So yeh. Let me walk you through something that’s been consuming my thinking lately—something I believe will fundamentally reshape how we build, deliver, and monetize software over the next decade.
The thesis: AI could disrupt software more profoundly than SaaS ever did.
This isn’t just about making software development faster or adding chat interfaces to existing products. It’s about a potential fundamental shift in what makes software valuable—and what might happen if the core assumption that "software is expensive to create" gets completely upended.
The Mental Models Behind This Analysis
I’m synthesizing several frameworks to understand how AI disruption will unfold:
-
Disruptive Innovation (Clayton Christensen) – New technologies start simple, serve overlooked needs, improve rapidly, then move upmarket to displace incumbents. AI follows this pattern exactly.
-
First Principles Thinking – Strip away assumptions to fundamental truths. The core assumption breaking here: "software is expensive to create." When that becomes false, everything downstream changes.
-
Value Chain Decomposition (Michael Porter) – Understanding where value is created and captured in software delivery. As creation costs fall, value migrates to quality, integration, and outcomes.
-
Jobs-to-be-Done (Clayton Christensen) – Users don’t want software—they want progress toward goals. When AI can deliver that progress more directly, intermediary platforms become friction.
-
Platform Economics (Parker, Van Alstyne, Choudary) – Understanding network effects, switching costs, and why some platforms remain resilient despite technological disruption.
The synthesis reveals something none address alone: when software creation costs approach zero, the entire value chain inverts—from selling software to orchestrating outcomes, from features to reliability, from platforms to protocols.
The Evolution of Software Value
The Historical Model
Let’s start with a framework. The WordPress and broader open-source ecosystem has historically been valuable because writing software manually is expensive—in resources, time, and risk. These costs made ready-to-use, open source software asymmetrically more attractive.
The SaaS Disruption
When SaaS and managed services came along, they disrupted the software market again by greatly reducing the resources, time, and risk involved in configuring, hosting, and managing that software.
You can visualize this evolution in three steps:
- Manual software development → Expensive custom builds
- Open source software → Shared development costs, but still deployment overhead
- SaaS platforms → Infrastructure and operations handled for you
The AI Inflection Point
But that’s not the end of the story. The next evolution could be AI-generated, bespoke, on-the-fly software.
Previous evolutions were built around the core tenet that writing software is expensive. This next evolution could break that fundamental assumption. And if that happens, everything changes.
The Disruption Sequence
Here’s what I believe could be coming, and the order it might happen:
1. Premium Plugins and Themes Suffer First
When it becomes cheap enough to simply generate a new solution instead of buying a premium plugin or theme, why would most people pay? The value proposition collapses for anything that doesn’t offer truly unique functionality or ongoing value.
2. CMS Overhead Gets Questioned
Content management systems exist because creating web content has been complex. But imagine copying a Google Doc into an AI and asking it to make it into a web page—complete with responsive design, SEO optimization, and performance tuning.
"No CMS" is already a trend with static site generators and Jamstack. What happens when the barrier to homebrew essentially disappears? When the CMS itself is just cognitive overhead for simple content publishing?
3. Agent Abstraction Layer
Site management agents might increasingly abstract away web services into a "virtual workforce."
Imagine a "community agent" that works to gather reviews, ratings, and comments, and manages spam. Whether it uses a traditional anti-spam service or an AI for spam protection likely won’t be a conscious human decision for most end users.
The agent makes optimization decisions based on cost, performance, and effectiveness—shifting vendors automatically if needed.
4. Serverless Becomes Default
Where people still need dynamic websites, they might increasingly see traditional hosting as an expensive way of doing it. Paying for a web server running 24/7 just in case you need server-side scripting or database access could become increasingly archaic in a serverless world.
It’s like leaving your car’s engine idling constantly just in case you need to drive somewhere. The cognitive overhead that comes with platforms that can do so much but where most users only use a small percentage becomes unjustifiable.
5. Atomic Content > Sites and Platforms
Historically, there’s been value to "websites" because they represented a local maximum in terms of end user experience and creator operational efficiency (training, management, etc).
AI promises a world in which systems can seamlessly reformat content from human-readable to JSON or vice versa, translate between languages, and even transcreate formats (text to video, audio to text, etc).
I believe that individual pieces of content (with appropriate schema) might already be the way people actually communicate online—it’s just that to-date, it’s been much more efficient to create better experiences by using centralized databases, content schemas, social networks, and algorithms.
If AI can handle the orchestration layer—reformatting, optimizing, and distributing atomic content across whatever surfaces make sense—the platform could become less important than the content itself.
6. Legacy System Replacement Accelerates
External to the web, AI-accelerated software development will make capex projects to rip out legacy software systems and replace them with either something modern or custom drastically cheaper than before.
Where there’s high fixed opex cost (enterprise SaaS licensing, for example), this becomes increasingly attractive for businesses to drive profitability. Early movers in this space will be fascinating to watch.
The Core Principle
Any business model that assumes software has value primarily because it is expensive to create is at risk.
