Five Major Predictions Shaping 2026
It’s that time of year, so I’m breaking out the crystal ball with my tech predictions for 2026. Just a warning, my mother was a famous astrologer, so I relish this just a little too much. Of course, many of these are AI-centric, but there is one in which AI is subtext, not the dominant theme.
So with no further ado, here are my top five predictions for 2026.
1) AI Skill Testing Arrives
Debate all you want, one thing that’s become clearer is that people who use AI are faster than those who do not. Employers are increasingly demanding that job applicants possess AI skills, perhaps their best way to bring their workforce up to speed with new productivity methods. And while many claim to have such skills, some just know how to ask basic questions of ChatGPT.
Moving forward, employers will increasingly ask candidates to take AI tests to ensure skills are present, and may hire the most proficient AI users. This was not just my thinking, but also Gartner, which made it their top prediction of the year.
Gartner anticipates 75% of employers will demand candidates take AI tests, but I do think that’s only true of medium and large enterprises that can purchase third-party testing, such as Codility, for developers.
Further, as time progresses, employers will demand continued testing as part of ongoing employment. After all, AI evolves quickly than any technology we’ve seen in recent memory. Yesterday’s prompt engineer may not adapt well to today’s increasingly nimble and agentic LLMs and AI-infused SaaS platforms. Expect employers to require ongoing skills development, including training and certifications.
2) Agentic AI Catches Up to the Hype
So many LLM-based AI workflows are triggered manually, disaggregated in flow, and require human connectivity. Wasn’t the promise of agents that they would reduce the rote nature of managing cojoining and linear tasks? We’ve seen a lot of hype and many failures. Yet big tech cos keep touting all the agentic work they have already automated (and the people they claim they no longer need).
So what gives?
Time has given developers the ability to work through data integrations via the MCP protocol and refine automations with LLMs to create better and more polished AI agents. The result is the rise of new user-driven, agentic-like flows in Google Workspace, ChatGPT, and Anthropic, new platforms that have taken on the risk of competing with larger platforms, and software-specific AI integrations that automate previously tedious tasks.
