TL;DR
The „AI will kill SaaS“ narrative is overblown. Yes, AI-powered development tools are making it easier to build custom software, and yes, most SaaS features go unused. But the real story is more interesting than simple disruption. Maintenance costs, security nightmares, compliance headaches, and network effects create powerful moats that protect the right kind of SaaS. What we’re seeing isn’t the death of an industry, it’s a transformation that will reward companies built around community, compliance, and compounding data value instead of just feature checklists.
Introduction
Every few years, someone predicts the death of SaaS. First it was open source. Then it was containerization. Now it’s AI-powered personal software creation.
The pitch sounds compelling: if AI can help anyone build exactly what they need, why pay for bloated enterprise software where you use 12 features out of 500? Why subscribe to Notion when you could spin up your own custom task manager in an afternoon?
Here’s the thing. I’ve watched enough „X will kill Y“ predictions fizzle out to know that the truth is usually messier and more interesting than the headline. So let’s dig into what’s actually happening with AI, SaaS, and the future of who builds what.
The feature bloat problem is real and it hurts
Let’s start with the uncomfortable truth: most SaaS is genuinely wasteful.
Pendo analyzed 615 software products and found that 80% of features are rarely or never used [1]. Just 12% of features generate 80% of daily usage. That’s not even a normal Pareto distribution, that’s extreme concentration. Across publicly-traded cloud companies, this represents an estimated $29.5 billion in wasted development investment [2].
But wait, it gets worse. Cloudnuro’s 2025 research found that 51% of enterprise SaaS licenses go completely unused [3]. Not underutilized. Unused. Organizations are wasting an average of $18 million annually on subscriptions that nobody even logs into.
This creates a theoretically exploitable gap. If you can build precisely what you need with AI assistance, you avoid paying for bloated feature sets you’ll never touch. The appeal is intuitive. A personal task manager tailored to your exact workflow instead of Notion’s 500+ features of which you use 12.
Consumer SaaS in „nice-to-have“ categories faces the most exposure here. Design tools, productivity apps, note-taking. These categories have low switching costs and minimal consequences if something breaks. If your custom-built note app loses data, you’re annoyed. If your custom-built HIPAA-compliant medical records system loses data, you’re sued.
But here’s what the „just build it yourself“ framing misses entirely: software products aren’t static artifacts you build once and forget. They’re living systems that need continuous feeding. Security updates. Dependency management. Integration changes when APIs evolve. Debugging when requirements shift.
The feature utilization data tells you nothing about total cost of ownership. And that’s where things get interesting.
AI coding gains are real but wildly inconsistent
The empirical evidence on AI coding assistants reveals something strange: results are all over the map.
GitHub and Microsoft ran controlled experiments showing developers completing tasks 55.8% faster with Copilot [4]. MIT, Harvard, and Microsoft field studies found 26% more tasks completed [5]. Sounds great, right?
Then METR (July 2025) dropped a critical contrarian study. Experienced developers working in their own codebases were actually 19% slower with AI assistance [6].
The perception gap is striking. Developers expected 24% speed improvement before tasks and believed AI helped them 20% faster afterward [7]. Despite objective measurement showing 19% slowdown.
This suggests AI coding productivity might be more about reducing cognitive load than actual speed gains. Valuable, sure. But not the dramatic democratization people claim.
Code quality presents additional concerns. CodeRabbit analyzed 470 open-source pull requests and found AI-generated code contained 1.7x more issues than human-only code [8]. Security vulnerabilities were particularly elevated: 1.88x more improper password handling, 2.74x more XSS vulnerabilities. Uplevel’s study of 800 developers found 41% higher bug rates in the Copilot group with no significant improvement in cycle time or throughput [9].
The pattern emerging: AI coding assistants provide substantial value for boilerplate generation, documentation, and learning new languages. But they may not accelerate complex, domain-specific development where context and architectural judgment matter most.
You know, exactly the kind of personalized software the hypothesis envisions individuals building.
