Issue No.1 - March 2026 - Open Question
Begin
→
A Narrative Moodboard - The AI Era - No Verdict
SaaS
is Dead
- or is it?
Five tiers. Six narratives. Six counters.
The fault lines. The geography. The adversarial questions.
One open question.
The fault lines. The geography. The adversarial questions.
One open question.
For thinking,
not deciding.
March 2026
not deciding.
March 2026
Page 02 - The Provocation
"The question has been asked as if SaaS is one thing. It is not."
Personal Software and Salesforce are both called SaaS. The death narrative is completely true for one of them, completely false for the other, and painfully contested for everything in between.
This issue maps the full territory - without picking a winner.
This issue maps the full territory - without picking a winner.
$34B
Salesforce FY2024 Revenue
Still Growing
Still Growing
Revenue, not ARR
~$200M
Cursor Est. ARR 2025
Fastest ramp in category
Fastest ramp in category
A wrapper that won
5–6×
ARR Multiples Today
vs 20× peak in 2021
vs 20× peak in 2021
Valuation, not revenue collapse
300M+
Llama Family Downloads
All versions, 2024
All versions, 2024
Open source signal
$10B
ServiceNow Subscription Rev.
Accelerating with AI
Accelerating with AI
Enterprise SaaS thriving
Section 00 - Before the Debate
"What exactly is SaaS?" - A spectrum, not a category
Five different conversations. One label.
Almost every "SaaS is dead" argument is actually about Tier 5. Almost every counter-argument is also about Tier 5. But the death is most real at Tier 1 and 3 - and nobody is talking about that.
Meanwhile, Tier 2 (Micro-SaaS) may be thriving precisely because of AI - lowering the cost of building to near-zero for a solo founder.
01
Personal
Tools built for yourself. AI workflows. Notion setups. Custom GPTs. This was always a category - AI exploded it into something new.
$0 - built, not bought
02
Micro-SaaS
Solo founders or 2-person teams. One problem, one niche. No investors, no sales team. Stripe + a landing page. Invisible to the VC narrative.
$1K – $50K MRR
03
Prosumer
Freelancers, creators, consultants, solopreneurs. Most exposed to AI replacement - and most likely to build their own alternatives.
$20 – $99/mo per user
04
SMB SaaS
Small business tools. Largest customer count segment. Being squeezed from above by enterprise tools dropping prices and below by AI tools doing 80% for free.
$500 – $5K/mo per co
05
Enterprise
Multi-year contracts, procurement cycles, compliance requirements, organizational memory. Where "SaaS is dead" is least true - and most loudly debated.
$50K – $10M+ ACV
Use ← → arrows or swipe to navigate · keyboard arrows supported
Section 00 of 07 - Full Territory Map
I
Section 01 - The Claim
Six ways people are arguing SaaS is dying
Each narrative is tagged to the tiers it actually applies to. Most of them don't apply to most founders.
Personal
Micro-SaaS
Prosumer
SMB
Enterprise
Narrative 01 / 06
SMB
Enterprise
The Replacement Thesis
"AI agents don't improve workflows. They dissolve them."
When an agent can draft, send, track, analyze, and report - why maintain five separate tools that each own one step? The unit of software shifts from the application to the outcome. The app layer collapses. What remains is infrastructure.
Applies most forcefully at Tier 4–5 where multi-tool stacks are the norm.
Applies most forcefully at Tier 4–5 where multi-tool stacks are the norm.
Counter + Reference
The Counter - Context Gap
Tasks need context. Context needs continuity. Enterprise workflows run on audit trails, approval chains, compliance frameworks built over years. An agent that executes without institutional context is a demo, not a deployment.
The Signal
"Devin ships code. But who reviews it? Who owns the ticket? Who updates the sprint? The agent replaces the task - not the system around it."
Observation - Enterprise AI deployment patterns, 2025
Narrative 02 / 06
Micro-SaaS
Prosumer
The Wrapper Collapse
"Everything built in 2023 on top of GPT-4 is already obsolete."
