Will Your Next Developer Be an AI Agent?

JetBrains is a Business Reporter client.

The future of business will involve agentic AI. Already, most companies, some 93 per cent, are planning to trial agents for software development. 

However, even as the use of AI rises, trust in AI agents is declining, in part because businesses face considerable challenges with their adoption. For example, only 27 per cent of organizations believe AI agents will become fully autonomous, down from 43 per cent in 2024. Trust is a critical barrier that must be overcome. 

The difficulty is that AI technologies are growing faster than businesses can manage. Models are getting smarter every month, and this makes it hard for organizations to adopt new solutions. Some companies have been successful with AI, but many others (up to 40 per cent) are rolling back their investments, particularly in software development. 

This is changing the profession of software engineers. However, at JetBrains we do not believe that the use of AI in code generation will result in developer jobs being lost. Code is not the only thing that developers deliver, being something they only spend about two hours a day on. Instead, this change will open up even more job opportunities, allowing developers to focus on bigger and more creative tasks than just coding. 

Because of this we have developed Junie, a coding agent that can execute tasks such as code generation, testing and verifying automatically. We did this because we are certain that agentic AI will transform the software development lifecycle and the process of coding. However, there are several possible scenarios that we see as being strongly possible. 

Scenario 1: the developer experiments with AI 

Developers trial their use of agents. They delegate specific tasks: for example, adding a new sorting option on the e-commerce store or even creating a new app from scratch to an agent. This saves the coder substantial time but still requires proper review and control. 

Many developers are experimenting with AI agents, with varying results. Some of them are recording substantial productivity gains, but others are seeing little value in AI. 

Success depends on the tasks given to the agent. Some developers adopt a cautious approach, using AI in “safe to fail” tasks such as prototyping, where the business will not suffer if the AI fails to deliver quality outputs. Other programmers have been more ambitious, attempting to use AI agents throughout the entire software development lifecycle, but they are often met with organizational obstacles to adoption. 

Success in this scenario involves harvesting the low-hanging fruit in code generation as a means of building knowledge, while avoiding undue risks. 

Scenario 2: the developer leads a team of AI agents 

Developers manage a “team” of AI agents. Some of these have specialist capabilities such as code generation or testing. Others have wider capabilities and can be assigned to different tasks. 

In organizations where there is confidence in the ability of AI agents to add value, the developer’s role changes from that of an individual to that of a team leader who assigns different agents highly defined tasks. These may vary, depending on the agent, from simple ones (such as changing button colors in the interface) to more complex ones (rewriting code from one language to another, for example). 

By delegating some tasks to AI agents, developers are freed up to work on multiple tasks simultaneously. Productivity gains are seen in the whole engineering process and not just in code generation. 

However, this scenario comes with some important challenges. Some developers will be unwilling to delegate to a machine tasks that they themselves previously undertook. They may not trust them, or they may be frightened of the impact on their jobs. There will therefore be cultural obstacles to overcome. 

Scenario 3: building an autonomy island within your business 

Businesses reorganize the software development lifecycle to involve other departments, such as sales and compliance. The full workflow is automated, not just individual tasks. We call it an autonomy island, a low-risk area of your business that can be fully co-developed with an agent. They will “communicate” with each other – sometimes even without the oversight of humans. 

In this scenario, companies rebuild the software development lifecycle so that AI agents are fully integrated at all stages. The degree of autonomy agents are given will vary: where it is safe for them to do so, they may operate with little intervention from humans. At other stages, humans will have more control. 

This scenario goes beyond the use of AI agents by the software engineering team. Other parts of the organization, such as marketing, sales and compliance, may be involved, and experimentation can take place anywhere within the organization. The task delivery will be fully executed by the agents – connecting design, minor copywriting, software delivery, marketing, and other functions. 

Imagine the situation when adding a website page – or even publishing this article – can be fully done by the agents. Once the text or interview is written, the agent will design it, program it, test the page, publish it, advertise automatically and send the report to you. We already have the agents and technology, but the organization needs to connect them and make it a common practice in the team. 

Building success with trust 

These three scenarios represent different approaches to AI. Whatever scenario an organization selects, leaders need to trust that AI will deliver positive value. This may not only be limited to coding efficiency: it could involve greater speed, stronger security or better functionality. Understanding that AI delivers more than just cost savings is essential. 

Cyber-security risk will be central to trusting AI. Organizations must proactively search for threats. To that end, having some form of human oversight will be essential wherever AI is used. And employees must be trained so that they avoid the danger of subordinating their own knowledge to computer-generated outputs. 

In addition, developers need to trust that AI will have a beneficial impact on their careers. They should accept that their roles may change, and that more of their role will involve managing and validating outputs from AI agents. But they will also need reassurance that their roles will not be diminished. 

Humans and AI agents working together 

AI agents have the potential to deliver substantial value across business, especially in the software development cycle. When working with AI agents, developers will have roles that go beyond coding, including a responsibility to understand the wider business. At the same time, they will need to manage the outputs of AI agents. AI is not a magic wand, and the role of the human is always going to be central.

JetBrains helps developers increase productivity, manage quality and safety and keep AI under control, which is why over 11 million developers worldwide trust us with their code and business.

This article originally appeared in Business Reporter.

Header image credit: iStock-2201418920