SaaS Isn’t Dead, It’s Getting a Brain
Animation by Grok
Will AI crush SaaS? This month Gerry Gimenez, our recently joined Director of Software & AI coverage, and Andrew Birmingham of Mi3, duke it out over thge future of SaaS in an AI world. In this article, Gerry makes the case that the death of SaaS is greatly exaggerated. You can find Andrew's bear case, ‘Will SaaS Survive the AI Shockwave?’ here.
The death of SaaS is greatly exaggerated
When Marc Andreessen famously declared in 2011 that ‘software is eating the world’, he kicked off the Software as a Service (SaaS) gold rush that defined the 2010s. Now, as generative AI crashes onto the scene, the tech commentariat has rushed to write SaaS’s obituary. ‘AI is the death of SaaS’, the ‘SaaS-Pocalypse’, they proclaim. But here’s the thing about obituaries: they’re often premature.
We’ve been advising tech companies since the dial-up era, and we’ve seen these ‘sky is falling’ moments before. The move to cloud was supposedly going to kill enterprise software. Mobile was going to kill desktop. Each time, the doomsayers were wrong. Not because change wasn’t real, but because they fundamentally misunderstood what was happening.
The real story isn’t about death, but evolution. What we’re seeing isn’t SaaS companies being replaced by AI. It’s SaaS companies becoming AI companies and, crucially, AI companies becoming SaaS companies.
SaaS is embracing AI
McKinsey estimates generative AI can add as much as $US4.4tn annually to the global economy. But here’s what’s striking: 70-80 per cent of SaaS companies have already integrated AI into their products, while 80-90 per cent of organisations will be using AI-powered applications by end of 2025. This isn’t cannibalisation; it’s convergence. Craig Blair, partner and co-founder at Airtree, told us that ‘over half of the companies we’ve backed in Airtree’s last two funds are AI-native or have incorporated AI fully into their products’.
Take a walk through the B2B software landscape today, and you’ll see transformation everywhere. Salesforce didn’t curl up when ChatGPT launched; it embedded Einstein AI and launched Agentforce. CEO Marc Benioff noted that ‘the speed of innovation is far exceeding the speed of customer adoption’. Adobe responded to AI tools like Midjourney and DALL-E by embedding AI into their own products. Adobe’s AI offerings contributed more than $125mn ARR in 2025. And Microsoft has integrated multiple AI models into its 365 product suite, offering customers both OpenAI and Anthropic models within Copilot.
The pattern is clear; successful SaaS companies aren’t running from AI. They’re running toward it.
The new AI giants… Still looks like SaaS
Now flip the script. Look at the hot AI companies making headlines – OpenAI, Anthropic, Perplexity. How do they actually make money? Through good old-fashioned SaaS business models.
Anthropic, valued at $183bn and approaching a $5bn annualised run-rate, doesn’t sell you AI in a box. It offers Claude through APIs and subscription tiers, billing 300,000 business customers as of August 2025, with sevenfold growth in large enterprise accounts year-over-year.
OpenAI? Same textbook. Over 3 million business users, more than 4 million developers, and 800 million weekly users on freemium and subscription models. Their business is built on SaaS 101: recurring revenue, usage-based pricing, and cloud delivery.
Even the pricing wars mirror the great cloud migration. Costs to query models like GPT-3.5 have plummeted 280-fold since late 2022, dropping from $20 to just $0.07 per million tokens. Remember AWS, Azure, and Google Cloud slashing prices to grab market share? We’re watching the same movie, just with new actors.
AI isn’t a new business model. The transformation isn’t in deployment or go-to-market strategy – it’s in what software can now do.
Article continues below.
Like this article? Join the thousands of tech founders, board members and investors who subscribe to our free monthly newsletter, Tech Round-Up. Sign up below!
The real transformation: Intelligent software
The fundamental shift isn’t about SaaS versus AI; it’s about what software is. Traditional SaaS gave you tools to work faster. AI-powered SaaS gives you a colleague who can work alongside you.
Your CRM predicts which deals will close and drafts personalised outreach. Your analytics platform spots anomalies you missed and explains what they mean. Your code editor writes entire functions based on your comments.
According to Gartner, by 2028, 33 per cent of software applications will include agentic AI – autonomous systems capable of setting their own goals, adaptively learning, and executing multi-step workflows. By the same year, around 15 per cent of day-to-day work decisions will be made autonomously, up from less than one per cent in 2024. We’re not talking about better spell-check, but software that reasons, decides, and acts.
Gartner warns, however, that more than 40 per cent of agentic AI projects may fail due to escalating costs, unclear business value and ROI, and inadequate risk controls and governance.
The Inconvenient truth: AI implementation is challenging
Here’s where the narrative gets uncomfortable. Behind the AI hype, some numbers are concerning.
MIT’s Project NANDA reported that 95 per cent of task-specific GenAI deployments failed to reach production with measurable P&L impact, despite $30-40bn spent. The core problem: most GenAI systems don’t retain feedback or improve over time. They are ‘brilliant one-shot wonders’ that can’t learn from mistakes. That’s exactly what traditional SaaS platforms spent two decades mastering.
