Will AI replace software?

Recently," AI will replace software This not only caused an unprecedented drop in software stocks, but also plunged the entire market into deep worry after superimposing concerns about AI profitability. If there is a reason for concerns about the profitability of AI, after all, the cost of AI has been high, and the profit model is still through traditional means such as advertising and recommendations. However, it is somewhat overreacted to the judgment that AI development will subvert the software industry.

The background of this discussion is the recent emergence of a series of landmark AI applications. First of all, the sudden popularity of clawbot allows users to build personalized AI systems on local devices, thereby realizing automated operations and process execution, and to some extent has the function of "personal assistant". Subsequently, Anthropic released cowork, further enhancing the Agent's ability to perform complex tasks and demonstrating the potential of AI in cross-process collaboration. These developments are a signal that AI is gradually moving from auxiliary tools to system applications with execution capabilities. In the near future, it is completely foreseeable that AI will integrate personal devices and data resources, and users only need to use instructions to let AI call various processes to complete the required work, from the current auxiliary "secretary" role to a "practical worker".

One of the more critical issues is that some software may no longer constitute rigid requirements. In AI-driven workflows, the system relies more on callable segmentation modules. For simple scenarios, AI can even dynamically generate or temporarily build the required capabilities. Looking at it this way, this will indeed be a big blow to the software industry.

What impact will AI have on software?

Before jumping to conclusions, let's take a look at the substantial impact that the development of AI will have on the software industry.

Software development just got easier

I don't think anyone will question this anymore: it used to take a team a long time to develop a software, but with AI, only a few senior engineers need to do architecture design, code integration and final debugging to complete a simple software, like clawbot is done by one person (although the analogy here is not entirely appropriate, it can be seen that AI has greatly improved a programmer's development ability). Although AI cannot write a very complex program (and is unlikely in the next few years), it is completely feasible to cut the program into small parts and then have experienced engineers debug and integrate. This greatly shortens the software development cycle and greatly reduces the cost of software development.

What impact will this have?

First of all, plug-ins, tool software programs, most of these programs are charged monthly, may disappear completely in the future, because it is very simple to let AI write a plug-in, and AI can also be customized according to the specific situation, which will be more in line with its own business logic than ready-made plug-ins on the market.

Secondly, there will be more and more replacements in large software, although it is not realistic to write a large software completely, but some functions can be realized by AI, so that the achievement of the goal does not need to rely entirely on a specific program. This makes some lightweight users consider other alternatives.

The charging model is subverted

Now almost all SaaS billing models are monthly billing + capitation billing, and this model may not continue. As mentioned above, simple tool software will lose user dependence, and the number of users who use it will decrease. So the monthly billing model is not cost-effective for users, originally this model is very unpopular, and now there is finally a chance not to use it, so why not run to tell each other?

Secondly, it will become more and more difficult to charge per capitation, because AI has improved work efficiency, and the original work of dozens of people now only requires a few people, and it can even be done with an AI "head", and all others use this AI machine/Agent to use the entire set of software. In this case, software companies need to change their business models and find other ways, such as charging based on function and frequency (similar to tokens).

Software composition reshaped

Nowadays, many large software is very functional and all kinds of interactions are also perfect. But after AI matures, users may only need some of the features of these software. This is because AI breaks down the work and plans and implements a workflow in the most economical way. For AI, large software consumes different resources than small functional modules, and he will inevitably choose low-consumption processes, which will force large software companies to modularize their software to retain heavy AI users.

Although many software functions are very complete, most users actually only use some of the functions, and most of the existing charging strategies are formulated around these popular functions.

Interaction mode changes

Now the interaction mode between software and users is through the front end of the software, while AI only needs to call APIs, which is completely different. Many complex software often requires a lot of time and effort to design and develop the front-end, and even different customer types will have different designs, which is one of the reasons why software is difficult to replicate. Users don't see the front end of the software, and those complex buttons and menus are gone.

