Happy New Year from Robotic Online Intelligence (ROI) - for our clients, contacts, and followers - wishing you healthy levels of control over your automation and LLM deployments in 2026!

In 2025, in regard to AI applications in data/information businesses and investment research, the big theme for us has remained 'making LLMs work in practice'. 

It's been an ongoing challenge to figure out the right balance between the real opportunities from new methods and models, and the over-hyped ideas and the 'enterprise reality' when it comes to high-accuracy scenarios.

Here's the 2026 thesis (for our niche):

1. For now, what works is still the largely deterministic workflow automation, but plugging in the various AI/LLM components, orchestrated by the human experts.

That's at the heart of our approach with the Kubro(TM) system, with a growing focus on workflow/orchestration.

2. 'Deep research', whether using apps one-off or through custom agentic workflow systems with search APIs, is where we see the greatest practical progress and new applications.

This is also where we focus our attention in early 2026 with 'everything via APIs / MCP' options.

3. In the world of automation, ironically, in a way, the human factor becomes the major source of differentiation, as, from a competitive edge perspective, everyone can use the same tools now.

This adds consultancy services for clients for their internal deployments.

4. In the financial sector, small open-source models fine-tuned on internal proprietary data, engineered for a very specific use-case, seem to be the way forward.

That's our R&D effort underway, specifically for real estate.