Will companies increasingly build their own internal software instead of buying SaaS?

David Frankel · Oct 14, 2024 · hedged

AI will not bring about the end of vertical SaaS; most companies still won't build their own solutions

The underlying AI tools are becoming commoditized (visible in open source), so incumbent software and data owners can access them cheaply and enrich their toolsets rather than being displaced

Scope: acknowledges others may disagree on commoditization

65:51 20VC: Investing Lessons from FC Seeding Uber, Airtable and Coupang | Why Pro Rata is the Original Sin in VC | Why Liquidity Has Died in 2024 | Why LPs are Pissed with VCs | The Hard Truth About Seed Fund Economics with David Frankel @ Founder Collective

Harry Stebbings · Dec 12, 2025

The viral 'SaaS is over' takes ignore how much implementation complexity is invisible from the outside

Guests who have actually built internally keep pointing to complexity that outside observers don't see

73:03 20Growth: How Wiz Built a $30BN Brand in Enterprise | What Worked vs What Was a Mega Failure: Lessons Learned | Why Marketers Make the Worst CMOs & What To Look for in Growth with Raaz Herzberg

Anish Acharya · Feb 9, 2026

The market has oversold software; the story that everything will be vibe coded is flat wrong

IT spend is only 8-12% of enterprise spend, so even fully rebuilding ERP or payroll saves at most that, with large risk; companies will instead point AI at extending their core advantage or optimizing the other 90% of spend

Scope: there will still be specific secular losers and disadvantaged business models

8:06 20VC: Is SaaS Dead in a World of AI | Do Margins Matter Anymore | Is Triple, Triple, Double, Double Dead Today? | Who Wins the Dev Market: Cursor or Claude Code | Why We Are Not in an AI Bubble with Anish Acharya @ a16z

Eran Zinman · Mar 2, 2026

Vibe coding will not disrupt software companies, including in consumer and SMB segments — its impact will be marginal at most

People underestimate how hard it is to maintain and evolve software over time; businesses need to focus on their core operations, and since software is a small expense for most companies, staffing a person or team to vibe code apps is a huge relative cost

Scope: allows for some marginal impact

9:25 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

Carles Reina · Apr 11, 2026

The SaaSpocalypse narrative is largely exaggerated, but the underlying trend of companies building their own applications is real

Scope: only a 'very large portion' is exaggerated, not all of it

47:42 20Sales: ElevenLabs: Why We Set a 20x Sales Quota | How to Structure Sales Compensation Plans | Customer Success: 'Total BS' or Growth Engine? | Building an AI Sales Machine: What Tools & Tactics Must CROs Adopt Today with Carles Reina

Chad Peets · May 23, 2026

Some SaaS is dead, but not all of it

45:09 20Sales: The $100M CRO Bubble: Why Anthropic Are Causing a Comp Crisis | Why You Should Never Hire From Salesforce or Service Now | How to Hire, Train and Forecase in a World of AI with Chad Peets and Chris Degnan

Harry Stebbings · Jul 13, 2026 · speculative

A plausible counter-argument is that vibe coding is merely a TAM reduction for Wix rather than an existential threat — perhaps 20% of the market defects, leaving 80%

Some subset of SMBs will build their own tools even if the majority won't

Scope: framed as a counter he puts to Avishai for confirmation

12:29 20VC: Wix's Founder on What Wall St Gets Wrong About AI and Wix | Will Base44 Win the Vibe Coding Wars | The Truth About the Economics of Vibe-Coding | The Buyback Disaster: Lessons Learned with Avishai Abrahami

Fred Turner · Jul 18, 2026 · hedged

Infrastructure-like software and deeply integrated tools such as Sentry and Slack will persist, though Slack becomes replaceable if prices rise too far

They've become part of your infrastructure and people have built many integrations and workflows on top of them, raising switching costs

Scope: Slack replacement contingent on price increases

46:43 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

David Frankel · Aug 8, 2026 · hedged

The SaaSpocalypse is likely overdone for deeply embedded, mission-critical software, and the market is underestimating the difficulty of replicating the last 5%

When billions of orders or mission-critical biotech research run through your system you are very hard to dispense with, unlike less embedded tools you can replicate by playing around with Claude

Scope: only for highly embedded products; less embedded software is genuinely easy to displace

39:48 20VC: The AI Boom Will Create Enormous Roadkill: Who Wins & Loses | Why Founders Should Never Take Multi-Stage Money at Seed | Why Triple, Triple, Double, Double is Good Enough

Maor Shlomo · Nov 24, 2025

Vibe coding will become the largest category in the software industry, with existing categories like CRM, support, task and project management folding into it

There will be a point where it is easier to build your own Salesforce-type CRM than to buy an off-the-shelf license

Scope: timing uncertain — could be a few months or a few years

7:45 20VC: Base44's Maor Shlomo on How Vibe Coding Will Kill SaaS and Salesforce | Why it is BS that Vibe Coding Platforms Do Not Have Defensibility and Bad Margins | Why He Worries About Google, Not Replit and Lovable | Why Long Anthropic, Not OpenAI?

