AI Technology Radar: What’s Here, What’s Coming, and What’s Next
- Clyne Albertelli
- Aug 25
- 3 min read
Artificial Intelligence is evolving at pace. Some tools are already embedded in our daily work, others are just starting to appear in pilot projects, and many more are still being shaped in research labs. To make sense of this, we use a three-horizon radar:
Business Horizon: technologies available today.
Engineering Horizon: next-generation systems being tested.
Scientific Horizon: frontier research that may define the next decade.
This article explores each horizon and the categories of AI that matter most for business, society, and innovation.
The Three Horizons of AI
Business Horizon (0–5 years)
These technologies are available and proven. They can be purchased, licensed, or adopted today.
Engineering Horizon (5–10 years)
This covers systems that are moving out of the lab and into trials. They are not yet mature but will become mainstream soon.
Scientific Horizon (10–20 years)
This is the frontier. Universities and research labs are pushing boundaries with experimental approaches that could transform what AI can do in the long term.
Categories of AI
Predictive Intelligence
What it is: AI that analyses past data to forecast future outcomes and prevent problems.
IBM: Enterprise tools for anomaly detection and predictive maintenance. Widely used in industries such as energy and finance. IBM Watsonx
Amazon AWS: Cloud AI services that streamline predictive modelling, data labelling, and digital twins. AWS Machine Learning
“Predictive AI shifts maintenance from reactive to proactive, cutting downtime and cost.”
Multi-modal AI
What it is: Models that combine text, images, video, and speech for a more complete understanding.
OpenAI: Multi-modal systems that can answer questions about images, summarise videos, and process spoken commands. OpenAI Research
Meta: Real-time translation and transcription tools, breaking down barriers between languages in workplaces. Meta AI
Explainable AI
What it is: AI that makes its reasoning visible and understandable, rather than acting as a “black box.”
University of Oxford: Research led by Prof. Thomas Lukasiewicz, holder of the AXA Chair in Explainable AI for Healthcare. The team works on neural-symbolic AI, combining logic with neural networks to make clinical decisions transparent. Oxford AI
Anthropic: Developing safety-aligned models with ethics embedded into decision-making. Anthropic
“Explainable AI builds the trust needed for adoption in sensitive areas such as healthcare.”
Reinforcement Learning and Autonomous Agents
What it is: AI trained by reward systems, creating agents that can operate with greater independence.
MIT Lincoln Laboratory: The Mission-Ready Reinforcement Learning (MeRLin) project tested agents in the cooperative card game Hanabi. Research showed humans prefer predictable, transparent AI partners over opaque high-performing ones. MIT MeRLin
NVIDIA: Developing AI agents that can plan and execute multi-step tasks in logistics, robotics, and operations. NVIDIA AI
Generative AI for Content
What it is: Systems that create new content, from text to video, based on prompts.
Stability AI: Advanced text-to-video generation, producing high-resolution, realistic video from text instructions. Stability AI
Microsoft: Integrated generative co-pilots across Office, GitHub, and Azure, turning them into everyday assistants for writing, coding, and analysis. Microsoft Copilot
Neuromorphic Computing
What it is: Hardware and software inspired by the brain, designed for efficient, adaptive learning.
DeepMind: Exploring neuromorphic systems that could reduce the energy demands of large AI models and accelerate their learning capabilities. DeepMind Research
“Neuromorphic computing could unlock AI that is fast, efficient, and closer to human cognition.”
Why This Matters
Tracking AI across all three horizons helps us stay grounded and focused. It's about recognising what we can use today, what's approaching readiness, and what could potentially shape the exciting future ahead.
Predictive systems can already reduce downtime and save money.
Multi-modal models will soon enable seamless collaboration across text, images, and speech.
Explainable AI is laying the foundation for trust in high-stakes environments.
Reinforcement learning agents prove that performance is not enough; predictability and collaboration also matter.
Generative AI is making creativity and automation accessible to everyone.
Neuromorphic computing holds the promise of human-like efficiency in machines.
By looking across the horizons, organisations and individuals can prepare not only for the tools they can use today, but also for the innovations that will shape how we work and live tomorrow.
Comments