Additionally, nearly 60 per cent of AI leaders cite integrating AI with legacy systems and addressing risk and compliance concerns as their primary challenges. Translation: bolting AI onto existing software is much harder than it looks in the demo video.
These enterprise deployment challenges coexist with broader GenAI adoption success. A Wharton study reveals 75 per cent of leaders are seeing positive returns, while BCG reported future-built firms are achieving 1.7x revenue growth and 1.6x EBIT margins compared to laggards.
The future is hybrid, not binary
Despite all the disruption, AI isn’t replacing SaaS anytime soon. As Symon Vegter, partner at Advent Partners told us, ‘We see AI as helping transform software. But not replace it’. Business workflows are deeply embedded in existing platforms, and decades of operational refinement aren’t easily replicated.
Businesses require auditability, control, and verification features that current AI agents struggle to deliver, especially in regulated sectors. Try telling your compliance officer an AI ‘just decided’ to approve a transaction with no audit trail. Many industries require governance capabilities SaaS platforms excel at.
Three key SaaS advantages are still hard for AI to replicate: deep vertical experience building specialised solutions, access to vast customer data for benchmarking and machine learning, and the integration of people, process, tech and governance.
Bain’s analysis outlines four AI-era SaaS scenarios: AI enhancing SaaS (strongholds), spending compression (open doors), AI outshining SaaS (gold mines), and AI cannibalising SaaS (battlegrounds). Only one involves total disruption; the rest show AI boosting incumbents, shifting budgets, or creating new opportunities.
The winners won’t be pure SaaS or pure AI – they’ll be those blending AI intelligence with SaaS scalability.
SaaS fundamentals with AI advantages
Who wins in this ‘AI eating SaaS’ paradigm?
Firstly, companies that build with both SaaS and AI. The SaaS incumbents that moved fast, such as HubSpot’s Breeze AI, Atlassian’s Rovo, and Xero’s JAX, protected their moats by deepening them with AI. These companies leverage unified customer data for contextual help, natively within their ecosystems.
Next, AI-native companies that mastered SaaS fundamentals. Alex Danieli, from Teoh Capital, shared with us that ‘My view of AI is that the way to win is distribution. And SaaS, in particular vertical SaaS, already has that distribution’. Anysphere’s Cursor AI scaled from $1m to $100m in annual revenue within a year by applying SaaS best practices to cutting-edge AI.
Finally, SaaS companies that understand orchestration beat AI models. Microsoft’s multi-model approach (letting customers choose between OpenAI, Anthropic, or its own MAI models) is smart because it recognises model quality commoditises quickly. Real value comes from integration, deployment, security, and the countless unglamorous details that make B2B software work at scale.
The intelligent SaaS playbook
Some best practices are emerging:
Embed AI across workflows, not as a bolt-on: Winners aren’t just adding an AI tab, they’re rethinking workflows with AI at the heart. MIT data shows external partnerships deliver a 67 per cent deployment success rate, compared to 33 per cent for internal-only builds. Crucially, pursue growth while maintaining profitability (aiming for the Rule of 40 or above).
Build defensible data moats: Scale alone isn’t enough. Winning data assets must be unique, comprehensive, and continuously improving. Established SaaS platforms benefit from aggregate customer data that enhances AI models in ways individual companies can’t replicate.
Prioritise vertical solutions: Horizontal AI can quickly become a commodity. Domain-specialised agents, engineered for industry-specific workflows, remain more defensible. Your expertise in pharma supply chains or insurance claims is harder to reproduce than basic generative features.
Embrace the pricing model shift: Seat-based pricing is giving way to hybrid models. Over a third of companies now mix seats, usage-based fees, and outcome-based charges. Traditional SaaS finance teams hate this, but customers love paying for outcomes, not licences. Proponents of the ‘SaaS is dying’ thesis argue AI agents will replace per-seat apps, that ‘agents are the new seats’. Microsoft’s Satya Nadella argued that ‘The application layer is collapsing into agents’. Business logic moves to an AI tier, interfaces recede into APIs, pricing shifts from licences to outcomes, and rising inference costs plus cheap custom builds erode traditional moats.
Laggards Lose, Winners Converge
We’ve worked with enough technology companies to spot the patterns. SaaS isn’t dying, it’s evolving. AI isn’t killing SaaS, rather, it’s driving the biggest shift in B2B software since the cloud.
By 2030, the distinction between SaaS companies and AI companies will feel as relevant as the one between Internet companies and regular companies today.
SaaS is getting smarter, faster, and frankly, more dangerous for those that don’t adapt. The threat is real for laggards, but for companies moving fast, the opportunities are real.
The death of SaaS? That’s one funeral we won’t be attending. We’ll be too busy working with clients who are building software that doesn’t just serve human intelligence but augments it – profitably, defensibly, and at scale.
_________________
Gerry Gimenez joined North Ridge Partners from Amazon recently, as Director of Software & AI coverage. His experience spans M&A, corporate development, management consulting (at PwC Strategy&), and transaction law (at Lefosse/Linklaters). He holds an MBA from the University of Chicago Booth School of Business, where he was awarded an Instituto Ling fellowship.