AI Agents, the Future of Work and Software Development

For 40 years we've gotten used to clicking buttons and selecting items using a graphical user interface (GUI). This has been the standard way to deliver applications for the software industry for as long as most of you have lived. That's all about to change because of 2 words; AI Agents!

The point about a graphical user interface is that it allows you to "provide instructions to the machine." The computer uses your input as arguments to functions it executes. Such functions can be for instance "send email", "create contact", or "generate invoice." How your GUI looks like depends upon what functions you want it to be able to execute. Today we can have systems where the instructions are provided with natural language. See an example below.

If you were to use a GUI to accomplish the above, you'd have to use your browser to go to TripAdvisor, maybe select a country, then select a city, before you have an answer. For the above example this would be easy, and TripAdvisor have probably spent millions of dollars making it as intuitive to use their GUI as possible - But most of us are dealing with hundreds of different applications every single day. These applications have different GUIs, and learning how to use these apps takes a lot of our time and resources.

With AI Agents you can give instructions using natural language, the same way you tell humans to do things. This eliminates the need for a graphical user interface, and allows you to use your natural communication to "execute functions."

AI Agents becomes a better solution for how we interact with our computers

The GUI becomes Obsolete

If you stop and think about this for a while, you realise that AI Agents are basically making the graphical user interface obsolete. Using natural language to give instructions is simply a superior way. Language is the first thing we learn as we're born, and our brains have evolved over millions of years to be better at understanding language than checkboxes, dropdown menus, and buttons.

In addition to being a better way to deliver applications, it can also do things almost impossible to do using a GUI. Let me give you an example from one of our clients.

I realise the above is possible to describe and create a design for using Figma, but in addition to simply providing a better user experience (UX) for the end user, it's also 1,000 times easier to implement for the developer using natural language.

If you were to create a GUI design for a mortgage calculator such as the one above using Figma, you'd end up with at least 3 input fields, maybe some checkboxes, and at least one or two buttons. When creating an AI agent it's a simple natural language instruction, optionally executing some AI function to do the actual calculation if you don't trust the LLM's calculations.

I have seen UX designers spend months on something such as the above. The above took us 5 minutes to implement

How AI Agents Work

AI Agents are quite easy to understand. The same way your existing systems have a GUI, they also typically have an API. API means Application Programming Interface, and APIs exist such that computer programs can interact with each other. APIs allow one program to use another program.

The program doesn't differentiate between an API call or a human being sitting in front of a computer using a GUI. This allows you to connect an LLM to an API, and have the LLM generate function invocations. If the user provides incomplete information to the function, the LLM can be configured to ask for more information.

In the above example the AI Agent didn't know how to answer my question without asking me for additional information. The AI Agent therefore asks me about a country or city, such that it can scrape the web for weather information, and tell me how the weather will be tomorrow.

This of course results in a far more intuitive way to interact with your computer, and over time it will lend itself to speech recognition and voice synthesis, allowing you to exclusively use voice commands.

As long as your application has an API, or the AI Agent can somehow access its database, it can basically do anything the original app can do - And more!

The GUI burden

Let's use your CRM system as an example here. When you started using your CRM, how much time did you spend learning it? If you're an average human you'd probably answer weeks, possibly months. In fact, there are "CRM experts" on LinkedIn who's sole income originates from teaching others how to use some specific CRM system.

Learning how to use a specific application is sometimes so hard that when companies are looking for new employees they will often ask for things such as.

• Expertise in HubSpot
• Certified SalesForce user
• NetSuite Qualifications

The reason is because all of the above systems have thousands of different forms, checkboxes, input fields, and choices. Learning how to use them all takes years! Learning enough to be productive often takes months!

However, if you replace the GUI completely with natural language, you can take the original thing you wanted to do, type it into a textbox, and click a button. And if you think about it, every single thing you're trying to achieve starts out with natural language. Some examples are given below.

• Send marketing email to all clients about summer sale
• Refund John's latest payment
• Generate monthly revenue report

How many clicks and selections would the 3 above tasks result in if you were using a GUI? With an AI Agent you simply give it your intentions using natural language and tell it to do it. For the record, all of the above are examples of real prompts we have been creating for clients of us.

Wrapping up

AI Agents are destined to completely change the way we work. But they're also destined to change how we deliver applications. Starting out with a complex graphical user interface isn't necessarily given today. Using natural language to deliver AI Agents instead is probably 1,000 times easier and less resource demanding.

In addition, the AI Agent is self describing, completely eliminating the need for documentation. If you want to know what an AI Agent can do, you can just ask it.

This reduces the cost to implement software by several orders of magnitudes - Implying applications that would cost you millions of dollars some few years ago, could probably be done with some few thousands dollars today - Assuming you're happy with a natural language interface of course as an alternative to a GUI.

At AINIRO we've got expertise in delivering AI Agents. If you want to know more about how AI Agents can help you, you can find more information about it below.

• Read more about AI Agents

Have a Custom AI Solution

At AINIRO we specialise in delivering custom AI solutions and AI chatbots with AI agent features. If you want to talk to us about how we can help you implement your next custom AI solution, you can reach out to us below.

• Contact AINIRO