Low-code platforms show us the limits
Low-code and no-code platforms are basically a preview of what democratized development looks like. The market has grown substantially to $13-14 billion in 2023 [10], with Gartner predicting 70% of new applications will use low-code technologies by 2025-26 [11]. Citizen developers are projected to outnumber professional developers 4:1 by 2025-26 [12].
So it’s working, right? Everyone’s a builder now?
Not quite. CIMI Corporation found that 54% of citizen development projects were considered failures after the first year [13]. Only 20% were deemed clear successes.
But here’s the twist: when organizations implemented proper governance through IT-established frameworks, success rates climbed to 81% in enterprise settings [14].
The lesson for AI-powered personal software is crystal clear: democratized development succeeds when it operates within guardrails provided by professional infrastructure. Not as a replacement for it.
The organizations achieving low-code success are embedding it within corporate IT frameworks, not replacing enterprise systems with individual creations. They’re augmenting professional teams, not replacing them.
Maintenance costs compound faster than you think
This is the most underappreciated counterforce to personal software creation: ongoing maintenance.
McKinsey research indicates that technical debt accounts for 20-40% of an organization’s entire IT estate value [15]. Companies pay 10-20% additional costs on top of any project to address it. The Consortium for Information & Software Quality estimates global technical debt costs at $1.52 trillion annually [16].
For software maintenance specifically, industry benchmarks suggest 15-25% of original development cost annually [17]. Total maintenance consumes 50-90% of software lifecycle costs according to academic research [18].
Let’s make this concrete. Say you build a medium-complexity application in a week with AI assistance. Cool. Now you need continuous investment that compounds quickly: security patches, dependency updates, API changes from integrated services, debugging as requirements evolve.
Gartner’s maintenance cost trajectory shows costs increasing over time: 10-25% in years 1-2, rising to 20-40% by year 6+ [19]. A mid-sized bank case study showed a $5 million system accruing $2.1 million in annual maintenance within three years. Over 40% of original development cost [20].
Personal software faces the same dynamics but without professional engineering support to manage it efficiently. You’re on your own when that critical dependency breaks at 2am. You’re on your own when a security vulnerability gets disclosed in a library you used. You’re on your own when the third-party API you integrated against changes its authentication flow.
Shadow IT security research underscores the risk. Gartner projects that 75% of employees will use shadow IT by 2027 [21]. Nearly 1 in 5 organizations has suffered cyberattacks directly linked to shadow IT, with $4.2 million average remediation costs [22].
Self-built software that handles meaningful data creates personal security liability without professional security teams monitoring for vulnerabilities. That’s not a theoretical concern. That’s a „you get personally sued“ concern.
Network effects create moats that personal software can’t cross
Industry research indicates network effects account for 70% of all value created by technology companies since the early 1990s [23].
Platform SaaS products with strong network effects create switching costs that personal software fundamentally cannot replicate. Slack’s communication patterns mean your whole team needs to move. Figma’s collaborative design means everyone in your workflow needs access. Salesforce’s marketplace of integrations means you’d need to rebuild hundreds of connections.
McKinsey found that B2B companies with strong lock-in strategies achieve 13% higher revenue growth compared to industry peers [24]. HubSpot customers using 3+ product hubs experience 35% better ROI than single-product users [25]. A typical enterprise SaaS platform offers 200+ pre-built integrations that would require substantial custom development to replicate.
The defensive moat is particularly strong in categories with collaboration features where value increases with network participants, marketplace dynamics connecting buyers and sellers, data accumulation where historical data becomes more valuable over time, and ecosystem lock-in where workflows build around platform capabilities.
Consumer SaaS without these dynamics faces disruption. Standalone productivity tools, simple utilities. But consumer SaaS with strong network effects will prove resilient.
I can build my own note-taking app. I can’t build my own Figma where my entire design team already collaborates daily with dozens of integrated plugins and shared component libraries.
