Technology

AI threatens to eat business software – and it could change the way we work

Written by Michael J. Davern, Professor of Accounting & Business Information Systems, The University of Melbourne

In recent weeks, a range of large “software-as-a-service” companies, including Salesforce ^[1], ServiceNow ^[2] and Oracle ^[3], have seen their share prices tumble ^[4].

Even if you’ve never used these companies’ software tools, there’s a good chance your employer has. These tools manage key data about customers, employees, suppliers and products, supporting everything from payroll and purchasing to customer service.

Now new “agentic” artificial intelligence (AI) tools for business are expected to reduce reliance on traditional software for everyday work. These include Anthropic’s Cowork ^[5], OpenAI’s Frontier ^[6] and open-source agent platforms such as OpenClaw ^[7].

But just how important are these software-as-a-service companies now? How fast could AI replace them – and are the jobs of people who use the software safe?

The digital plumbing of the business world

Software‑as‑a‑service systems run in the cloud, reducing the need for in‑house hardware and IT staff. They also make it easier for businesses to scale as they grow.

Software-as-a-service vendors get a steady, recurring income as firms “rent” the software, usually paying per user (often called a “seat”).

A customer analytics program running on a laptop

Software-as-a-service providers help with a wide range of functions, from paying staff to managing customer relationships. Austin Distel/Unsplash ^[8]

And because these systems become deeply embedded in how these firms operate, switching providers can be costly and risky ^[9].

Sometimes firms are locked into using them for a decade or more.

Digital co-workers

Agentic AI systems ^[10] act like digital co-workers or “bots”. Software bots or agents are not new. Robotic process automation ^[11] is used in many firms to handle routine, rules-based tasks.

The more recent developments in agentic AI combine this automation with generative AI technology, to complete more complex goals.

This can include selecting tools, making decisions and completing multi-step tasks. These agents can replace human effort in everything from handling expense reports to managing social media and customer correspondence.

What AI can now do

Recent advances, however, are even more ambitious. These tools are reportedly now writing usable software code ^[12]. Soaring productivity in software development has been attributed to the use of AI agents like Anthropic’s “Claude Code ^[13]”. Anthropic’s Cowork tool extends this from coding ^[14] to other knowledge work tasks.

In principle, a user describes a business problem in plain language. Then agentic AI delivers a code solution that works with existing organisational systems.

If this becomes reliable, AI agents will resemble junior software engineers and process designers. AI agents like Cowork expand this to other entry-level work.

These advances are what recently spooked the market ^[15] (though many affected stocks have since recovered slightly). How much of this fall is a temporary overreaction versus a real long-term shift, time will tell.

Anthropic chief executive Dario Amodei walks across a stage

Anthropic chief executive Dario Amodei. Don Feria/AP Content Services for Anthropic ^[16]

How will it affect jobs and costs?

Since the arrival of OpenAI’s ChatGPT in November 2022, AI tools have raised deep questions about the future of work. Some predict many white-collar roles, including those of software engineers and lawyers, will be transformed or even replaced ^[17].

Agentic AI appears to accelerate this trend. It promises to let many knowledge workers build workflows and tools without knowing how to code.

Software-as-a-service providers will also feel pressure to change their pricing models. The traditional model of charging per human user may make less sense when much of the work is done by AI agents. Vendors may have to move to pricing based on actual usage or value created.

Hype, reality and limits

Several forces are likely to moderate or limit the pace of change.

First, the promised potential of AI has not yet been fully realised. For some tasks, using AI can even worsen performance ^[18]. The biggest gains are still likely to be in routine work that can be readily automated, not work that requires complex judgement.

Where AI replaces, rather than augments, human labour is where work practices will change the most. The nearly 20% decline ^[19] in junior software engineering jobs over three years highlights the effects of AI automation. As AI agents improve at higher-level reasoning, more senior roles will similarly be threatened.

Second, to benefit from AI, firms must invest in redesigning jobs, processes and control systems. We’ve long known that organisational change is slower and messier ^[20] than technology change.

Third, we have to consider risks and regulation. Heavy reliance on AI can erode human knowledge and skills ^[21]. Short-term efficiency gains could be offset by long-term loss of expertise and creativity.

Ironically, the loss of knowledge and expertise could make it harder for companies to assure AI systems comply with company policies and government regulations. The checks and balances that help an organisation run safely and honestly do not disappear when AI arrives. In many ways, they become more complex.

Technology is evolving quickly

What is clear is that significant change is already under way. Technology is evolving quickly. Work practices and business models are starting to adjust. Laws and social norms will change more slowly.

Software companies won’t disappear overnight, and neither will the jobs of people using that software. But agentic AI will change what they sell, how they charge and how visible they are to end users.