SaaS reinvented

SaaS is ~~dead~~ reinvented.

The last 12 months have been challenging for anyone in the software industry. No one has been immune, but the global SaaS providers have taken the brunt of enterprise buyer uncertainty and declining sentiment. As we explored in our recent article AI with Real ROI, organisations are increasingly scrutinising digital investments and prioritising solutions that deliver measurable business value. We examine the cause of this disruption and our view on where we believe the Software as a Service market is likely to settle. Building on this shift toward value-driven technology, we have also launched an AI Application Catalogue, a practical collection of use cases aligned to the services we deliver, helping organisations identify where AI can generate measurable impact across supply chain, procurement, logistics and operations.

Andreessen Horowitz, released an article called Good news: AI will Eat Application Software by Alex Immerman and Santiago Rodriguez (Good news: AI Will Eat Application Software | Andreessen Horowitz). The perspectives are well informed, unsurprising given one of the founders Mark is largely credited with creating the Internet and getting some of the biggest technology predictions right over the last two decades.

https://a16z.com/good-news-ai-will-eat-application-software/

The article described a SaaSpocalypse using Hamilton Helmer’s Seven Powers (a company’s competitive advantages used by the acclaimed Acquired podcast) scale, network effects, counter positioning, switching costs, brand, cornered resources, and process power to argue software as a class is going to remain and thrive, but will change. They predict a bifurcation of software solutions as some incumbents adapt at their core inclusive of commercial models, features, and functions, whilst others find themselves subject to being the next Blackberry business school case study for all the wrong reasons.

We launched IRIS, an AI First consulting and services company powered by Argon & Co in mid-2024. We had some beliefs based on the potential we saw with Large Language Models (LLM) and built our solutions and services around these fundamentals. The potential for a new class of AI applications and associated management platforms to emerge seemed obvious to us at the time. We found ourselves in a unique position where our services market fit and product market fit simultaneously worked. Something I have not experienced in my career as we follow our hunches and build platform-based applications that address the deficiencies of legacy applications without adding technical complexity. Feel free to peruse our rapidly expanding catalogue of AI applications and platform for more context.

We authored an article explaining the four forces that continue to shape our industry and the applications market. I will not go over old ground, but most executives we speak to broadly agree with our perspective. It has become extremely difficult to make no regrets decisions for application investments.

In the article we made six predictions, controversially forecasting SaaS is dead or reinvented within 36 months. We based our prediction on the challenge of data access limited to the remit of the SaaS domain, existing fragmented application estates, over indexing on Excel and the commercial pressures starting to mount as per task pricing targeted labour budgets for new market entrant software services.

The core change underpinning the sustained change in the applications market is the speed of innovation by the AI coding agents. The industry was rocked recently when Claude Opus 4.6 and Codex 5.3 were released. The graphic is a visual representation of the concept of exponentiality that Dario Amodei of Anthropic speaks to with great passion and references as early as October 2024 in his essay ‘Machines of loving grace’. (Dario Amodei — Machines of Loving Grace. The capability of these models is astounding but requires the right people and a very measured approach.

https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/

We have Opus 4.6 leaping from independent coding of 6.5 hours to 14.5 hours and what makes this improvement so important are two factors. Opus and Codex had most of their code base built by the prior models – self-improvement at scale. And secondly the release speed. Some of us will remember waiting months for the next ChatGPT release. GPT 5.3 was released 27 days after Codex 5.2 and we only see this continuing to accelerate.

Looping back to the Andreessen Horowitz article, finding an application that exhibits SaaS like attributes with a competitive moat continues to be challenging but it has not stopped the flood of investment, particularly in agentic based solutions. Some of these are sticking and experiencing rapid revenue growth. The new category of AI coding tools like Cursor, Codex, Antigravity and Claude Code have driven downward pressure on SaaS market sentiment. Domain and functionally specific solutions like the legal platform Harvey and the Deep Research service Perplexity have captured the hearts, minds, and wallets of customers despite frontier model providers deploying comparable services to create revenue.

Where does that leave SaaS and the broader application category? It is easy to jump on the bandwagon and proclaim it is doomed, but we have a more precise set of views formed through shared experiences with our clients, and our research. Application cohabitation will be the reality of the near future and where to draw the line will be determined by many factors, some we detail below.

The user experience

One of the beauties of ChatGPT is the way we now dynamically use information, whether for our personal or professional lives. Consumer products like NotebookLLM, and Nano Banana have opened different ways of interacting with systems and these are coming to the enterprise. The move from highly prescriptive business processes at a system level to a more intuitive, ‘what should I focus on today’ design is becoming a reality, and we spoke about this as the Experience Layer in our original article. This is about the workforce changes we are about to witness, the role may be the same, but the task is done with AI augmentation.

Process commoditisation or process differentiation

High volume, repeatable, standard processes of execution don’t need bells and whistles. They need reliability and it is the domain of SaaS for the near future. Systems of execution taking inputs from systems of innovation. Innovative, customer led personalisation is the domain of the AI/ agentic application, particularly when paired with an Experience Layer.

Systems economics and commercial innovation

The days of high-cost license lock-ins are numbered, and the term platform hostage can be consigned to the past. Per task services emerged early with start-ups like Sierra, with an understanding IT budgets are finite, but people budgets are not. Our independent research provided some assumption-based ROI comparisons we found both interesting and aligned to our commercial experiences with our clients.

SaaS 2.0

Coding acceleration goes both ways. Existing SaaS providers have the same opportunity to re-architect and re-platform their systems at speed and in a way previously never contemplated. The confidence level of this platform transformation will come down to the people. Turns out, the software industry built to underpin process automation through applications is as change resistant as any other part of your business. Our hunch is the variability of development velocity and quality currently being measured and reported is a factor of the teams, not the tools. This is our experience and means great teams become greater.

Do we feel SaaS is dead/ reinvented? We tend to agree with the bifurcation prediction made by the authors of the Andreessen Horowitz article but with an extension.

Those SaaS providers unwilling or unable to embrace the new ways of working to rearchitect their existing applications face a slow extinction event. In the early days, mid-market solutions will most be at risk if they can’t demonstrate value adding AI centric features.

Inventive courageous SaaS providers willing to ‘innovate themselves’ with the right talent to make it happen will have a fighting chance.

New emergent AI applications will continue to grow the total addressable market at an accelerated rate. Global forecasts for AI Software are in the order of 35% CAGR for the next 5 years. That growth funds a lot of AI application category winners.

The AI technology responsible for the industry disruption of SaaS could be incumbents lifeline.