European regulations are a compliance nightmare
The European regulatory environment creates substantial barriers to personal software adoption that Americans often underestimate.
GDPR compliance alone costs an average of €1.3 million annually for mid-sized firms [26]. Over €1.6 billion in fines were issued in 2024. The EU AI Act, effective August 2024, adds additional requirements for any software incorporating AI capabilities, including conformity assessments and technical documentation requirements.
For personal software, you bear complete GDPR accountability as sole data controller. Full responsibility for security measures, breach notification within 72 hours [27], and data subject rights processing. Enterprise SaaS vendors absorb much of this compliance burden, spreading costs across their customer base and offering pre-built GDPR features.
Here’s a comparison:
GDPR liability: Self-built means full responsibility. Enterprise SaaS means shared via processor agreement.
Security certifications: Self-built means you must build internally. Enterprise SaaS provides SOC 2, ISO 27001.
Breach notification: Self-built means direct 72-hour obligation. Enterprise SaaS means processor notifies controller.
AI Act conformity: Self-built means full assessment required. Enterprise SaaS vendor may absorb.
The CLOUD Act creates additional complexity. US-headquartered providers may be compelled to provide data regardless of where it’s stored, creating conflicts with GDPR Article 48 [28]. This drives demand for European SaaS providers and sovereignty-compliant solutions.
Categories where personal software actually has advantages for highly sensitive use cases still require substantial compliance expertise to execute properly. You can’t just „build it yourself“ when regulators can fine you 4% of global revenue.
Market predictions point to transformation, not death
Analyst projections show the SaaS market continuing strong growth: from approximately $218-300 billion in 2024 to projections of $887 billion to $1 trillion+ by 2030. IDC’s „Is SaaS Dead?“ analysis acknowledges traditional SaaS may face short-term negative impact from AI but predicts recovery by 2030 [29].
More nuanced predictions suggest structural transformation in the near term (2-5 years): pricing model evolution with a shift from seat-based to consumption and outcome-based pricing, with 70% of software vendors expected to refactor pricing strategies by 2028 [30]. AI feature integration where 95% of organizations expected to adopt AI-powered SaaS by 2025. Consolidation where the average company is using 106 apps, down from 130 peak in 2022 [31]. And vertical SaaS commanding premium valuations at 8.6x revenue multiples versus 6.7x for horizontal solutions.
Long-term (10+ years) we’ll see commoditization of simple tools where horizontal productivity SaaS faces margin compression as AI alternatives mature. Micro-SaaS proliferation where more individuals sell personalized solutions, but most remain small (under $1M ARR). Platform consolidation where network-effect-protected platforms absorb point solutions. And compliance-as-moat where regulatory complexity creates barriers to entry that favor scaled vendors.
The micro-SaaS explosion will create fragmentation, not disruption. Individuals will increasingly build and sell personalized solutions, but structural economics constrain how large these can grow. The 80% unused features finding cuts both ways: highly specialized tools can capture underserved niches, but total addressable markets remain small.
What gets disrupted and what stays protected
The segments most vulnerable include simple productivity tools without network effects like note-taking, to-do lists, and personal CRM. Consumer utilities with low switching costs. „Nice-to-have“ subscriptions where failure consequences are minimal.
The segments most protected include collaboration platforms with network effects like Slack, Figma, and Notion for teams. Compliance-heavy enterprise SaaS like HR systems and financial software. Data-moat platforms where accumulated data creates value. Marketplace SaaS connecting multiple parties. Vertical SaaS with deep industry specialization.
Evidence from the vertical SaaS premium (8.6x vs 6.7x multiples) suggests markets are already rewarding specialization. AI-enabled development will accelerate this by reducing the minimum viable team required to build software, enabling individuals to capture niches that previously required larger organizations.
However, most will remain lifestyle businesses rather than venture-scale outcomes. Which is fine! Not everything needs to be a unicorn. But it does mean we’re not seeing the „death“ of SaaS. We’re seeing its fragmentation into protected platforms and niche micro-products.
