Artificial intelligence (AI) is penetrating ever deeper into personal daily life and corporate workplaces. Some argue more verification is needed on whether it boosts productivity, but companies and the government remain in attack mode on AI. A growing number of firms are even changing their names to include AI or AX (AI transformation). Major companies are declaring themselves “AI native,” beyond “AI first,” and challenging new business models. To mark its 19th anniversary, DigitalToday has compiled the latest AI trends and AX issues unfolding across industries. [Editor’s note]

[DigitalToday reporter Chi-gyu Hwang] AI agents that handle specific tasks without human intervention have already been seen as a major variable in the tech world. Lately, they appear to be cementing their place as a key word driving shifts in the AI landscape and trends. Major tech companies are accelerating moves to restructure their business models around AI agents.

The origin of the AI agent boom was, as expected, coding. Since the emergence of ChatGPT in late 2022, coding has been a killer app for generative AI. It has also shown the strongest fit with AI agents, emerging as a stage that directly demonstrates innovation in AI agents.

Competition among companies is also intense. Anthropic’s coding AI tool Claude Code is growing rapidly on the back of agent functions. Rivals, spurred by that, are investing large sums to respond. The “agentic coding” craze, in which AI agents code instead of humans, is intensifying.

Claude Code, released last year, sparked a craze among developers, entering a run rate of $1 billion in annualised revenue in less than a year. Anthropic had $2.5 billion in annual recurring revenue as of March and more than 300,000 corporate customers from Claude Code. It also secured an image of leading the agentic AI trend rather than being the No. 3 player behind OpenAI and Google in the AI field.

As Claude Code gained huge popularity among developers, competitors such as OpenAI and Cursor also put agents at the forefront. OpenAI substantially strengthened AI agent capabilities in its coding AI Codex and set pricing cheaper than Anthropic. Cursor, which has grown on a code-editor base, also recently added a large number of agent features through updates.

Moves by various companies seeking to expand agentic AI beyond coding into other areas of work are also accelerating. Efforts are active to put AI agents into real-world use in workplace applications and e-commerce as well.

Anthropic introduced Claude Cowork, an AI agent tool for non-developers. Microsoft unveiled Copilot Cowork, an AI agent function, in its productivity platform Microsoft 365. Copilot Cowork supports AI agents in handling long-running, multi-step tasks that had required constant human oversight. When users state the results they want, the agent makes a plan and carries out the necessary tasks in sequence.

Other major enterprise software companies, including Salesforce, are also rapidly reshaping their product lineups around AI agents, a trend that is also leading to major changes in enterprise software user interfaces.

◆ Rapid reshuffle in enterprise software

Discussion around AI agents is evolving beyond changing the enterprise software experience to predictions that they will dismantle the existing industry order and reshape the sector itself. A scenario in which AI agents replace enterprise software altogether, dubbed the “SaaSpocalypse” narrative, has gained some traction in the tech world.

The crux of the SaaSpocalypse is that as more companies make their own software using AI rather than buying or subscribing, the footing of existing software vendors weakens. Many view it as an “overreaction” that will be hard to realise, but the assessment is that it is at least resonating among investors. Despite the AI agent boom, shares of major software companies have been falling this year.

Even if the SaaSpocalypse itself is an overreaction, it is not easy to dismiss the case that relative undervaluation of software companies lacks any basis. As companies gain other options, a realistic risk is that software firms may find it harder to grow revenue than before. Recent share prices are not unrelated to that.

AI is already having a significant impact on software business models. In particular, as more companies change how they set prices to fit AI, pricing is spreading that focuses beyond the number of users to usage, and further to outcomes.

The narrative that AI could replace or significantly disrupt existing tech products is spreading beyond software to other fields. Recently, a new AI model under development by Anthropic, Claude Mythos, was reported to be strong at finding security vulnerabilities, in a development that also saw shares of major security companies fall.

◆ Profitability still unclear; focus on whether innovation spreads beyond coding

AI-related companies are growing faster than those in other fields, but uncertainty remains on profits. Companies that provide infrastructure needed to develop or run AI are making big money. Companies that sell AI are still unable to escape losses, as before.

OpenAI and Anthropic, both seen as representative global AI startups, are preparing for record initial public offerings this year, but criticism persists that their internal structures are weak. Surging costs required to train new AI models are also cited as an Achilles’ heel.

According to a Wall Street Journal report, OpenAI expects to spend $121 billion on computing power needed for AI research in 2028. Even if revenue nearly doubles in 2027, it could burn $85 billion in cash flow in 2028. It is a structure in which money going out still far exceeds money coming in.

Anthropic does not appear likely to spend as much as OpenAI, but it is in a similar situation in the narrative of rising computing costs. The two companies are releasing new versions of AI models at a faster cadence than ever, and there is no sign this race will slow.

It is still unclear whether areas where people can feel AI productivity will expand beyond coding. Tech companies are trying to apply the agentic AI success formula that worked in coding to areas outside coding, but it is uncertain whether it will lead to changes of the same scale.