The end of SaaS?
In 1999 Mark Benioff and 3 co-founders launched Salesforce, a new software company with a mission to disrupt the world of CRM Software. Until that point, CRM software had been on-premise deployments of tools like Siebel and Oracle. Benioff’s vision was that in the future companies wouldn’t buy software and then pay to have it maintained, they would instead subscribe to use it for an ongoing fee. Salesforce was the first cloud-hosted Software as a Service (SaaS) product built from scratch. But their big breakthrough came when in 2000 they staged fake protests at Siebel’s user conference with anti-software “protesters” proclaiming the end of software which bold move catapulted them into buyer’s consciousness.
The move to SaaS was not overnight, but it is estimated that SaaS represented over 70% of software used in business by 2024, a figure predicted to rise to 85% by 2025.
But today, the question increasingly being asked is whether that is a peak from which the SaaS market will now start to decline, as a new model is emerging – one that in a perhaps tongue-in-cheek play on words is being titled “Services as Software”.
Having been at the heart of the SaaS world with my previous business, Thoughtonomy, and now being at the heart of the Services as Software world with OpenDialog, my answer to that question is a resounding yes. Like the move to SaaS, it won’t be overnight, but it will happen. Why? Because it makes so much sense for buyers, and also offers software providers access to an entirely new market opportunity – if they are bold enough to embrace the new paradigm.
The end of RPA?
If we think about how a business delivers white-collar work, it’s clear that despite advances in technology, there remains a huge reliance on people. While automation – and particularly the use of frictionless automation technologies such as RPA – has improved efficiency and reduced reliance on human labour, that really only worked for structured, defined business processes based on structured input and data. So efficiency improved for defined, routine back office processes, but many other tasks remain the domain of human workers.
But AI – particularly Agentic AI – completely changes the landscape. Powered by large language models (LLMs), AI Agents can converse in any language, to read and interpret structured and unstructured content, reason and make decisions, generate output, set priorities, and direct tasks. The scope of work output that can be automated by AI Agents includes not only what is currently addressed by RPA but also the many, many more tasks that couldn’t be automated before. It fundamentally changes how work gets done.
How work gets done
But there’s another shift too. A responsibility shift. In today’s world, a business hires human resources to deliver output. They may hire them directly, or outsource to a third-party service provider who then provides the staff, perhaps from a low-cost location as part of an efficiency play. The business, or perhaps the service provider, will use software tools – like Salesforce, for example – to get work done. They may also use automation tools, such as RPA, to deliver some of the work. But ultimately, the business or their outsourcer is responsible for configuring, managing and orchestrating both the automated and the human work to achieve their desired outcomes.
In the world of Services as Software, where, as highlighted above, the AI Agent can be given goals and can then converse, read, interpret, reason, make decisions and take actions, generate output, set priorities, and direct tasks, the provider of the Agent – the software provider – can take responsibility not just for the tools and systems on which work gets done, but for the work itself.
Paying for Work
We are starting to see this emerge in software provider business models. Companies like Salesforce, Zendesk, 11x, and OpenDialog are offering commercial models where the software provider takes responsibility for the delivery of work, and customers pay only for that software when work is executed – in the form of interactions, completed tasks, or even successful outcomes such as tickets closed, queries answered, or requests handled.
It’s easy to see why that would be attractive to customers. An AI Agent, able to operate 24/7 and infinitely scale to meet demand represents a compelling way to improve productivity, reduce costs, and break the linear relationship between growth and headcount, but also to provide the potential to create and deliver services and solutions which would be impossible or economically unviable with human resources. With the right commercial agreements in place, businesses can map a direct relationship between value in the form of work, and costs in the form of payment for AI Agent software. Rather than paying SaaS subscriptions for unused or underused licenses or products, flexible usage-based pricing represents the procurement of software in its purest form.
The Opportunity for Businesses
Let’s consider a hypothetical example of an insurance company “INS”, and a software provider “PAS,” whose core insurance administration system is used by INS on a per-seat cost basis. INS has 50 agents with an average loaded cost of $35,000 per annum, each of whom manages an average of 500 customer interactions a month across quoting, selling, contracting, handling queries, managing adjustments, processing claims and renewals. PAS charges a seat-based fee of $160 per user per month for their software.
Seeking to reduce operational costs and improve customer experiences, INS starts working with a Services as Software company, OpenDialog, who provide an AI Agent Management System to help organisations build AI Agents to automate tasks – in this case, in the form of customer engagements within the Insurance industry.
After training, OpenDialog Agents are able to execute many of the tasks being delivered by INS’s human agents. OpenDialog doesn’t charge a SaaS subscription fee for those Agents but instead for the work those agents deliver, let’s say at $2 per task. Within a month, those AI Agents are autonomously delivering 20% of the work being done by the human agents, equating to 5,000 tasks per month.
Example FTE cost before and after AI Agents
For INS, the cost of the insurance platform from PAS has declined, but they are now investing in OpenDialog such that in total they are now paying $8,400 per month more for software. But importantly, when it comes to the Services as Software they buy, they are only paying for work being done. And the cost of that work is far lower than the cost of human resources doing it. In this example, INS is able to reduce the workload by 10 FTE, a saving of around $30,000 per month. And that’s only at a 20% automation rate – in reality, over time, automation rates as high as 90% have been achieved. The prize for INS in taking this approach is a significant reduction in costs – and the Services as Software commercial model means there is no commercial risk in investing in software that may not be being fully used.
The Opportunity for Software Companies
Of course, the only loser in this example is PAS, the software provider, who sees a decline in user-based subscriptions. This is exactly why companies like PAS are developing technology partnerships with OpenDialog to incorporate AI Agent functionality into their offerings—thus supplementing declining seat-based software revenues with new revenue streams for the execution of work in their platform. By using their own expertise to build AI Agents in a Management System designed specifically by OpenDialog for that purpose, they can leverage their domain expertise and Intellectual Property in how a business in their core domain operates by using their software, without needing to invest significant time and effort into creating the underlying Agent architecture.
The market opportunity is vast and untapped. Instead of convincing customers to migrate from a PAS competitor or spend more on enhanced features, they can access a customer’s people spend – in our example a market opportunity five times bigger than the revenue from PAS’ subscriptions, and already something the customer will have budgeted for.
This is one example of how the software landscape is developing. Agentic AI has made this possible, and forward-thinking buyers and providers are rapidly making it a reality.
Want to learn more? Get expert advice on AI Agents and see what we can do to transform your business.