Competitive edges for the survivors
The research points to several defensible competitive positions for SaaS companies navigating this transition.
Network effects and ecosystem lock-in remain the strongest moat. Products where value increases with users, where integration ecosystems create switching costs, and where collaborative features are core will face the least disruption from individual alternatives.
Compliance and trust certification becomes more valuable as regulatory complexity increases. Enterprise buyers in regulated industries cannot rely on self-built tools. Gartner surveys show 48% cite security as their top purchasing criterion [32]. SOC 2 certification, HIPAA compliance, and AI Act conformity create moats that individual builders cannot easily cross [33].
Outcome-based pricing alignment allows SaaS companies to capture value proportional to customer benefit rather than competing on per-seat economics against free alternatives. Usage-based and success-based models may prove more resilient than traditional subscription structures.
Vertical depth over horizontal breadth enables charging premiums for specialized solutions that understand industry-specific workflows. The 8.6x vs 6.7x valuation multiple for vertical SaaS suggests markets already recognize this value.
AI-native architecture that compounds with usage data creates new network effects. OpenAI’s function-calling ecosystem and similar „AI cloud“ platforms may create the next generation of platform lock-in [34].
Conclusion
The hypothesis that AI-powered personal software will cause the „death“ of consumer SaaS oversimplifies a more complex transformation.
Yes, the 80% unused feature finding creates real vulnerability. Yes, AI development tools are democratizing software creation in ways that will compress margins and fragment markets for commoditized horizontal tools.
But the countervailing forces are substantial. Maintenance burden (15-25% annually, compounding) makes software ownership expensive over time. Security risks from unmanaged personal software create liability that enterprises won’t accept. Compliance requirements, particularly in the EU, favor certified vendors over individual builders. Network effects protect platform businesses in ways personal software cannot replicate.
The most likely future is bifurcated: commoditized productivity tools face margin compression and micro-SaaS competition, while network-effect-protected, compliance-heavy, and vertical-specialized SaaS continues growth with defensible positions.
Multi-billion-dollar horizontal SaaS platforms won’t „die“ but will need to demonstrate value beyond feature sets that individuals can replicate. The SaaS companies that thrive will be those building moats around community, compliance, and compound value from user data, not just software functionality.
So no, AI won’t kill SaaS. But it will force the industry to justify its existence based on what individuals actually can’t build themselves. And that’s probably a good thing.
References
[1] Pendo (2019). „The 2019 Feature Adoption Report.“ https://www.pendo.io/resources/the-2019-feature-adoption-report/
[2] Pendo. „Feature Adoption Benchmarking.“ https://www.pendo.io/pendo-blog/feature-adoption-benchmarking/
[3] Cloudnuro (2025). „50+ Essential SaaS Statistics and Industry Trends for 2026.“ https://www.cloudnuro.ai/blog/saas-statistics
[4] GitHub/Microsoft. „Research: quantifying GitHub Copilot’s impact on developer productivity and happiness.“ https://github.blog/news-insights/research/research-quantifying-github-copilots-impact-on-developer-productivity-and-happiness/
[5] MIT/Harvard/Microsoft (2023). „The Effects of Generative AI on High-Skilled Work.“ https://economics.mit.edu/sites/default/files/inline-files/draft_copilot_experiments.pdf
[6] METR (July 2025). „Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity.“ https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/
[7] METR (July 2025). „Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity.“ https://arxiv.org/abs/2507.09089