Eran Zinman · Mar 2, 2026 · hedged

Companies should build their own AI tooling rather than buy, because tools still need heavy customization to be useful and the space is moving too fast

You still need to highly customize tools today to properly enjoy them, and things are moving so fast that building your own is the right path at this point

Scope: 'at this point' — time-bound judgment; Monday still uses some third party for support

24:10 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

Carles Reina · Apr 11, 2026

Companies should build their own internal software for many applications, including core ones like CRM, though not for infrastructure-type products like Gmail or Google Drive

A CRM is essentially just a pool of data and doesn't take much time to build; if a custom tool fits better into your ecosystem of applications and is easy enough, you should build it

Scope: ElevenLabs itself still uses Salesforce and hasn't done this; excludes large-scale products like Gmail/Google Drive

46:51 20Sales: ElevenLabs: Why We Set a 20x Sales Quota | How to Structure Sales Compensation Plans | Customer Success: 'Total BS' or Growth Engine? | Building an AI Sales Machine: What Tools & Tactics Must CROs Adopt Today with Carles Reina

Ryan Petersen · Jun 20, 2026

Selling SaaS to tech companies is going to be a tough business because those customers can now build the tools themselves

Flexport has already built its own agents and software for some workflows, which changes the leverage in vendor negotiations

44:48 20VC: Why Remote Work is White Collar Fraud | Why Revenge and Patriotism are the Best Founder Traits | Two Questions Every Founder Needs to Ask | The Wild Story of Raising $1BN from Masa Son in an Hour Long Meeting with Ryan Peterson, Founder @ Flexport

Fred Turner · Jul 18, 2026

He buys the 'SaaS is dead' thesis — companies will increasingly replace vendor SaaS with internally built software

He sees it in the contracts they're cancelling: a vibe-coded internal CRM works better, is more integrated with their agents, and killed a $600k/year Salesforce contract that nobody was using anymore

Scope: based on his own company's experience

45:21 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

Fred Turner · Jul 18, 2026

For most businesses of reasonable scale, building custom internal tools is worth the engineering hours despite maintenance cost

Custom-built systems fit your workflow better, and you replace multiple FTEs of administrators maintaining archaic platforms with one great engineer who can also ship custom features on demand

Scope: maintenance is the most challenging piece; businesses of reasonable scale; open question whether most firms have the tech resources yet

45:52 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

Fred Turner · Jul 18, 2026

Off-the-shelf insurance-specific platforms like claims systems are worth replacing with in-house builds

The vendor claim system they migrated to a few years ago is hard to use, its API barely works, data is hard to extract and they aren't allowed to manage it — yet this is how insurance companies run

Scope: their own migration completing in July

47:34 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

Matt Swulinski · Aug 15, 2026

You can build sophisticated agentic marketing systems without knowing how to code, by feeding 90-95% of your daily work through the AI and giving feedback each day

He built his marketing OS by asking Claude Code how to build an OS, then improved it daily through the terminal feedback loop

Scope: he does not code

56:56 20Growth: How to Build a $100M Growth Engine: Lessons from Wispr Flow and Superhuman | Why You Should Do Paid Ads Today and How To Do Them | How to Build the Best Referral Programs and How to Crush UGC with Matt Swulinski

David Cahn · Aug 5, 2024

In evaluating companies, investors should listen to what customers do rather than what they say — customers paying for an expensive product while saying they will rip it out and build it themselves usually keep paying

Across Marqeta, UiPath, Snowflake and Databricks, customers said the products were too expensive or a band-aid they'd replace, yet years later none had churned and the companies became very valuable; people pay for expensive products for a reason

Scope: applies especially to expensive products customers are already paying for

47:28 20VC: Sequoia's David Cahn on AI's $600BN Question | Why the Data Centre is the Most Important Asset | Servers, Steel and Power: The Core Pillars Powering the Future of AI

