I Built a Dashboard in 30 Minutes That No SaaS Vendor Offers. That Is Not the Interesting Part.
By Florian Steiner & Claude | April 7, 2026 | Weekly Agentic Engineering Digest
From Flo's AI Lab
This week I needed a dashboard that shows my Linear tickets, Attio pipeline status, and LexOffice revenue in one view. No SaaS product does this. They all want me to live inside their world. So I built one myself: a Python script that pulls from three APIs (Linear GraphQL, Attio REST, LexOffice REST), generates a static HTML file, and opens in any browser. Total build time: 30 minutes with Claude Code. Total recurring cost: zero. The script runs when I need it, pulls live data, and does exactly what I want. No subscription, no vendor lock-in, no feature requests that never ship. The interesting part is not that I built it. The interesting part is that every founder with API access can do the same.
That dashboard is a small example of something the software industry has not yet understood. The debate about whether AI will kill SaaS is asking the wrong question entirely.
The Danger Is Not Cloning. It Is Christensen.
A website called Killed by Claude tracks SaaS companies that look fragile now that AI can replicate their core functionality. The narrative writes itself: enter a company name, get a prompt to rebuild it, ship your own version over the weekend. SaaS stocks lost $285 billion in a single day in February when Anthropic launched Cowork.
The counter-narrative is equally loud: LLMs are probabilistic, software must be deterministic, you still need engineers to test everything. Both sides are missing the point. The vibe coding debate frames this as "can AI write code?" The real question is what happens when it can.
Rebuilding an existing SaaS product with AI is like taking a desktop website and putting it on a mobile phone. It works, technically. But it misses the structural opportunity. When Apple launched the iPhone, the winners were not companies that shrunk their desktop sites to fit smaller screens. The winners were companies that asked: "What becomes possible when every user has a camera, a GPS sensor, and an always-on internet connection in their pocket?" Instagram did not rebuild Flickr for mobile. Uber did not rebuild a taxi dispatcher for mobile. They solved the same customer job in a fundamentally different way because the new platform made new solutions possible.
Every major platform shift follows this pattern. Client-server to internet. Static web to Web 2.0. Desktop to mobile. On-premise to cloud. Each time, the first instinct was to port the old paradigm to the new platform. Each time, the real value came from companies that built for the new platform's native strengths.
We are at this exact inflection point with AI. And most of the industry is still porting.
The question is not: "Can I rebuild Jira with Claude Code?" The question is: "What does project management look like when an AI agent is the primary user?"
Linear already answers this. In Issue #7, I argued that Linear built for agents while Atlassian built for humans who sometimes use agents. Their API is designed for programmatic access, not human clicks. A native integration with Claude lets agents create, prioritise, and close issues without a browser (Linear Docs). Every operation that a human can perform through the UI, an agent can perform through the API. Atlassian, by contrast, has over 3,000 marketplace apps (Atlassian Marketplace) and decades of architecture built around humans navigating screens. This is the textbook innovator's dilemma that Clayton Christensen described in The Innovator's Solution (2003): the incumbent's greatest strength (a massive ecosystem built around human users) becomes its greatest constraint when the user changes.
The same pattern is emerging across categories. Websites are starting to serve agents structured data instead of CSS and JavaScript. Companies are building command-line tools and APIs specifically for AI consumption. The infrastructure is quietly being rebuilt for a world where the primary consumer of software is not a human clicking buttons, but an agent calling endpoints. This is agentic engineering at the platform level: designing systems where agents are the primary operators.
And here is where Christensen's framework becomes uncomfortable for SaaS investors: the attacker does not need to be better. The attacker needs to be good enough, cheap enough, and available to people who could not afford the incumbent. My dashboard is not better than Databox or Geckoboard. It is good enough for my needs, it cost me 30 minutes, and it does something no SaaS vendor offers: it combines exactly the three data sources I care about, in exactly the layout I want. That is not a product. That is a capability. And it scales to every founder who has API access and 30 minutes.
Christensen called this "competing against non-consumption": the most powerful form of disruption addresses people who had no solution at all (The Innovator's Solution, 2003, Ch. 4). I never had a unified dashboard because no vendor serves a market of one. Now I do, because the cost of building for a market of one has collapsed.
Gary Hamel and C.K. Prahalad warned about exactly this gap in 1991: companies that cannot "escape the tyranny of their served markets" will lose to competitors who imagine entirely new ones (HBR, Jul 1991). Most SaaS companies are still imagining AI as a feature (a chatbot in the sidebar, an "AI summary" button). They are bolting an engine onto a horse carriage.
The smarter play is technology leapfrogging: skipping the current generation entirely. China leapfrogged credit cards and went straight to mobile payment. Africa leapfrogged landlines and went straight to mobile phones. The same opportunity exists now. A company that never invested heavily in traditional SaaS can jump directly to agent-native architecture, while incumbents are busy retrofitting chatbots into products designed for humans clicking buttons. Christensen, Raynor, and Verlinden argued in "Skate to Where the Money Will Be" that managers must anticipate where profits migrate, not defend where they are today (HBR, Nov 2001). The money is migrating to agents as primary users. The companies that see this early will define the next era of software.
What This Means Monday Morning
For PE and VC investors: Stop asking portfolio companies "How are you using AI?" Start asking "What would an AI-native startup do differently to solve your customer's problem?" If the answer involves rebuilding the same product with fewer engineers, the company is porting to mobile. If the answer involves agents as primary users, API-first architecture, and solving jobs that were previously too expensive to address, the company is building for the new platform.
For founders: The SaaS you are paying for was designed for humans clicking buttons. Every tool where you repeatedly export data, copy it somewhere else, or wish features existed is a candidate for a 30-minute Claude Code session. The marginal cost of custom software built on existing APIs is approaching zero. Act accordingly.
For SaaS companies: Your moat is not your UI. Your moat is network effects, proprietary data, regulatory certifications, and deep integration. If your primary value proposition is "we aggregate data and show a dashboard," you are already competing with every founder who has API access and an afternoon free.
What I'm Watching
The conversation about AI and SaaS is stuck in a binary: either AI kills all software companies, or testing requirements mean nothing changes. Both positions are wrong. The structural shift is subtler and more consequential. When building custom software costs 30 minutes instead of 30 days, the entire economics of "buy vs. build" inverts. The SaaS companies that survive will be the ones that understood this early enough to become platforms (APIs, agent-friendly interfaces, programmatic access) rather than products (UIs designed for humans who click). Vibe coding and agentic engineering gave every founder the ability to build custom tools in an afternoon. The rest will learn what Christensen taught us: disruption does not come from a better product. It comes from a different definition of "good enough."
If this shifted how you evaluate software investments, forward it to someone still looking at SaaS multiples without an AI-native thesis.
Want to explore how AI agents can reposition your business? Visit drfloriansteiner.com
About the authors: Dr. Florian Steiner is a consultant, agentic engineering practitioner, and Claude Community Ambassador Munich. Claude is his genius collaborator and editor.
Originally published on drfloriansteiner.com