APP Store？

In the foreseeable future, AI can become an important entrance to user interaction, just like Apple's iPhone opened the APP era. So, will there be a similar pattern to the App Store? I think it is very possible that if AI becomes an entrance, then all software that can be called through this AI needs to be licensed by the platform, and users also need to purchase the software/functions they need through the platform to integrate an AI work platform that suits them. In the future, several large AI platforms will have high bargaining power, and may even charge a "tax" on the software on the shelves (refer to the Apple tax), which will further compress the profits of software companies.

Advantages of existing software

For complex and changeable large-scale software, whether engineering or industry software, it is almost impossible for AI to imitate and replicate. What AI is good at is that the user gives it a very clear goal and asks him to disassemble it and complete it one by one, but if it is a complex process and environment, it is difficult for the user to even accurately describe all the situations, and it is difficult for AI to help.

This is also the biggest advantage of most SaaS companies. Of course, users can disassemble the entire process into a single task by themselves, and then let AI perform it step by step, I believe this is not difficult to achieve, but in this way, the user's time cost will increase, and after weighing the pros and cons, users may still consider using professional software. To explain with a practical example, for example, the annual tax return, the forms are public, and how to file taxes is also very clear. But people are more willing to spend money on software or find accounting than to do it themselves. For AI, tax filing is a simple single task, I believe that AI can be replaced in the near future, but it will be a complicated thing to sort out all your tax forms and hand them over to AI, and here software companies have an advantage, if they split the functions to AI, or integrate AI by themselves, it will greatly improve the competitiveness of the software. Tax filing software + AI sells you 50 yuan, or do you organize the materials yourself and teach AI to do it step by step? And after you let AI file taxes independently, who is responsible for the results, wouldn't it be more troublesome if something went wrong?

Therefore, software companies may be more powerful and more convenient for users to complete tasks because AI is more powerful. Build a ship by yourself, I may still be at ease, but to build an aircraft carrier, no matter how similar the appearance, I will still worry about whether the interior has been perfected, with this mind and time, it is better to spend some money to solve it, in essence, the software company is to share the cost of using the aircraft carrier together. This is also the idea of many SaaS companies now: I know you can solve the problem yourself, but I can make you more efficient and free up your time to do your core business.

Will software be replaced by AI?

Obviously the answer is: no. Thinking carefully about the above analysis, the impact of AI on software companies is huge, but the core is still that as long as AI cannot develop complex software on its own and quickly, it will always need software to work for it. It is almost unrealistic for AI to develop complex software, because its understanding of complex workflows and the human world is still far from it, which is difficult to make up for through its own learning. If AI can do this one day, I believe we will face a much bigger problem than software replacement.

Software will not be replaced by AI, it does not mean that all software will survive, it subverts the profit model, usage scenarios and business logic of the software industry. The moat of SaaS itself is not the difficulty of software development, their real value is to handle complex processes that AI cannot do in the short term. And those small software, especially tool software, will be more impacted.

To put it another way, we are talking about ready-made software now, and software companies themselves have a strong technical team and development capabilities, won't they embrace AI and make their software an AI entrance? Make the interaction mode that originally needed to find buttons and click buttons become communication with AI? Will this reduce the cost of software use and learning, so that users will rely more on the software itself? For example, Photoshop is extremely expensive to learn, but users can clearly depict their goals, so should we consider designing an AI entrance, whether it is desktop software, page side, mobile APP or even cloud, as long as the user gives instructions, he can complete the modification of the picture? For heavy users, a semi-automatic mode may be considered—tasks that can be clearly explained are handed over to AI, and those that cannot be clearly operated by themselves. Wouldn't this be more attractive than a fully automated model that was completely handed over to third-party AI?

In summary, AI completely replacing software is a fantasy, but the impact on the software industry exists, he will reshuffle the entire industry, and will also subvert the mode of user use of software in the near future, and those companies that can adapt and change will not be greatly impacted, and it will also be a great opportunity for small software companies.