Harry Stebbings · Aug 5, 2024

Many great investments have been missed because investors believed customers could build the product themselves, when in practice customers rarely do

49:02 20VC: Sequoia's David Cahn on AI's $600BN Question | Why the Data Centre is the Most Important Asset | Servers, Steel and Power: The Core Pillars Powering the Future of AI

David Frankel · Oct 14, 2024

Large enterprises that claim they can build it themselves typically deliver nothing, so there remains a real business in smart software solutions

When he sold ISP and hosting services, a national railway owner said its own fiber network made him unnecessary — and six months, a year, two years later they had built nothing

Scope: the 'we don't need you' reaction is a short-term phenomenon

66:32 20VC: Investing Lessons from FC Seeding Uber, Airtable and Coupang | Why Pro Rata is the Original Sin in VC | Why Liquidity Has Died in 2024 | Why LPs are Pissed with VCs | The Hard Truth About Seed Fund Economics with David Frankel @ Founder Collective

Matan Grinberg · Jun 13, 2026

Kirkland's attempt to build in-house is good for Harvey, because trying it yourself is what makes you realize how difficult it is and that you should buy from an expert instead

Nothing like trying to do something yourself to reveal the difficulty and that in-house capability doesn't matter for you

Scope: 'that is my sense'

10:11 20VC: Who Wins the Model War: OpenAI, Anthropic or Open-Source | Token Maxing, AI Hangovers & The Coming ROI Reckoning | Labour Displacement Fears are BS & Overblown | From Physicist to Sequoia Founder with Matan Grinberg, Founder @ Factory

Avishai Abrahami · Jul 13, 2026

The fear that developers would replace Atlassian by building their own tools has not materialized.

He held that view himself when buying Atlassian stock and was proven wrong; laziness in not selling saved him.

Scope: offered as an admission of his own bad judgment as a non-professional trader

19:55 20VC: Wix's Founder on What Wall St Gets Wrong About AI and Wix | Will Base44 Win the Vibe Coding Wars | The Truth About the Economics of Vibe-Coding | The Buyback Disaster: Lessons Learned with Avishai Abrahami

Harry Stebbings · Aug 24, 2026

'Customers will build it in-house once they get big' is a bogus reason to pass on a company, because they don't actually build it themselves

The classic Stripe or search API objection has repeatedly proven wrong — large customers just don't build their own

69:34 20VC: Inside Sequoia's Investment Committee: Lessons from Don Valentine, Doug Leone and Alfred Lin | How the SpaceX and Citadel Deals Went Down | What Sequoia Specifically Looks for in Founders with Julien Bek

Harry Stebbings · Oct 14, 2024

The market drastically overestimates the technical sophistication of ordinary companies to build their own vertical solutions with AI

These companies struggle even to onboard Slack

66:20 20VC: Investing Lessons from FC Seeding Uber, Airtable and Coupang | Why Pro Rata is the Original Sin in VC | Why Liquidity Has Died in 2024 | Why LPs are Pissed with VCs | The Hard Truth About Seed Fund Economics with David Frankel @ Founder Collective

Harry Stebbings · Mar 10, 2025

The idea that most enterprises will build and maintain their own verticalized AI tools is absurd.

Most enterprises, especially in Europe, don't even know what Slack or Notion are; no company can maintain the ~172 tools it runs; and organisations need an external party to blame, which is why consultants exist.

49:20 20VC: Lessons from Investing $2BN and Returning $8BN in Cash | Why Most Venture Partnerships are Broken | We Sold Salesforce Early and Lost Out on Billions | Are The Best Deals Always Expensive and Competitive with Jake Saper @ Emergence Capital

Harry Stebbings · Jul 27, 2025

The claim that companies will stop buying SaaS and instead build their own internal versions of everything is ridiculous

You would end up having to maintain and continuously improve 170 internal products

19:33 20Product: Should We Kill the PM Role Entirely | How Does Product Design Change Most in a World of AI | How Do Monzo Build Product Today: What Works, What Does Not | Why Most Product Sucks and What Makes Truly Great Product

Harry Stebbings · Dec 1, 2025

SaaS still has life and will not be replaced by companies building their own software or by foundation models

The average company uses 80-100 SaaS products, which they could neither build nor maintain; most businesses (plumbers, law firms, restaurants) can barely use Wix or Squarespace; and foundation model companies won't pursue narrow verticals like patent creation

Scope: admits a biased perspective given his job; the more vertical the software, the more defensibility

