Can companies succeed by bolting AI onto an existing product, or must they be built AI-native?

Harry Stebbings · Nov 15, 2023

Project management tools like Asana or Monday could use AI foundationally — connecting to mail and calendars to proactively create tasks and shape how you spend your time — which is more than a sprinkle

The AI could become instrumental to the creation of the to-do list itself rather than an add-on

Scope: framed as a push-back on Des

6:29 20VC: How to Survive and Thrive in a World of OpenAI, Are LLMs Being Commoditised, Where Does the Value Lie; Infrastructure or Application Layer, How Apple Could Win in a World of AI, How Amazon Could Threaten OpenAI and Why Google Struggle with Des Trayn

Gokul Rajaram · Mar 16, 2026

Bolt-on AI strategies have a real ceiling; they only work when the company reframes what the product does and rebuilds the experience end to end, rather than adding a thin layer over a GPT or Anthropic model.

Adding AI search is an upgrade, whereas rebuilding search with new UX primitives is a different product; the win comes where AI changes the experience and economics, not just margin, and requires tuning the model to your own customer base.

21:22 20VC: The 8 Moats of Enduring Software Companies: How to Analyse for Durability and Defensibility in a World of AI | Why Dropouts are "AI Maxing" the World & Remote Early-Stage Companies are Dying with Gokul Rajaram

Gokul Rajaram · Mar 16, 2026

Overvalued slow-growth private software companies split into two outcomes: zombie companies that bolt on AI, fail, and get sold to PE, or companies whose leadership burns the bridges and builds a completely new AI-native product.

Bolt-on AI as a last resort doesn't succeed, and the PE exit may not even be available because PE is struggling to digest earlier purchases at reset prices; Intercom's Fin and Podium show the new-product path reaching $100M+ in a couple of years.

Scope: some verticals may instead see companies merge with each other for scale

29:39 20VC: The 8 Moats of Enduring Software Companies: How to Analyse for Durability and Defensibility in a World of AI | Why Dropouts are "AI Maxing" the World & Remote Early-Stage Companies are Dying with Gokul Rajaram

Aaron Levie · Apr 20, 2026

Software companies right now must simultaneously re-architect for the AI shift and act as a bridge so their customers aren't crushed by it

The pace of change forces both fast internal repositioning across business, product and partner ecosystem, and customer hand-holding into the future

36:09 20VC: Everyone is Wrong; We Will Have More Developers in Five Years | Why Frontier Labs Will Be Way More Valuable Than They Are Today | Are SaaS Companies Cooked: Which Thrive & Which Die with Aaron Levie, Founder at Box

Nikesh Arora · Jun 22, 2026

The long-term winners will be companies that fundamentally rethink their workflows and organizations around AI, not those that marginally adapt existing workflows

True benefit comes from redesigning the workflow rather than making the old way slightly more efficient

Scope: long term

11:46 20VC: Nikesh Arora on the Frontier Model Problem: Breadth vs Depth | The Future of Token Costs | Memory Becoming the Moat | Where Value Accrues: Infra, Models, or Apps? | Why Enterprise AI is Not Ready & Systems of Record vs Systems of Intelligence

Fred Turner · Jul 18, 2026

Had he known the AI/LLM wave was coming, he would have designed the health insurance business fundamentally differently

Health insurance is essentially a marketplace business that just moves bits around a database, and in the last 18 months AI has changed how they do nearly every workflow

Scope: retrospective on 2022 design decisions

40:37 20VC: $5BN in Revenue, 7 to 7,000 Employees in 9 Months, 206,000 Tests in a Single Day: The Craziest Story in Startups: Curative with Fred Turner

Daniel Dines · Dec 18, 2024

Incremental building is not enough in the AI era; AI-first products must be built from the ground up

Cursor was built from the ground up as an AI-first product, which is why it works

10:03 20VC: UiPath's Daniel Dines on Why Agents Do Not Mean RPA is F***** | Why We Have Reached the Upper End of Scaling Laws | The Future of Work in an Agent World and What Everyone Misunderstands About Enterprise AI

Nabeel Hyatt · Apr 4, 2025

Investors should lean toward the revolution and evolution categories and refuse adaptation bets, accepting higher risk

Spark's largest exits and most satisfying work over time came from revolution and sometimes evolution; the more disruptive journey is worth traveling even when it doesn't work out

Scope: higher risk, won't always work out

47:05 20VC: Why To Win in AI, Investors Need to Change Their Approach | Why VC is Run by Principals and Associates and is a Broken System | The Bull Case for Anthropic & Whether Deepseek Changes Their Strategy with Nabeel Hyatt @ Spark Capital

Anton Osika · Aug 18, 2025

How applications are built needs rethinking: Lovable currently encodes decades of software best practices, but future applications will all contain AI and extremely seamless payment and checkout flows

The future of software won't look like the accumulated best practices Lovable currently reproduces

16:40 20VC: Lovable CEO Anton Osika on $120M in ARR in 7 Months | The Honest Truth About Defensibility and Unit Economics for AI Startups | The State of Foundation Models: Long Grok, Short OpenAI, Why | Replit vs Lovable vs Bolt: What Happens

Harry Stebbings · Mar 2, 2026 · hedged

Bolt-on AI strategies at scaled companies are fundamentally different from AI-native companies in how they build, integrate, and adopt AI as a platform

How you build, integrate and approach AI as a platform is very different when AI is foundational rather than added on

Scope: distinction is scale vs. AI-native, not public vs. private

26:19 20VC: Monday.com CEO on Is SaaS Dead: Will Everything Be Vibe Coded | Will Systems of Record Become Valueless Databases in an Agentic World | Will LLMs Own the Value in the Application Layer with Eran Zinman

