, make results the product. Track outcomes: reports generated, emails drafted, actions completed. Then make them shareable. Let users send, export, or broadcast those results easily. That’s your new growth loop.

3. Rethink onboarding as “delegation setup”

It’s not about teaching users where the buttons are. It’s about giving the agent access, context, and preferences to do the job right. Did they connect their calendar? Upload data? Define success criteria? That’s the new onboarding checklist.

4. Design for trust, transparency, and reversibility

Users need to feel safe letting agents act on their behalf. Let them preview outputs, approve actions, see what happened, and undo mistakes. The more control you give users over the agent, the more they’ll let go — and that’s where retention lives.

The Bottom Line

AI agents break classic PLG. They short-circuit UI journeys, skip onboarding, and render usage-based metrics meaningless. But they also open the door to a new kind of growth, one rooted in intent, outcomes, and delegation. If you can capture what the user wants, deliver it autonomously, and prove it’s working, you win.

PLG isn’t dead. But your user might not be the user anymore. The agent is. And it’s time we started building for that.

Thanks for reading. By way of background, I am an early-stage investor at Wing and a former founder. Please reach out to me on X @zacharydewitt or at zach@wing.vc. Some of the early-stage PLG + AI companies that I have the privilege to work with and learn from are: AirOps, Copy.ai, Deepgram, Hireguide, Slang.ai, Tango and Workmate.

Operating Benchmarks (from PLG Startups):

I will continue to update these metrics and add new metrics. Let me know what metrics you want me to add (zach@wing.vc)

Organic Traffic (as % of all website traffic):

• Great: 70%

• Good: 50%

Conversion rate (website → free user):

• Great: 10%

• Good: 5%

Activation rate (free user → activated user):

• Great: 50%

• Good: 30%

Paid conversion rate (free user → paid user):

• Great: 10%

• Good: 5%

Enterprise conversion rate (free user → enterprise plan):

• Great: 4%

• Good: 2%

3-month user retention (% of all users still using product after 3 months):

• Great: 30%

• Good: 15%

Conversion from waitlist to free user:

• <1 month on waitlist: ~50%

• >3 months on waitlist: 20%

For more detail on acqusition rates by channel (Organic, SEM, Social etc), please refer to this prior Notorious episode.

Financial Benchmarks (from PLG Public Companies):

Financial data as of previous business day market close.

Best-in-Class Benchmarking:

15 Highest EV/ NTM Revenue Multiples:

Complete Dataset (click to zoom):

Note: TTM = Trailing Twelve Months; NTM = Next Twelve Months. Rule of 40 = TTM Revenue Growth % + FCF Margin %. GM-Adjusted CAC Payback = Change in Quarterly Revenue / (Gross Margin % * Prior Quarter Sales & Marketing Expense) * 12. Recent IPOs will have temporary “N/A”s as Wall Street Research has to wait to initiate converge.

Recent PLG + AI Financings:

Seed:

• Agree, an AI-powered agreements platform that automatically detects and labels all input fields and signature blocks in contracts, has raised $7.2M at a $30M valuation. The round was led by Pelion Venture Partners, with participation from Better Tomorrow Ventures, Everywhere Ventures, 8-Bit Capital, FirsthandVC, Hustle Fund, Singh Capital Partners, Trust Fund and Blank Ventures.

• Fastino, an AI model architecture company building task-specific language models, has raised $17.5M. The round was led by Khosla Ventures, with participation from Dropbox Ventures, Valor Equity Partners and Insight Partners.

• Korl, an AI-first platform that generates customer-ready presentations and messaging for roadmap, QBR, and renewal conversations, has raised $5M. The round was led by MaC Venture Capital and Underscore VC, with participation from Perceptive Ventures.

• Rork, an AI platform for building mobile applications using only natural language prompts, has raised $2.8M. The round was led by Cheyenne Partners and Innovius Capital, with participation from Salesforce Ventures, Snowflake Ventures and Norwest Venture Partners.

• SSOReady, an open-source authentication infrastructure for B2B software companies, has raised $3.3M. The round was funded by Y Combinator.