49:35 20VC: Scale, Surge, Turing, Mercor: Who Wins & Who Loses in Data Labelling | Is Revenue in Data Labelling Real or GMV? | Why 99% of Knowledge Work Will Go and What Happens Then? | Why SaaS is Dead in a World of AI with Jonathan Siddharth @ Turing

Ryan Petersen · Jun 20, 2026

Engineering talent should be pointed at replacing expensive labor and improving the core product, not at rebuilding SaaS vendors' software

Replacing SaaS at scale would consume too much engineering capacity for too little value; the leverage is in the core business

Scope: some of the threat to SaaS vendors may be a bluff; he could not actually replace all of them

47:24 20VC: Why Remote Work is White Collar Fraud | Why Revenge and Patriotism are the Best Founder Traits | Two Questions Every Founder Needs to Ask | The Wild Story of Raising $1BN from Masa Son in an Hour Long Meeting with Ryan Peterson, Founder @ Flexport

Harry Stebbings · Aug 14, 2024 · hedged

Building your entire internal software stack — compliance, HR, payroll, project management — is hard to justify given the ongoing burden of maintaining, updating and now adding AI to all of it.

Once you build it you own maintenance and modernization forever across the whole stack.

Scope: framed as not yet understanding the argument rather than flat rejection

10:26 20VC: Lessons from Mark Zuckerberg, Keith Rabois & Tobi Lütke | Why Remote is a Bad Idea for 90% of Companies | The Framework for How Shopify Builds Product Today | What Humans Get Wrong About Marriage and Kids with Kaz Nejatian, COO @ Shopify

Bret Taylor · Oct 2, 2024

Almost every company that chose to build its own software in an area where a SaaS solution was available has regretted it

Software must be perpetually maintained rather than written once, so in-house builds carry ongoing cost

Scope: applies to areas where a SaaS alternative exists; "almost every" rather than all

0:00 20VC: Bret Taylor: The AI Bubble and What Happens Now | How the Cost of Chips and Models Will Change in AI | Will Companies Build Their Own Software | Why Pre-Training is for Morons | Leaderships Lessons from Mark Zuckerberg

Harry Stebbings · Jan 15, 2025

The 'just build your own software' thesis ignores the burden of continuously upgrading and improving the resulting tools — no company wants to maintain 150 different internal tools.

Self-built internal tooling implies an ongoing maintenance and improvement obligation at scale.

24:06 20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia

Varun Mohan · Jun 2, 2025

Because bespoke software is now cheap to build, non-technical employees can build the internal tools that companies previously bought as expensive SaaS, displacing that spend.

Those tools were bought because they were bespoke single-function tools no company wanted to build itself; ephemeral, simple software is now cheap enough that non-developers build it — one non-developer internally replaced over $500K of sales tool spend.

14:58 20VC: Windsurf Founder on Will Model Companies Own the App Layer | Why Moats Do Not Exist in a World of AI | Why the Notion of Single Person $BN Companies is BS | Lovable vs Bolt & Cursor vs Windsurf: How Does it All End with Varun Mohan

Maor Shlomo · Nov 24, 2025

Software will become more liquid: buyers will start from templates or open-source projects and vibe code their own features rather than accept one-size-fits-all products

It makes more sense for the buyer — the code and data are theirs, no vendor lock-in, no feature bloat, and they can adapt it to their needs

Scope: not everyone will move there; enterprises will take a lot of time; not necessarily via a single prompt producing a full-featured product

9:00 20VC: Base44's Maor Shlomo on How Vibe Coding Will Kill SaaS and Salesforce | Why it is BS that Vibe Coding Platforms Do Not Have Defensibility and Bad Margins | Why He Worries About Google, Not Replit and Lovable | Why Long Anthropic, Not OpenAI?

Sebastian Siemiatkowski · Feb 16, 2026

The cost of creating software is going to zero, meaning anyone will be able to generate software at any time

He saw this coming one to two years ago when AI coding capability became clear

4:48 20VC: SaaS is Dead: Why Systems of Record Will Die in an Agentic World | What Revenue Multiple Will Software Companies Trade At? | From 7,000 to 3,000: We Need Less People Than Ever with Sebastian Siemiatkowski

Harry Stebbings · Aug 14, 2024

Shopify builds its own internal tooling not out of a refusal to partner but because no external vendor shares Shopify's opinions about how companies should be built

The tooling needs to be as good and as opinionated as Shopify wants, and nobody outside holds those opinions