Des Traynor · Nov 15, 2023

A truly AI-first project management product is possible and would deserve the label, but nobody has built it yet — the category is still in the salt-and-pepper phase

The job involves reading email, checking tasks, looking at GitHub commits and corresponding with a calendar; a bot that did all of that would qualify, but it doesn't exist

Scope: concedes Harry's point about what's possible; claim is about the present state

6:53 20VC: How to Survive and Thrive in a World of OpenAI, Are LLMs Being Commoditised, Where Does the Value Lie; Infrastructure or Application Layer, How Apple Could Win in a World of AI, How Amazon Could Threaten OpenAI and Why Google Struggle with Des Trayn

Des Traynor · Nov 15, 2023

Most of the AI features shipped by scale-up software companies in the last few months are pointless, driven by marketing signalling rather than user value

Teams reach for the easiest-to-use AI features and reverse-engineer a demand for them inside the product so they can wave a flag and get the marketing juice of being 'AI something'

Scope: specifically about the scale-up space; analogy to early JavaScript effect libraries

51:18 20VC: How to Survive and Thrive in a World of OpenAI, Are LLMs Being Commoditised, Where Does the Value Lie; Infrastructure or Application Layer, How Apple Could Win in a World of AI, How Amazon Could Threaten OpenAI and Why Google Struggle with Des Trayn

Nabeel Hyatt · Apr 4, 2025

Most AI startups today are either evolved products that aren't good enough or adapted products with a coat of AI paint on top, and almost none are revolution

We're in an 'industrialization of startups' playbook where founders churn out arbitrage plays to get through an incubator and raise a seed round

46:43 20VC: Why To Win in AI, Investors Need to Change Their Approach | Why VC is Run by Principals and Associates and is a Broken System | The Bull Case for Anthropic & Whether Deepseek Changes Their Strategy with Nabeel Hyatt @ Spark Capital

Eran Zinman · Mar 2, 2026

Monday's early AI features were 'AI dust' sprinkled on top of the product that did not change the value delivered to customers

They didn't yet understand what to build, so they shipped things like AI formula builders and AI columns that sugarcoated the existing product

Scope: describes Monday's own first phase, roughly 1.5-2 years ago

26:51 20VC: Monday.com CEO on Is SaaS Dead: Will Everything Be Vibe Coded | Will Systems of Record Become Valueless Databases in an Agentic World | Will LLMs Own the Value in the Application Layer with Eran Zinman

Christian Lanng · Sep 27, 2023

Copilots are the last breath of existing enterprise applications trying to remain relevant, and are even worse coming from startups

The problem isn't navigating a bad application with AI help — the better solution is to remove the bad application and talk straight to the AI

57:05 Copilots are a transitional artifact not the future

Jerry Murdock · Feb 28, 2026

Bolting AI onto an existing company can work well enough to get an exit, but building AI-native makes you a better company

Thinking AI-native forces deeper engagement with what's actually driving the change

5:46 Bolt on can buy an exit ai native builds the better company

Jerry Murdock · Feb 28, 2026

Bolt-on AI strategies only work if you are world class at what you're attempting — trying is not enough

Like the Olympics, lots of people try to win a gold medal and very few do

24:04 Bolt on only works if you are world class

Howie Liu · Aug 25, 2023

Building AI into an existing product for an existing customer base is necessarily slower than building a clean-slate new product, but it buys the advantage of compounding existing value.

Incumbents face resource scarcity and cannot cut corners on things like security the way a cavalier new startup can; incumbency is a spectrum and it's always a tradeoff.

17:09 Incumbents move slower integrating ai but gain compounding value advantage

Des Traynor · Nov 15, 2023

Intercom was in pole position for the AI era because it went all-in on messaging from the start, and needed no substantial rebuild

A chatbot is the intersection of two megatrends, AI and messaging, and Intercom already had messaging as its core mode of interaction within a complete CS platform

7:40 Architecture already aligned with ai mode avoids rebuild cost

Paul Erlanger · Jun 27, 2026

The future is mostly fast prototyping tools but Figma-style tools hold a strong position, because humans still want manual control over the interface

People want to feel in control, so the winning pattern is generate-then-manipulate; it is easier to layer a commoditized LLM onto human-centric software than for LLM-first tools like Lovable to build the human-centric layer

29:33 Layering ai onto human control beats ai first tools

David Frankel · Aug 8, 2026

Every company is now an AI company — saying you're not AI is like saying you're not using the internet — so 'is it AI?' is not a useful financing filter.

Any competent founder uses the most contemporary tools available; the real differences are in approach and domain depth

9:43 Ai is universal so the ai label carries no signal

Nabeel Hyatt · Apr 4, 2025

AI startups can be usefully sorted into three categories — adaptation (bolting AI onto an existing product), evolution (a genuinely new workflow native to the medium), and revolution (a platform that could only exist because of the new technology)

The same lens worked in the mobile revolution — NYT mobile site as adaptation, Instagram as evolution, Uber as revolution — and maps cleanly onto AI, where 2023's incumbent launches were adaptation and Granola/Descript are evolution

44:29 Adaptation evolution revolution framework distinguishes cosmetic ai from genuine ai native innovation

Rob Go · Jun 23, 2023

He has reversed from AI skepticism to being all in, because the present moment resembles the early days of the internet or cloud, and every team he invests in should be AI native even if the company isn't an AI company.

He can now distinguish teams that build every software product expecting AI tools to create efficiencies and new capabilities from teams that treat AI as a passing fad — the same tell that separated internet-native founders from others.

48:08 Every founding team should build ai natively even outside ai labeled companies