[8] CodeRabbit (2025). „State of AI vs Human Code Generation Report.“ https://www.coderabbit.ai/blog/state-of-ai-vs-human-code-generation-report
[9] Uplevel (2024). „Gen AI for Coding Research Report.“ https://resources.uplevelteam.com/gen-ai-for-coding
[10] Gartner/Kissflow. „Gartner Forecasts Low Code/No Code Platform Market for 2026.“ https://kissflow.com/low-code/gartner-forecasts-on-low-code-development-market/
[11] Joget. „Low-Code Growth: Key Statistics That Show Its Impact.“ https://joget.com/low-code-growth-key-statistics-that-show-its-impact/
[12] Kissflow. „Citizen Developers will Outnumber Professional Coders by 2025.“ https://kissflow.com/citizen-development/future-of-citizen-developers/
[13] TechTarget. „Why projects fail in the citizen developer model.“ https://www.techtarget.com/searchsoftwarequality/tip/Why-projects-fail-in-the-citizen-developer-model
[14] Hostinger (2025). „26 low-code trends for 2025: Key statistics and insights.“ https://www.hostinger.com/tutorials/low-code-trends
[15] McKinsey (2020). „Tech debt: Reclaiming tech equity.“ https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/tech-debt-reclaiming-tech-equity
[16] CISQ (2022). „The Cost of Poor Software Quality in the US: A 2022 Report.“ https://www.it-cisq.org/the-cost-of-poor-quality-software-in-the-us-a-2022-report/
[17] Appinventiv. „How to Accurately Estimate and Minimize Software Maintenance costs.“ https://appinventiv.com/blog/software-maintenance-cost/
[18] PMC. „Which Factors Affect Software Projects Maintenance Cost More?“ https://pmc.ncbi.nlm.nih.gov/articles/PMC3610582/
[19] Idea Link. „Software Development vs Maintenance: The True Cost Equation.“ https://idealink.tech/blog/software-development-maintenance-true-cost-equation
[20] Idea Link. „Software Development vs Maintenance: The True Cost Equation.“ https://idealink.tech/blog/software-development-maintenance-true-cost-equation
[21] Gartner (2023). „Gartner Identifies Top Security and Risk Management Trends.“ https://www.valencesecurity.com/resources/blogs/gartner-saas-applications-outside-of-it
[22] Zluri (2025). „Shadow IT Statistics: Key Facts to Learn in 2025.“ https://www.zluri.com/blog/shadow-it-statistics-key-facts-to-learn-in-2024
[23] NFX. „70 Percent of Value in Tech is Driven by Network Effects.“ https://www.nfx.com/post/70-percent-value-network-effects
[24] McKinsey. „Net revenue retention strategies for B2B tech growth.“ https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/the-net-revenue-retention-advantage-driving-success-in-b2b-tech
[25] HubSpot. „ROI of HubSpot’s Customer Platform.“ https://www.hubspot.com/roi
[26] CookieYes. „Top 10 GDPR Compliance Cost and How to Manage Them.“ https://www.cookieyes.com/blog/gdpr-compliance-cost/
[27] GDPR Info. „Art. 33 GDPR – Notification of a personal data breach to the supervisory authority.“ https://gdpr-info.eu/art-33-gdpr/
[28] Wire. „CLOUD Act – What It Means for EU Data Sovereignty.“ https://wire.com/en/blog/cloud-act-eu-data-sovereignty
[29] IDC (2024). „Is SaaS Dead? Rethinking the Future of Software in the Age of AI.“ https://www.idc.com/resource-center/blog/is-saas-dead-rethinking-the-future-of-software-in-the-age-of-ai/
[30] CIO. „New software pricing metrics will force CIOs to change negotiating tactics.“ https://www.cio.com/article/4097012/new-software-pricing-metrics-will-force-cios-to-change-negotiating-tactics.html
[31] SellersCommerce (2025). „27 Eye-Opening SaaS Statistics In 2025.“ https://www.sellerscommerce.com/blog/saas-statistics/
[32] Gartner Digital Markets (2024). „Why Security Is a Priority for Software Buyers in 2024.“ https://www.gartner.com/en/digital-markets/insights/2024-buying-trends-software-security
[33] Vanta. „How much does GDPR compliance cost?“ https://www.vanta.com/collection/gdpr/gdpr-compliance-cost
[34] Multiples.vc. „Vertical SaaS Valuation Multiples and Public Comps.“ https://multiples.vc/vertical-saas-valuation-multiples