First-wave AI startups were thin UX layers on OpenAI APIs - no proprietary data, no workflow depth, no switching costs. As foundation models improve, the wrapper loses its margin.
The real threat to Micro-SaaS is not VC funding. It's that GPT-5 will do what you built - for free - in six months.
The real threat to Micro-SaaS is not VC funding. It's that GPT-5 will do what you built - for free - in six months.
Counter + Reference
The Counter - Depth Wins
Cursor is a wrapper. It reached ~$200M ARR faster than almost any dev tool in history. The dismissal conflates building material with building. The wrapper that earns workflow depth + compounding usage data is not the same thing as the wrapper that doesn't.
Narrative 03 / 06
Prosumer
SMB
Enterprise
The Pricing Crisis
"Per-seat pricing was a proxy for value. AI breaks the proxy."
SaaS pricing assumed humans were the unit of consumption. When AI handles the work of 10 seats, you can't sell 10 seats. Revenue per customer compresses structurally - not cyclically.
GitHub Copilot is the case study nobody wants to discuss aloud: more output, fewer seats needed.
GitHub Copilot is the case study nobody wants to discuss aloud: more output, fewer seats needed.
Counter + Reference
The Counter - CFO Legibility
Usage-based pricing was supposed to fix this. It created a different problem. Finance teams hated unpredictability. Investors punished variance. Per-seat was a feature - legible, forecastable, beloved by CFOs. The replacement model is not obvious yet.
The Signal
"GitHub Copilot didn't kill developer seats. It made each seat 55% more productive - and companies still bought the same number."
GitHub Copilot impact study - developer productivity, 2024
Narrative 04 / 06
Micro-SaaS
Prosumer
SMB
The Open Source Threat
"Llama and Mistral just broke the SaaS pricing moat forever."
Open-source AI means you can self-host intelligence. The moat isn't the model anymore. Hits Tiers 2–4 hardest - the technical prosumer or small team that can now build what they used to buy.
Enterprise (Tier 5) is largely unaffected - they buy compliance wrappers, not models. For the solo technical founder, the build/buy calculation just shifted permanently.
Enterprise (Tier 5) is largely unaffected - they buy compliance wrappers, not models. For the solo technical founder, the build/buy calculation just shifted permanently.
Counter + Reference
The Counter - Compliance Moat
Open source gives you the model. Not the compliance wrapper, the SLA, or the support contract. Self-hosting requires ML engineers, GPU infrastructure, security review, ongoing maintenance. For most enterprises the fully-loaded cost exceeds the SaaS subscription it replaces.
Narrative 05 / 06
Prosumer
SMB
Enterprise
The Vertical Exposure
"Vertical SaaS sold you a workflow with a UI. AI delivers the workflow without the UI."
Legal research tools, HR screening platforms, customer support software - these are workflows with a front-end. AI now delivers the outcome without the interface layer.
Harvey AI vs. LexisNexis is the live experiment. Deep domain knowledge survives. Surface-level workflow automation does not. Which one are you?
Harvey AI vs. LexisNexis is the live experiment. Deep domain knowledge survives. Surface-level workflow automation does not. Which one are you?
Counter + Reference
The Counter - Domain Depth
Deep vertical SaaS isn't a UI over a workflow. It's a decade of domain knowledge encoded in software - healthcare billing logic, financial compliance rules. AI can replace the interface; it cannot replicate encoded institutional knowledge built through thousands of customer conversations.
The Signal
"Harvey AI raised $100M to replace legal research tools. But LexisNexis still has 40 years of case law annotation that no model can replicate from scratch."
Harvey AI vs LexisNexis - vertical depth case study, 2025
Narrative 06 / 06 - The Underreported One
Personal
Micro-SaaS
Prosumer
The Personal OS Shift
"AI doesn't kill SaaS at the top. It bypasses it at the bottom - quietly."