This doesn’t mean all software becomes worthless. It means the source of value shifts:
- From creation cost to quality and reliability
- From features to integration and orchestration
- From one-time builds to continuous optimization
- From selling software to sharing success
Opportunity Spaces
Rather than fighting this disruption, what if we planned for the end of SaaS as we know it? Here are some strategic opportunities:
Bet on Being a Toolkit for AI
Instead of trying to bake more and more functionality into monolithic products, build a toolbox of great software modules that can be composed by AI in service of meeting business requirements.
The question becomes: what capabilities are genuinely valuable as building blocks? What represents real differentiation that AI can’t easily replicate?
Think of it like ingredients for cooking. You don’t need to pre-package every possible meal. You need high-quality, reliable ingredients that can be combined in infinite ways.
Radically Disrupt Hosting
What if instead of selling expensive monthly recurring hosting subscriptions, platforms use serverless infrastructure to massively reduce costs (for both providers and users) and focus instead on monetization through shared success?
Use AI to reengineer hosting solutions toward serverless architectures—make basic managed platform use totally free but capable far beyond current free tiers. Then focus on monetization through whatever business model customers are following:
- Revenue share on payment processing
- Affiliate commissions on sales
- Value-added services (premium support, advanced features, professional services)
This tightly couples platform success with customer success—a powerful economic incentive alignment.
Move Toward Agent-Centricity
Offer specialist AI agents for all sorts of jobs. Given an unlimited budget, what would your dream team look like?
- Business analysis
- Technical direction
- Design
- Development
- QA and testing
- Release management
- Ongoing maintenance
- Growth marketing
- Project management
- Product management
- Customer success
- Business strategy
- Incident response
- Migration specialists
- Technical coaching
Which of these couldn’t be an AI agent? Which would people pay for?
The value shifts from "we built this software" to "we provide this ongoing capability."
Create AI Quality Agents
Use AI to help design and implement mass testing protocols. The resistance to comprehensive quality frameworks has always been resourcing—the cost of writing tests, auditing code, fixing issues, translating content.
But AI makes this increasingly less expensive. What if every plugin, every theme, every piece of open-source software could be continuously:
- Tested for compatibility
- Audited for security
- Improved for performance
- Translated to multiple languages
- Documented comprehensively
Quality becomes affordable at scale.
Make Platforms More Radically Open
My personal favorite: what if platforms participated at a more fundamental level in the economics of the web through protocol-based architecture?
This relates to the atomic content vision. Instead of walled gardens where content is trapped, imagine content as portable, owned assets that can be:
- Published to any surface
- Monetized through any mechanism
- Owned and controlled by creators
- Indexed and discovered across networks
Yes, this involves decentralization technologies. Yes, this means rethinking business models. But it also means aligning with where I believe the web is heading—toward creator ownership and protocol-based interoperability.
What if major platforms were protocols instead of walled gardens?
What This Means for Open Source
For platforms built on open-source software, there’s a window of opportunity to adapt, respond, and reimagine how they bring value in an AI world.
Complex software is beyond AI’s immediate ability to recreate. So platforms must evolve quickly enough to always be more attractive to use—including by AIs—than even an AI-derived bespoke solution.
This requires:
- Recognizing the threat – Not dismissing AI disruption as hype
- Deploying AI aggressively – Using AI to accelerate every aspect of open-source development
- Defining quality frameworks – Having clear standards for what "good" means
- Focusing on composition – Making software modular and AI-friendly
- Betting on outcomes – Selling results, not software
The Uncomfortable Questions
This transformation raises hard questions:
If software becomes cheap to create, what’s valuable?
Quality, reliability, integration, ongoing optimization, trust, support—things that require sustained effort and expertise.
If platforms become toolkits for AI, who pays for them?
The beneficiaries of success: creators who make money, businesses that grow, users who achieve their goals.
If hosting becomes free, how do hosting companies make money?
Shared success, premium capabilities, professional services, advanced features.
If anyone can create software, why open source?
Shared quality standards, ecosystem effects, distribution, trust, governance, longevity.
These aren’t rhetorical questions. They’re strategic challenges that need real answers.
The Alternative Path
Right now, there’s a choice:
Option A: Chase the current leaders by trying to match their SaaS features, fight on their terms, play catch-up in markets they already dominate.
Option B: Bet on where the puck is going—skip ahead to post-SaaS models, agent-centric systems, serverless infrastructure, protocol-based interoperability.
Option A is lower risk but also lower potential return. You might close some gap, but you’ll never leapfrog.
Option B is higher risk but potentially transformative. If you’re right about the direction, you’re positioned ahead of the curve when others scramble to adapt.
The metaphor I keep coming back to: Are you building a fiber optic network or betting on satellite internet?
Both can work. But they represent fundamentally different strategic visions.
Looking Forward: Scenarios for 2025-2026
Thinking ahead 12-18 months from publication, here are scenarios that might unfold—though predicting AI adoption timelines is notoriously difficult:
What Could Move Faster Than Expected
Agent systems might proliferate rapidly. The combination of improved LLM reasoning (OpenAI’s o1, Anthropic’s Claude 3.5, and newer models), better tool-use capabilities, and frameworks like LangChain, AutoGPT, and CrewAI could enable functional agent systems to reach production quickly.
