AI Agents: Where we're going, we don't need to click anything
It's a no brainer really. But we are geniuenly witnessing an mini-paradigm shift unfolding before us in AI. We're clearly moving past the basic chatbot era and heading toward much more sophisticated Agentic AI era.
The Shift from Reactive to Proactive AI
When I look at AI Agents, I see they're giving traditional software a sense of "agency". What do I mean by this? This means the software isn’t just waiting for me to ask a question or fill in a dropdown anymore. In other words, Agents, can proactively suggest the next step or rote sequence in a workflow, moving us from that reactive "chatbot phase" to the proactive "task-fulfillment phase" of AI. Less in and out, more in and sideways, up and down - kind of thing. If this makese any sense?
Now, in theory, this transition should bring broad productivity gains because Agents move us toward a dynamic system or ecosystem of sub systems all working to find the best solutions to evolving problems. Software is no longer static.
Clearer Monetization and Diminishing ROI Debates
I believe the markets are currently poised for Semiconductor leadership to give way to the Software/Application layer accordingly. That's my hunch right now. I may not be right here, but from what I can see, this Agentic AI rise is expected to lead to clearer monetization and consequently, a diminishing focus on debates over ROI. In 2026, I think this means less CAPEX and more OPEX. The pressure will be on for many, specially zombie companies still hanging on.
We are already seeing real examples of this monetization trend. For instance, I read that Salesforce is closing numerous "Agentforce" deals. Using their Agents, customers can drive a reduction in the price per service call from about $6 per call to around $1. That's big. Same goes for SAP who mid year were also targeting a 300 million cloud user base with AI Agentic services, which could generate hundreds of billions in customer savings long-term. So, even capturing a small part of that opportunity means big revenue potential for SAP. On the mid-cap side of things, even companies like RELX are seeing new tools like Protégé, a next-gen AI Agent, continue to drive its transition toward autonmous analytics and decision tools.
Where We Go Next
I find the conceptual roadmap for AI development fascinating. Model capabilities are accelerating rapidly, moving from reliably supplementing 5-second or 5-minute tasks to providing solid output for tasks lasting between 15 minutes and 1 hour. The newest models, focusing on inference-time reasoning, are moving us toward the potential for automating tasks that take 5 hours and eventually 5 days.
And the buck doesn't stop there. Gartner estimates that by 2028, Agentic AI will be included in 33% of all enterprise software applications, up from less than 1% today. They also estimate these agents will allow 15% of day-to-day work decisions to be made autonomously.
On the implementation side of the equation, the length of software engineering tasks that AI can autonomously perform is doubling every 7 months.
Where could we go from here? I am not sure yet, but I believe this will evolve in 3 ways:
1) I am confident that this shift is going to redefine how we interact with software, moving us from passive response to true task-fulfillment. 2) 2026 will be a pivotal year for the widespread integration of AI Agentic capbility into the economy, according to conservative extrapolations of exponential performance trends and 3) I am confident that this shift is going to redefine how we interact with and build software ecosystems, moving us from passive interactions to true "autonomous task-fulfillment", requiring completely different approaches to design thinking and human interaction methodologies.