At Tier 1–3, the shift isn't from one SaaS tool to another. It's from buying software to building it. Custom GPTs, Claude Projects, n8n flows, local LLMs. The professional who paid for 12 tools is now configuring one AI context that does 80% of what the stack did.
This is the quiet death. No press release. Just 8 cancelled subscriptions.
This is the quiet death. No press release. Just 8 cancelled subscriptions.
The Inversion
The Counter - Micro-SaaS Renaissance
AI is not killing Micro-SaaS. It's enabling solo founders to build what used to require a team of 10. Cursor + Claude = one person shipping, supporting, and iterating in days. The "SaaS is dead" story at Tier 1–2 is almost perfectly inverted.
Signal from the field
"I built and launched a SaaS in 4 hours with Cursor. It has 40 paying customers. Two years ago this would have taken me 4 months and a co-founder."
- COMPOSITE VOICE · INDIE HACKERS / SHIP30 · 2024–25
II
Section 02 - The Counter
Six reasons the death notice is premature
Each counter is strongest at specific tiers - and genuinely weak at others. Read where it applies before deciding if it applies to you.
The Fault Lines
Strongest vs. strongest - where the real tension lives, mapped by tier
Replacement - Tiers 4–5
SMB
Enterprise
AI agents will perform multi-step enterprise workflows end-to-end. The app layer is redundant when the agent can read intent and execute autonomously.
TRUST
Accountability Gap - Tiers 4–5
Counter
When an agent makes a $2M procurement error, who do you call? Until agents are legally accountable, humans stay in the loop. Humans in the loop need tools.
Personal OS - Tiers 1–3
Personal
Micro
Prosumer
The prosumer cancels 8 SaaS subscriptions and builds a personal AI stack. The $29/month tool market is collapsing from the bottom up, not the top down.
DEPTH
Brittleness - Tiers 1–3
Counter
Personal AI stacks are non-transferable and require constant maintenance. Most prosumers who build their own tools eventually return to SaaS - they're in the business of their actual work, not software maintenance.
Micro-SaaS Renaissance - Tiers 1–2
Personal
Micro
AI enables solo founders to ship focused SaaS products faster than ever. Micro-SaaS is thriving. The supply of software is expanding, not contracting.
MARKET
Supply Glut - Tiers 1–2
Counter
More products being built doesn't mean more being bought. If AI lowers build cost AND compresses demand, you get more products chasing fewer paying customers. Supply glut, not renaissance.
Pricing Compression - Tiers 3–5
Prosumer
SMB
When AI handles the work of 10 seats, SaaS revenue per customer declines structurally - not cyclically. The seat count model is permanently broken.
TAM
Long Tail Expansion - Tiers 3–4
Counter
Millions of SMBs that couldn't afford 10 seats can now access AI-powered capabilities at 1-seat pricing. TAM expansion at the bottom may outpace revenue compression at the top.
The Missing Moats
The defenses that don't show up in strategy decks - but decide the outcome on the ground
Ecosystem Lock-in
"The real moat isn't features - it's the 500 apps that assume you exist."
Apollo, Clay, LinkedIn, Outreach - they all have native HubSpot and Salesforce connectors. Go custom, and you're building every connector yourself. That's a full-time job nobody budgeted for. An ecosystem moat that compounds with every integration.
Signal: 107-comment LinkedIn thread on vibe-coding a CRM · March 2025
Hiring Velocity
"Sales candidates already know HubSpot. Onboarding them onto your custom tool costs weeks."
Every job description says "HubSpot experience required." Vibe-code a replacement and you're opting out of an entire talent pool. Ten reps × one week of onboarding = ten weeks of selling time gone. HubSpot has an academy. Your franken-CRM has Jeff's Loom videos.
Victor Ugochukwu · Senior Product Marketing · B2B SaaS GTM Strategy
Maintenance Is the Product
"Building is easy. Maintaining is the part nobody talks about until it's too late."
