How AI will enable the next generation of business software
Y Combinator, one of the world’s most well known startup accelerators, regularly updates a wishlist of products that they hope someone will build. One of the requests in their current ranking was an “AI to Build Enterprise Software”:
“What if [...] you just give customers a simple starter product and have them tell your AI how they want it customized? In the future, every enterprise could have their own custom ERP, CRM or HRIS that is continually updating itself as the company itself is changing.”
There is tremendous value for businesses in having tailored software that doesn’t require any coding or complex configuration. Furthermore, AI can help accommodate the ever changing needs of businesses. Picture this - you spend weeks setting up the perfect CRM. Suddenly, your business model changes, you expand your team or need to address and adjacent use case, like project management or a customer portal. Instead of starting from scratch, AI can adapt your system to your needs without any delay.
This next generation of software will empower non-technical people and resource-constrained organizations to reap the benefits of tailor-made software. I will discuss this in more detail in my next post.
Here, I am eager to provide Skippet's view about the optimal approach to creating the fully adaptable AI-powered software.
The AI will only be as capable as the underlying platform
AI clearly was the missing piece of the puzzle for creating this type of software. It enables the computer to make the right decisions on behalf of the user. There are some nuances in how you apply AI to yield optimal results for this type of problem, that largely involve working around limitations of current AI models.
More importantly, however, the underlying AI models alone are not the key ingredient in building this self-adapting flexible software of the future. This is because the AI is only as capable as the tools that it can leverage.
In the extreme, you could unleash the AI on code. The current state of these 'co-pilots' for code are not at the level they need to be to have a non-technical person be confident that the output is what they want (they take a lot of hand holding still). In the future, creating apps from scratch with AI may technically work but will still be highly problematic for the non-technical user. Thus you need some abstracted platform where those applications are built.
Platform accessible to anyone
The platform where the AI crafts applications should be one that can be configured without any code and thus accessible to technical and non-technical people alike. This is for two reasons:
- Anyone should be able to understand the AI’s actions. After all, who would store their company’s sensitive information in a tool that was generated on the fly in a way that they don’t understand?
- Anyone should be able to make manual changes (or more precisely prompt the AI) to initiate those changes. Human language is inherently imprecise and thus corrections will always be needed. Using AI to iterate on your system is an area that we are actively working on in Skippet. If something is too laborious for someone to describe it should still be possible to change it manually.
There is an added benefit - a platform that a user understands, is also likely one that the AI will have less trouble handling. Computer and human intelligence hold some similarities!
It should also be noted that, although language may be an intuitive way to describe changes you wanted to implement, it is clearly not the ideal way to experience a system like this. That is to say, the generated applications must be intuitive to use, and having a platform which takes care of this by design is helpful.
Flexible platform
The platform also has to be extremely flexible for the AI to actually be able to accommodate the needs of users. This means enabling custom business logic as well as a multitude of formats for storing and visualizing data. As AI gets better this flexibility will become a competitive advantage over other solutions which are be constrained.
There is an inherent tradeoff between flexibility and complexity. Incumbents have been operating under certain assumptions of where that trade-off lies to make sure that their products are usable. While AI doesn’t completely remove this trade-off, it alleviates it substantially. Though the interface of a flexible/complex system is by nature more difficult to learn, with the help of the AI to setup complex workflows, the output just needs to be understandable to the end user so that they can verify and make adjustments
This creates a window of opportunity for new companies to disrupt the market.
Looking into the future
It is evident that despite the advances in AI, creating the next generation of software is harder than it may seem. Yet the opportunity is too great to miss out on. To see a glimpse of the future check out Skippet for yourself. We have lots of exciting updates coming both to the AI and the platform which will bring us closer to our vision of empowering humans to achieve more with computers using natural language.
