2024 was the year of AI assistants and this year 2025, is going to be the year of AI agents 

 

In 2024, some of us used AI assistants based on Large Language Models (LLMs) to handle specific predefined tasks for us, such as answering questions, summarizing content, generating images, software codimg and providing recommendations.  

 

Many of us used prompts - typed or spoken - to get answers to our queries from ChatGPT(OpenAI), Claude(Anthropic), Gemini(Google) and other LLM powered assistants.  

 

The recent "reasoning" models are the latest iteration of this assistant technology. 

 

But this year 2025, AI agents which will take things further, and are now being developed, tested and released. 

 

Agents operate independently with minimal or no human supervision and handle multi-step, complex workflows, problem-solving, and dynamic decision-making. 

 

One such agent is Operator, an agentic system recently limitedly released by OpenAI, the company behind ChatGPT. 

 

Operator works by - 

• taking screenshots of a web page and analysing it thereby "seeing and reading" the screen 

• it can then execute online tasks such as booking restaurants, ordering groceries, and purchasing tickets 

• it navigates websites, selecting, typing and clicking, without you the user making the input (like invisible hands using your keyboard and mouse) 

• it asks clarifying questions when needed, to complete tasks accurately 

• it returns control to you for sensitive actions like logins and purchases 

 

So instead of you asking Operator about where the next Taylor Swift or Beyonce concert is, you can ask it to buy you a ticket for a show at a time and place that it knows your schedule is free. 

 

It is technically interesting that OpenAI has chosen to use such a general user interface system (one that actually reads and types into the screen) rather than one that depends on code, to interact with its system.  

 

It is possible that this choice will make taking it up more open to a wider spectrum of adopters, but like most things AI at the moment, this remains to be seen. 

 

So, the AI train keeps rolling on, gathering speed.  

 

This revolution is not one which the older generation can sit out and let the younger generation take the lead with. Everyone must get involved and work through the changing realities - No, you can’t sit this one out. 

 

Register for ChatGPT today and have it open in a browser and ask it questions all day. Install the Gemini app on your phone and talk to it all day. Use Copilot in your Microsoft based applications to help you write, summarize, understand, create! 

 

Finally, let's look at two use cases which are 100% possible to implement right now given the current state of technology - even these will be overtaken in a matter of months. 

 

Hybrid help desk 

Companies could handle a lot more queries and reduce the cost of query resolution by following the steps below - 

• Sample the voices of their human operatives. 

• Then using the AI generated voice of each human, speak with and answer callers on maybe ten channels or phone lines with the same voice of the single human. 

• AI (what else) will monitor the "health" of each interaction and if any of the AI channels is not going well, the real human can quickly scan the record of the call and take over call in seamless manner. 

•  

This will make it possible for a single operator to "handle" up to ten calls at a time. 

 

Reversing offshoring 

Many companies offshore a good part of their technical and software development work, on the basis of cost, while keeping higher value technical tasks - envisioning, architecting and designing - onshore. 

But AI is rapidly getting to the stage where it will be able to do the bulk of the software coding, reviews and testing that is the raison d'être for moving work to lower cost jurisdictions.  

 

Soon it would make no sense to introduce the communication and co-ordination challenges resulting from doing a substantial part of software development work, thousands of miles and hours of time zones away. 

 

This will mean that a lot of work implementation will return to the US and Europe.  

 

In conclusion, the game is changing, the goal posts are shifting - you pick your metaphor. 

 

I like to think that this is going to be like the invention of the printing press - since books did not need to be copied by hand, knowledge became more widely available and there was a knowledge and productivity explosion which benefitted all of mankind beyond measure.