When a routing rule breaks or a field mapping drifts - it falls on an engineer. Your sales team can't fix it. The real product of SaaS was never the initial build. It was the 10,000 edge cases, the security patches, the uptime guarantees. You can't vibe-code a decade of operational maturity.
Saharsh Agrawal · CEO, Surface Labs · "We went back to HubSpot in two weeks"
Build vs. Buy Is Being Relitigated
"Devs could have built most software even in the past. Did we always build?"
AI didn't invent the build-vs-buy decision. It just made "build" cheaper. But cheaper to build doesn't mean cheaper to own. The question was never "can you build it?" It was always "should you maintain it?" AI changes the numerator but not the denominator.
Nikita Agarwal · Building Exosphere · Ex-Azure OpenAI · IIT Alum
Agent-Native Data Layer
"AI doesn't need opportunities and accounts and leads. Those are inherently understood by the LLM."
The CRM data model - opportunities, accounts, leads, contacts - was designed for humans navigating tables. Agents don't need that scaffolding. The real question is whether the relational data model itself is obsolete for agent workflows. If it is, the moat evaporates regardless.
Ashmer A. · Founder & CEO, Gravity · "Trying to build a CRM is an antiquated notion"
The Threat IS the Disruption
"Even if most teams end up back on HubSpot, the threat forces incumbents to move."
Every failed attempt to replace SaaS still exerts downward pricing pressure. SaaS companies are forced to drop prices - not because alternatives are better, but because the option to build exists. The disruption doesn't require replacement to work. The threat alone restructures the market.
Daniel Brady · AI GTM Operations Lead · Darren Edwards · AI Automation Specialist
Section 05 - The Emerging Consensus
The Third Way
Neither "SaaS is dead" nor "SaaS wins" captures what the smartest operators are actually doing. They're keeping the system of record and vibe-coding the workflow layer on top. The CRM stays. The custom intelligence layer grows around it.
Keep (System of Record)
CRM as single source of truth
Native integrations ecosystem
Compliance and audit trails
Hiring pool familiarity
Vendor-managed uptime
Security certifications
Native integrations ecosystem
Compliance and audit trails
Hiring pool familiarity
Vendor-managed uptime
Security certifications
hybrid
Build (Workflow Layer)
Custom lead scoring with AI
Agent-driven routing rules
Personalized outreach sequences
Internal dashboards & alerts
Workflow automations via MCP
Domain-specific intelligence
Agent-driven routing rules
Personalized outreach sequences
Internal dashboards & alerts
Workflow automations via MCP
Domain-specific intelligence
"Keep HubSpot for your system of record. Vibe-code the workflows around it. That's the third way that actually works."
- Shahed Islam, Co-Founder & CEO, SJ Innovation
- Shahed Islam, Co-Founder & CEO, SJ Innovation
Editor's Note - Narrative Hygiene
Everyone in this debate has skin in the game. Including us.
Every "SaaS is dead" post comes from someone building the replacement. Every "SaaS wins" post comes from someone whose revenue depends on it. The most viral post in this debate last week - a founder who vibe-coded a CRM and went back to HubSpot - was written by someone whose company sits on top of HubSpot. That doesn't make the points wrong. But it means every story in this space is also a positioning exercise.
When an incumbent says "SaaS wins"
"Their revenue depends on you believing this."
When a startup says "SaaS is dead"
"Their funding depends on you believing this."
When a VC says "the market is shifting"
"Their portfolio depends on you believing this."
When this magazine says "no verdict"
"Our identity depends on you believing this."
Read everything in this issue - and everything outside it - with this filter on.
The Geography of the Debate
Parallel realities - not counters to each other
🇺🇸
Silicon Valley / US Enterprise
"SaaS is being renegotiated, not killed."
Debate is about pricing models, agent integration, and whether incumbents can absorb AI faster than new entrants compound distribution. Revenue still grows; the story is about multiples and defensibility.
🇮🇳
India / South Asia
"SaaS never fully arrived. AI is the first software layer."