Major platforms could launch agent-centric products (Salesforce’s AgentForce, agent marketplaces) if the technical capabilities mature. The shift from "chat interface" to "autonomous agent" could happen faster than most expect.
Serverless could mature significantly. Technologies like Vercel’s Edge Functions, Cloudflare Workers, and AWS Lambda improvements might make serverless genuinely practical for production workloads. The cognitive overhead could drop rapidly.
Quality automation might work better than expected. AI-powered testing, security scanning, and code auditing tools could become remarkably effective. GitHub Copilot, Cursor, and similar tools might evolve beyond autocomplete to identifying bugs, suggesting improvements, and writing comprehensive tests.
What Could Move Slower
Business model innovation. Despite technical progress, most companies might continue trying to fit AI into traditional SaaS subscription models. Revenue-sharing and outcome-based pricing could remain rare if organizational culture resists change.
True atomic content. While protocols for portable content exist (ActivityPub, Bluesky’s AT Protocol), mainstream adoption could be slow. Walled gardens might remain dominant, and most creators might still not truly own their content or audience.
CMS disruption. Traditional platforms could prove more resilient than expected. The combination of ecosystem effects, SEO authority, and existing content might keep people locked into legacy systems longer than pure technical capability would suggest.
Key Uncertainties
-
Technology vs. business models. The tools might exist to build agent-centric, serverless, AI-powered systems, but monetizing them could require business model innovation that’s culturally harder than technical innovation.
-
Integration vs. features. The winners might not be building the best individual capabilities—they could be building the best integration layers that let agents orchestrate multiple tools.
-
Trust as bottleneck. People might be willing to let AI write code and create content, but much more hesitant to let AI make consequential decisions autonomously. The "human in the loop" could remain critical longer than technologists expect.
-
Open source’s role. Rather than becoming less relevant, open-source frameworks for agents (Eliza, LangChain, etc.) might become more important because they provide transparency and control that proprietary agent systems don’t.
The disruption appears real, but the path from "technically possible" to "mainstream adoption" often takes longer than technologists expect.
Where I Might Be Wrong
This analysis assumes several things that could turn out to be incorrect:
I’m assuming that AI capability continues improving at current rates and that software creation costs will continue falling. But there could be plateaus—both technical and economic—that slow or halt this trend. If model improvement slows, training costs don’t decrease, or energy constraints become binding, the "software creation approaches zero cost" assumption breaks.
I’m assuming that quality and trust won’t become bigger differentiators than I’m modeling. If AI-generated code proves fundamentally less reliable, more brittle, or harder to maintain than human-written code, the economics change entirely. Software creation might get cheaper, but the cost of ensuring quality could remain high.
I’m assuming that regulatory and legal frameworks won’t significantly slow AI adoption. But if liability, copyright, or safety regulations create substantial friction—particularly for autonomous agents making decisions—the timeline could be much longer. The "move fast and break things" era might be over for mission-critical systems.
I’m assuming that network effects and switching costs won’t protect incumbents as effectively as they have historically. But platform economics shows that entrenched platforms can be remarkably resilient. WordPress, Salesforce, and other incumbents have strong ecosystem effects that pure technical capability might not overcome.
An alternative view would emphasize that most businesses don’t actually want to build custom software—they want proven, reliable solutions with clear SLAs and vendor accountability. The "SaaS is dead" narrative might underestimate how much value businesses place on not having to think about infrastructure, security, compliance, and operations.
I could be wrong about the disruption sequence. Christensen’s disruption playbook shows that predictions about which segments get disrupted first are often incorrect. The actual sequence might be different, or certain segments might prove much more resistant than I’m modeling.
If I’m overestimating the pace of AI capability improvements, underestimating the importance of trust and reliability, or miscalculating the strength of incumbent advantages, this entire framework might be off by years or even invalid. But understanding the direction—even if uncertain about timeline—still helps inform strategic decisions today.
Conclusion
I wrote this because these ideas are intersectional with decisions we’re all making today—about product strategy, technology investments, business models, and where to focus energy and resources.
This is presented as thought leadership and an invitation to discussion. I hope it inspires ideas and debate.
The fundamental question: When the cost of creating software approaches zero, what becomes valuable?
My answer: Quality. Reliability. Integration. Trust. Shared success.
The companies and platforms that figure out how to deliver those things—and monetize them fairly—will thrive in a post-SaaS world.
Those that cling to business models predicated on software creation being expensive will find themselves disrupted by competitors who’ve adapted to the new reality.
Right now we can decide to roll out a fiber optic network and chase incumbents, or bet on next-generation infrastructure and skate to where the puck will be.
I know which bet I’d make.
Published: October 21, 2024
Related Content:
- AI Disruption Framework – Systematic framework for evaluating AI disruption
- Intent-Driven Commerce – How AI agents change e-commerce
- Intents All The Way Down – Foundation for intent-driven thinking