• TensorStax, an agentic platform designed to build and maintain data pipelines, has raised $5M. The round was led by Glasswing Ventures, with participation from Bee Partners, S3 Ventures, Mana Ventures and Gaingels.

• WisdomAI, an AI-powered business insights platform that uses intelligent agents to deliver deep understanding of company data, has raised $23M. The round was led by Coatue Management, with participation from Madrona Venture Group, GTM Capital and Menlo Ventures.

• Luzia, an AI-powered personal assistant that helps users manage daily tasks by answering questions and providing personalized support, has raised $13.5M. The round was led by Prosus Ventures, with participation from Khosla Ventures and Monashees.

Series A:

• Glide Finance, an AI-powered growth platform for credit unions and community banks, has raised $15M. The round was led by Acrew Capital, with participation from Pathlight Capital and Pear.

• Sett, an agentic AI platform that enables mobile gaming studios to create data-driven marketing and in-game content at scale, has raised $15M. The round was led by Bessemer Venture Partners, with participation from vgames and Saga Ventures.

• Trek Health, an AI-powered platform that helps healthcare organizations benchmark and negotiate their payer contracts using real market data, has raised $11M at a $28M valuation. The round was led by Madrona Venture Group, with participation from Founder Collective, Correlation Ventures, LifeX Ventures, SNR and Accrete Health Partners.

• Unblocked, an AI-powered coding assistant that answers contextual questions about lines of code, has raised $20M. The round was funded by Radical Ventures and B Capital.

Series B:

• 7Learnings, an AI-powered optimization platform for B2C companies, has raised $10M. The round was led by Accel and New Enterprise Associates, with participation from Tru Arrow Partners, Valor Equity Partners, Overmatch and StepStone Group.

• Carta Healthcare, a startup using AI, machine learning, and clinical expertise to automate the collection and analysis of clinical data, has raised $18.25M. The round was led by UPMC Enterprises, with participation from CU Healthcare Innovation Fund, Rex Health Ventures, MemorialCare Innovation Fund, Tampa General Hospital, Memorial Hermann Health System, Mass General Brigham Ventures, Storm Ventures, Paramark Ventures and Frist Cressey Ventures.

• Recraft, an AI-powered image generation platform designed for professional designers with a focus on brand consistency, has raised $30M at a $148M valuation. The round was led by Accel, with participation from Khosla Ventures, RTP Global and Madrona Venture Group.

• Relevance AI, an AI agent operating system that enables companies of all sizes to build and deploy AI agent workforces, has raised $24M. The round was led by Bessemer Venture Partners, with participation from Insight Partners, King River Capital and Peak XV Partners.

Series C:

• Classiq, a quantum development platform for designing, optimizing, analyzing, and executing quantum algorithms, has raised $110M. The round was led by Entrée Capital, with participation from Wing VC, IN Venture, Samsung NEXT Ventures, Phoenix Financial, Norwest Holdings, NightDragon, Clal Industries, Qbeat Ventures, Team8, NEVA SGR, HSBC Holdings and Hamilton Lane.

• Parloa, an AI Agent Management Platform that enables enterprises to build, test, and deploy AI agents, has raised $120M at a $1B valuation. The round was led by Altimeter Capital Management, General Catalyst and Durable Capital Partners, with participation from Senovo, EQT Ventures, Mosaic Ventures and Riverpoint Capital.

• Sanity, a content operating system designed to help organizations manage content across digital surfaces, has raised $85M. The round was led by GP Bullhound, with participation from Heavybit, Threshold Ventures, ICONIQ Growth, Shopify Blue Cloud Ventures and Monochrome Ventures.

• Statsig, a product development platform that helps software engineering teams test, analyze, and roll out new features, has raised $100M at a $1.1B valuation. The round was led by ICONIQ Growth, with participation from Sequoia Capital and Madrona Venture Group.

Series D:

• AI21 Labs, a startup developing enterprise AI systems and foundation models, has raised $300M. The round was led by Nvidia and Alphabet.