Software companies no longer see AI assistants as experimental extras. Now, they are building them in as a standard part of their products.
According to GitHub’s Octoverse 2025 report, “AI, agents, and typed languages” are behind the “biggest shifts in software development in more than a decade.”
The report also notes that over 36 million developers joined GitHub in a single year, and public projects using generative AI model SDKs reached more than 1.1 million repositories by 2025.
This matters for two reasons: AI is becoming a normal part of software development, and software vendors now have a strong reason to encourage users to interact with AI directly, rather than just letting it run in the background.
Why companies keep adding assistants
The business reasons are clear. In May 2026, ServiceNow announced it aims for “$30 billion-plus in subscription revenues” by 2030, with ServiceNow AI expected to make up “over 30%” of annual contract value.
They describe their platform as one for “governed, autonomous work.” Adobe took a similar approach with its Firefly AI Assistant, saying the tool brings creative software into a “single conversational interface” and can “orchestrate and execute complex, multi-step workflows” across Photoshop, Firefly, Premiere, Express, Lightroom, and Illustrator. In both cases, the assistant is not just a novelty.
It is meant to make big, complex software easier to use and to create a premium feature that companies can charge for.
Why the pitch is attractive to users
This approach works because modern software is powerful but often hard to use. Adobe’s Firefly AI Assistant announcement highlights this: creators can “simply describe the outcome they want,” and the assistant takes care of the steps.
This is attractive because most users do not lack ideas—they struggle with confusing menus and the hidden skills needed to use advanced tools well. The assistant acts as a shortcut through all that complexity.
So, software companies are not just following AI trends. They are also trying to solve a real user experience problem: many digital products have become too complicated for most people to use easily with traditional interfaces.
When assistants genuinely save time
There is strong evidence that assistants help when tasks are clear and users need support. In the NBER working paper Generative AI at Work, Erik Brynjolfsson, Danielle Li, and Lindsey Raymond studied 5,179 customer support agents using a generative AI assistant.
They found that the tool increased productivity by 14% on average, and by 34% for beginners and less experienced workers. It also improved customer satisfaction and employee retention.
This shows that assistants are most useful when they guide users through best practices they do not already know.
In short, assistants are valuable when they make learning easier, not just when they generate text quickly.
Why so many still feel intrusive
The problem is that simply adding a chat box does not make an assistant useful. A 2025 study on arXiv, “My productivity is boosted, but …” Demystifying Users’ Perception on AI Coding Assistants, found 1,085 AI-related extensions among 66,053 in the VS Code Marketplace, with over 90% released in 2023 and 2024.
This shows how fast the field is growing. But the same study found that users want assistants to be “context-aware, customizable, and resource-efficient,” which is why many tools still do not meet expectations.
Reviews about “context awareness” were more negative than positive, with only 38% rated positively.
The authors note that assistants can understand code when they have enough context, but often struggle to “retrieve and maintain relevant context.”
This explains why many assistants seem only partly helpful—they sound confident but do not always understand what the user actually needs.
Clutter is also a performance problem
That same arXiv study, “My productivity is boosted, but …” Demystifying Users’ Perception on AI Coding Assistants, shows that clutter is not just a design issue—it is also a technical one.
The authors found that developers may spend over half their coding time checking AI-generated suggestions, which adds to their mental workload instead of reducing it. They also found that resource use is a common complaint: 78% of related reviews were negative, and some users said assistants used “over 50% CPU and more than 1 GB memory.”
This is where the idea of putting an AI assistant everywhere starts to fall apart.
If the assistant interrupts work, uses too many resources, slows down the editor, or makes users check too much low-quality output, then the product is not really simpler. It just adds another layer to an already complex interface.
The trust problem companies underestimate
There is also a deeper issue: assistants do not just save time—they influence how people make decisions.
In the 2026 paper Adjust for Trust: Mitigating Trust-Induced Inappropriate Reliance on AI Assistance, researchers found that when trust was high, medical doctors mistakenly accepted 26% of AI misdiagnoses, compared to 8% when trust was lower.
When trust was low, doctors rejected correct AI diagnoses 68% of the time, up from 40% otherwise. The same paper found that explanations, counter-explanations, and forced pauses could reduce inappropriate reliance by up to 38% and improve decision accuracy by 20%.
This lesson goes beyond medicine. An assistant is only valuable when users trust it the right amount. If it leads to blind trust, it can be dangerous. If it feels unreliable, users will just ignore it.
Speed now, maintenance later
Even when assistants seem helpful at first, they can shift work to later.
The 2026 paper Debt Behind the AI Boom: A Large-Scale Empirical Study of AI-Generated Code in the Wild looked at 302,532 verified AI-authored commits from 6,299 GitHub repositories using five popular AI coding assistants.
It found that more than 15% of commits from each assistant introduced at least one issue, and 22.7% of these AI-related issues were still present in the latest version of the code. The authors say that AI-generated code can create “long-term maintenance costs” for real software projects.
This helps explain why companies want to add assistants, but users often have mixed feelings: assistants can make things easier at first, but they do not always reduce the total work that needs to be done.
So why are companies still doing it?
The reason is clear: the benefits are hard to ignore. Assistants help people get started faster, make complex tools easier to use, and fit well with today’s focus on AI.
Built-in assistants save time when they are focused, truly context-aware, and help users manage complexity. They become clutter when added just for branding, when they use too many resources, when their output needs lots of checking, or when they encourage trust without being reliable.
That is why every software company wants one now. An assistant is the easiest way to show a product is modern. The real challenge is making it truly helpful, not just always present.
