Gartner is on the 1. July 2026 given a number in the market, which since then runs through all IT media: $234 billion in terms of spending on enterprise software, agent AI is at stake by 2030. That’s about 20 percent of what companies are expected to spend on enterprise application SaaS by 2030. Notable is less the number than the mechanism behind it – and this also affects medium-sized users, although the forecast at first glance describes a problem of software manufacturers.
What Gartner actually says
The term Gartner coined for this is “agentic arbitrage”. The logic: Classic enterprise software is licensed per user. This model works as long as people log into surfaces, click and maintain data. But if an AI agent does the task across systems – Drawing data from the ERP, updating it in the CRM, closing a ticket – then the number of people who still log in drops. The software, as Gartner puts it, becomes invisible. This breaks the link between user growth and revenue growth, on which the business model of many providers rests.
Gartner analyst George Brocklehurst puts it on the formula that companies are increasingly no longer buying software primarily for people, but for agents. Important is the classification that goes down in many headlines: Gartner speaks explicitly no the demise of the SaaS model, but of a metamorphosis. “At risk” means: these expenses are under pressure to redistribute – not that they disappear.
The second number to put next to it
If you only read the 234 billion, you get a skewed picture. The same analyst firm predicts that over 40 percent of agent AI projects will be discontinued by the end of 2027 – due to escalating costs, unclear business benefits and inadequate risk controls. Gartner also describes a phenomenon called “agent washing”: Of the thousands of vendors who got the “agentic” label, Gartner only thought about 130 were actually substantial.
Both statements come from the same source, and they do not contradict each other. Technology is changing the economics of software – and the majority of today's projects fail for craft reasons. It is precisely in this tension that the task lies.
Where German SMEs really stand
The contrast with the analyst headlines is illuminating. According to KfW Research, nowadays about 20 percent of medium-sized companies AI This is five times higher than in 2016-2018 and corresponds to about 780,000 companies. For companies with more than 50 employees, it is 36 percent. For companies with 20 or more employees, Bitkom also uses 36 percent AI. The most frequently mentioned hurdles there: legal uncertainty and a lack of know-how, each over 50 percent.
This means in plain language: While Gartner talks about the redistribution of software billions by autonomous agents, the typical medium-sized company is currently using AI productively at all. Anyone who concludes that the topic is still far away, however, draws the wrong conclusion – because the decisions that are pending today are procurement and contract decisions, not technology decisions.
What exactly follows
- Check license agreements for agent suitability. Many SaaS contracts prohibit access by automated third-party systems or bind it to additional licenses. If you want to set an agent on your CRM in two years, the contract today decides whether you are allowed to do so.
- API capability becomes the selection criterion. So far, software has been selected for its surface. For agents, it counts whether there is a documented, complete interface – or whether the function can only be reached by click of the mouse.
- Keep process knowledge in the house. The more of your processes end up in the automation layer of an individual provider, the more expensive the exit becomes. That is the lock-in question of the next few years.
- Start small, but measurable. The 40 percent drop-out rate does not arise from poor technology, but from projects without a clear business case. An agent who takes on a concrete, countable operation beats every platform strategy.
Conclusion
The 234 billion is a forecast about software markets, not a guideline for SMEs. What follows for medium-sized companies is more unspectacular and important: Who buys software today, should judge it according to whether an agent can serve it later. And those who experiment with agents today should choose a process whose benefits can be calculated in hours or euros – otherwise the project statistically belongs to the 40 percent.
How AI agents can be proven and set up with your own data, we describe on our page for AI Consulting.
Sources: Gartner press release dated July 1, 2026 (“$234 trillion in Enterprise Application Software Spend Is at Risk from Agentic AI”), Gartner project termination forecast dated June 25, 2025, KfW Research (Focus on Economics No. 533, February 2026), Bitkom company survey. As of 6 August 2026.
Would you like to implement this in your company? We support you pragmatically – from the idea to the operation.