If 2023 was the year when ChatGPT reigned, and 2024 was when a thousand LLMs bloomed, 2025 will be the year of AI Agents. The first article in my troika of predictions for 2025 focused on how English will be the new coding, AI the new UI, AI and humans the new creators and AI creating a new customer (https://bit.ly/4fB6XBm ) The second one tackles Predictions 5 to 8, and the focus is on Agentic AI, or when AI is given ‘agency’ and the power to do action by humans.

Prediction 5 or 12: Agents are the new Platform: Bill Gates presciently blogged in 2023: “In the next five years…you’ll simply tell your device, in everyday language, what you want to do…and software will be able to respond personally because it will have a rich understanding of your life. This type of software—something that responds to natural language and can accomplish many different tasks based on its knowledge of the user—is called an agent.” (https://bit.ly/3tSMNkB). So, if you plan a vacation today, it means several hours spent tapping across multiple apps before you book a satisfactory itinerary. However, a future booking agent could select a hotel and airline based on your past preferences and pricing, design a daily itinerary based on your known interests, and then proceed to book flights and restaurants, after you have given it the permission and the agency to do so. Sarah Hinkfuss of Bain Capital described this well “We are used to ‘pulling’ information from computers, (AI agents will) ‘push’ finished work to us instead”. (https://bit.ly/3Ye4reU ).

“Agents are the new platform”, Gates declared, and hundreds of startups have heard his clarion call to build agents on top of the LLMs that Big Tech is rolling out – Microsoft with CoPilot Wave 2, Google with Gemini 2, AgentForce by Salesforce, etc. Minday scours the Internet and mines your preferences to find the best restaurant around you, and Relevance AI automates prospect meetings for harried sales reps. The CEO of fintech Klarna announced that customer service agents built on OpenAI platforms have ‘replaced’ 700 human agents.

6 of 12: SaaS is the New SaaS: as Software-as-a-Service gives way to Service-as-a-Software. AI will bundle up discrete steps in a typical enterprise workflow into an agentic AI software. SaaS companies rent out software today, which help humans in enterprises perform some service. These ‘New SaaS’ companies will sell the service instead, enabled by agentic AI. So, if QuickBooks sells a SaaS product today, which enables us to file our tax returns, agentic AI will focus on the end-service, and file our tax returns instead, with minimal humans in the loop. This is a dramatic change, and Foundation Capital touts this as a $4.6trn opportunity, and a disruption to the structure of the industry and the jobs it provides (https://bit.ly/473ByVO ). Bill Gates warned us about this too in the same blog. “Agents are not only going to change how everyone interacts with computers”, he wrote, “they’re also going to upend the software industry.” There has been a lot of hand wringing about Generative AI not having ‘real’ enterprise use case; it is agentic AI that will bring Generative AI to the enterprise

7 of 12: AI are the new Colleagues: Agents will make their way into the workforce too. Today, a team comprises of a bunch of humans, and the team leader’s job is to know their strengths and weaknesses and utilize them accordingly. This will change as a team has a mix of humans and AI agents, and the leader’s job will be to bring in humans for some jobs and agents for some other tasks. Diversity in an organisation will not only be across race and gender, but also across humans and AI, as in the Klarna example above. Jensen Huang and others are already considering bringing ‘millions of AI agents’ in their companies, besides thousands of humans. They are touting the new ‘one person unicorn’, which will a billion-dollar startup, with one founder and the rest of the team being all AI agents.

8 of 12: AI is the new Cloud: The cloud created new business models and sparked an innovation revolution, with startups now able to rent compute rather than pay huge amounts upfront for it. It also created innovative products and a new technology ecosystem. Much like we have three ‘hyperscalars’ building and managing the cloud, I believe we will have three or four large commodity LLMs – OpenAI, Anthropic, Google, Meta – and startups will rent out AI-as-a-service and build innovative startups and applications on top. I also believe that Microsoft will attempt to buy OpenAI and Amazon to buy Anthropic, as the M&A regime loosens up in the US, making the cloud and AI providers the same entity; AI will seamlessly integrate into the cloud as we know it.

I believe that the Agentic AI wave will be bigger than the Chat wave that has preceded it, and it will raise even bigger questions around the human and ethics aspects of AI than now. The last four trends tackle that, and I will talk over those in the last of the three articles next week. So, watch this space, or if impatient, navigate your way to this video (https://bit.ly/4fnaJ1h ) where I talk about all twelve trends at once