37:55 20VC: Lessons from Mark Zuckerberg, Keith Rabois & Tobi Lütke | Why Remote is a Bad Idea for 90% of Companies | The Framework for How Shopify Builds Product Today | What Humans Get Wrong About Marriage and Kids with Kaz Nejatian, COO @ Shopify

Harry Stebbings · Aug 14, 2024

Shopify partners exceptionally well, but can only partner best with companies that hold similar opinions about what good software looks like

The constraint is not an inability to partner but Shopify's strong opinions about good software; it has over 700 payment providers integrated, more than any other software platform

38:16 20VC: Lessons from Mark Zuckerberg, Keith Rabois & Tobi Lütke | Why Remote is a Bad Idea for 90% of Companies | The Framework for How Shopify Builds Product Today | What Humans Get Wrong About Marriage and Kids with Kaz Nejatian, COO @ Shopify

Kieran Flanagan · Jul 11, 2025 · hedged

Companies are better off building and customizing some AI go-to-market capabilities themselves rather than buying generic off-the-shelf software

Customization to company-specific needs materially improves results relative to prompts built to work across all companies

Scope: applies to use cases like prospecting outreach and personalization

9:26 20Growth: The Death of Growth Teams? | How Hubspot Use AI to Triple Email Conversion | The Future of AI SEO | Why Prompt Engineering is the New Coding | What Every CMO Needs to Know About AI in 2025

Harry Stebbings · Jun 1, 2026 · hedged

If agents can rebuild software internally, SaaS is dead for use cases like bespoke customizations and integrations — large companies will just build their own

A company needing very specific integrations, e.g. a real estate firm connecting to pricing providers, could have agents build it rather than buy

Scope: conditional on agents being able to build software end to end

37:40 20VC: Mercor CEO on Why Application Layer Companies Have No Defensibility, The Model is the Product | Token Spend Will Exceed Headcount Spend in 5 Years | The True Cost of Hiring AI Researchers in the Valley Today with Brendan Foody

Nikesh Arora · Jun 22, 2026

Enterprises should only build AI applications where they have unique proprietary knowledge, and wait for generic functions to arrive as off-the-shelf AI applications

Generic problems everybody needs to solve will be delivered as customizable AI applications in 12-24 months, just as they were with SaaS; building them yourself wastes effort

19:38 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

Nikesh Arora · Jun 22, 2026

Enterprises should not build their own proprietary AI marketing, HR or ERP stacks; they should wait for vendors to build those horizontally and only build the AI systems that are particular to them, where they own the information, context and memory.

Things that should be available to everyone are built more effectively and efficiently by a vendor at world scale — likely the next iteration of Salesforce, SAP or Workday.

37:21 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

Sebastian Siemiatkowski · Feb 16, 2026

Small businesses will not vibe code their own software; they will buy broad off-the-shelf AI products that act as a 'company in a box'

He built a test project stacking a Claude agent on top of open-source accounting and CRM software and it worked well, showing a small firm can get bookkeeping and account setup done conversationally without an accountant

Scope: explicitly about plumbing firms, electricians and other small companies

9:17 20VC: SaaS is Dead: Why Systems of Record Will Die in an Agentic World | What Revenue Multiple Will Software Companies Trade At? | From 7,000 to 3,000: We Need Less People Than Ever with Sebastian Siemiatkowski

Sebastian Siemiatkowski · Feb 16, 2026 · speculative

Small companies like plumbing firms will likely buy an off-the-shelf 'company in a box' from Anthropic or someone building on Claude rather than build their own, but who provides it is still unknown

Scope: applies to small companies, not large enterprises; explicitly says the answer is unknown

12:57 20VC: SaaS is Dead: Why Systems of Record Will Die in an Agentic World | What Revenue Multiple Will Software Companies Trade At? | From 7,000 to 3,000: We Need Less People Than Ever with Sebastian Siemiatkowski

Kaz Nejatian · Aug 14, 2024

If your business is building software that other people build with, you must build your own internal tools, because tools shape their users and there is no such thing as opinion-free software.

First we build our tools, then our tools build us; the way Google Sheets is built changes how you write and think, so you need to know where your opinions come from. Tesla builds its own factory robots rather than buying off-the-shelf ones for the same reason.

10:54 Companies whose business is software must build their own tools because tools shape users

Kaz Nejatian · Aug 14, 2024

The fundamental problem with business software is that it is under-premised because it is under-dated: the data you need to run the business doesn't exist, so companies duct-tape tools together and end up making every decision in a spreadsheet.