63M+ small businesses bypassed desktop SaaS entirely. Mobile-first, cost-sensitive, informal. AI-native tools on WhatsApp and voice interfaces are the first real software many of these businesses will ever use seriously.
🌍
Africa / Sub-Saharan Markets
"AI leapfrogs SaaS the way mobile leapfrogged landlines."
1.4B people. SaaS penetration a fraction of North America's. The M-Pesa parallel is real - AI tools built for informal markets and local languages are the dominant first software layer. Not disruption of SaaS; a parallel track.
🇨🇳
China / State-Adjacent Markets
"Sovereignty rewrites the stack. Western SaaS is a geopolitical risk."
Huawei Cloud, Alibaba Cloud replacing Salesforce and Workday - not because AI is better, but because sovereignty demands it. Data localization and state-backed platforms reshape the entire competitive landscape.
🇪🇺
Europe / GDPR Markets
"Compliance is making SaaS more valuable, not less."
EU AI Act, GDPR, DORA, MDR create an enormous compliance surface. SaaS vendors with established certifications benefit - self-hosting compliant AI infrastructure is prohibitive for most European enterprises.
🌏
Southeast Asia / APAC
"SaaS and AI-native coexist by segment. APAC is not one market."
Singapore and Tokyo operate on US enterprise timelines. Jakarta, Manila, Dhaka are closer to the India dynamic. "SaaS is dead" is simultaneously irrelevant and premature - depending which city you're standing in.
Section 06 - The Adversarial Frame
What a skeptic sees when they look at your business
These are not self-help questions. They are the questions a hostile investor - one who believes SaaS is dead - will ask you. If you can answer "yes" in under 10 seconds, the question is not hard enough.
01
Micro-SaaS
Prosumer
If your best customer's new hire started today and had never heard of your product - how long before they'd find an AI-native alternative that does 80% of what you do and stopped the evaluation there?
Weak answer sounds like:
"Our product has more features." - Features are not switching costs. They are reasons to switch.
02
Micro-SaaS
SMB
Name the specific thing your product knows about a customer after 12 months of use that couldn't be reconstructed in a week by an AI with access to their email and calendar.
Weak answer sounds like:
"We have their data." - Data without structural lock-in is just a migration project waiting to happen.
03
SMB
Enterprise
Is your product getting more valuable as more people use it - or is it the same product for user 1 and user 10,000? If the answer is "the same," you are selling a capability, not building a network. Capabilities are being commoditized.
Weak answer sounds like:
"We improve with feedback." - That's product management, not a network effect.
04
Micro-SaaS
Prosumer
SMB
If a technically proficient version of your ideal customer spent one weekend with Claude and Cursor - what percentage of your product's core value could they replicate? And what's left after that?
Weak answer sounds like:
"They're not technical enough." - Your customers are getting more technical every quarter. This is a countdown, not a defense.
No verdict. Just the territory.
"The question isn't whether SaaS is dead. It's which tier, which market, and which decade.
At Tier 1–2, AI is creating software, not killing it.
At Tier 3, AI is the quiet cancellation of 12 subscriptions.
At Tier 5, the death notice arrives in 2034, not 2024."
At Tier 1–2, AI is creating software, not killing it.
At Tier 3, AI is the quiet cancellation of 12 subscriptions.
At Tier 5, the death notice arrives in 2034, not 2024."
Thriving
Personal + Micro-SaaS
Tiers 1–2
Tiers 1–2
Disrupted
Prosumer Layer
Tier 3
Tier 3
Squeezed
SMB SaaS - from above
AND below simultaneously
Tier 4
AND below simultaneously
Tier 4
Renegotiated
Enterprise Platform
Death notice: 2034
Tier 5
Death notice: 2034
Tier 5
SaaS is Dead - A Narrative Moodboard · Issue No.1 · March 2025
Quotes marked Paraphrased reflect documented positions, not verbatim text.
For thinking, not deciding.
Quotes marked Paraphrased reflect documented positions, not verbatim text.
For thinking, not deciding.
"The map is not the territory - but the map is what we have."