Missing data forces CSV exports into Excel, so decisions get made there regardless of company size.

13:10 Off the shelf software lacks necessary data forcing reliance on spreadsheets

Kaz Nejatian · Aug 14, 2024

Outsourcing any core capability to an external partner cedes future decision-making, and over a hundred-year horizon it is very likely that partner's view of the world will diverge from yours

An external party's perspectives and opinions are an unknown you cannot control

38:09 Outsourcing core capabilities cedes long term control as partner views diverge

Maor Shlomo · Nov 24, 2025

Small businesses will want to build leaner, more customized software rather than own their code and data per se

Existing drag-and-drop, configurable SaaS tools are hellish to set up even for a software person, while an LLM can code something leaner that exactly fits a small business's process

11:18 Smbs build for simplicity and fit not ownership

Alex Bouaziz · Oct 22, 2025

Building your own internal infrastructure is the dumbest thing to do when you're small but the smartest thing to do as you scale, and Deel should have made that transition earlier than it did

Once you know where you're going, you can start making longer-term investments and bets; AI-routed internal ticketing has cut resolution time ~90%

35:50 Build own infrastructure only makes sense once scaled not when small

Jack Zhang · May 27, 2025

Pony Ma rejected the investment on the grounds that Tencent should build global WeChat Pay settlement internally rather than rely on a ten-person startup, since Tencent has hundreds of thousands of engineers

A company with that engineering scale should not need a tiny startup to do a better job

42:32 Large incumbents prefer building critical infra internally over relying on startups

Harry Stebbings · Jun 19, 2024

AI kills vertical SaaS, because it makes building and customizing equivalent vertical software for your own organization far cheaper

Cost of creating customized vertical software for your organization becomes much easier with AI

18:15 Ai collapses build cost making vertical saas obsolete

Michael Eisenberg · Jun 19, 2024 · hedged

Amazon has told its teams to stop buying software because AI can handle their business process needs

He has now heard it from three separate people, though it may be an echo chamber

18:35 Large enterprises already halting saas purchases because ai can replace them

Harry Stebbings · Feb 16, 2026 · hedged

Agents make data migration and integration between third-party tools so easy that owning and building every part of your stack in-house will look like a mistake in hindsight

If agents can move data between products, the integration pain that justified the compound/build-it-yourself approach disappears

12:22 Agent driven integration removes the case for building in house

Jonathan Siddharth · Dec 1, 2025 · speculative

Companies founded today use fewer SaaS apps and have fewer people than pre-ChatGPT cohorts at the same stage

50:57 New company cohorts already buy fewer saas apps

Amjad Masad · Apr 25, 2026

The decline in public SaaS valuations is justified: warehouse-native building maims growth for core SaaS companies, vertical SaaS is in trouble, and micro-entrepreneurs undercut point solutions on price

Survey SaaS is already being replaced wholesale by Replit builds, and solo builders on Replit are creating additional pricing pressure on vertical point solutions

20:55 Saas derating is justified warehouse builds and solo builders undercut

Avishai Abrahami · Jul 13, 2026 · hedged

The future of SMB website building will be a mix: many small business owners will stay on managed platforms like Wix because they don't want to invest effort outside their core business, while others will vibe-code custom tools on Base44.

A pizza-place owner's life is making pizza, not building software, so he'll stay with a done-for-him platform; but others will find vibe coding easy enough that custom-made beats templated, so both options need to exist.

13:04 Market splits between done for you platforms and custom builders

Ryan Petersen · Jun 20, 2026

The bigger near-term effect of AI on SaaS is renegotiation leverage rather than wholesale replacement — companies can use a documented in-house replacement as proof to force vendors to cut rates

Flexport has fully replaced one or two pieces of software and is having procurement build a case study to show other vendors what happens if they don't reduce contract rates

46:34 In house builds are renegotiation leverage not replacement

Julien Bek · Aug 24, 2026

The 'they'll build it themselves' objection is especially wrong for infrastructure companies

69:48 Infrastructure is least likely to be rebuilt in house

Jesse Zhang · Sep 19, 2025

Very few enterprises will build their own AI customer support system in house; the common outcome is partnering, sometimes in a hybrid where the customer owns select pieces

Teams experiment with GPT and quickly find production requires a whole framework for non-technical users plus alerting, monitoring, observability and A/B testing — which takes too long to build and isn't what their engineers should be doing

29:49 Production grade ai support requires a full framework that most enterprises wont build in house