Emerging Technology & Innovation Radar

150 technologies. 5 industries. One map.

An independent radar placing each technology into Adopt, Trial, Assess or Hold — with a written brief explaining why.

Methodology

How to read this radar — the full guide to rings, quadrants and how placements are decided.

Adopt
Proven and low-risk; invest now.
Trial
Worth piloting on a real project.
Assess
Worth understanding, not yet committing.
Hold
Proceed with caution, or de-prioritise.

Quadrants: AI & Agents · Platforms & Infrastructure · Techniques & Methods · Tools & Languages

Technology & SaaS

Where software, cloud and AI-native platforms are placing their next bets.

Edition 2026.1 · published 2026-10-01

Themes this edition

  • Agents leave the sandbox — Agentic workflows and autonomous coding agents move from demo to default — the question for 2026 is governance and cost control, not capability.
  • The inference bill becomes a board topic — Serverless GPU inference, SLMs and aggressive model routing are the direct response to runaway LLM spend — FinOps for AI is no longer optional.
  • Trust infrastructure catches up — Non-human identity, confidential computing and evaluation-driven development are the unglamorous plumbing that makes agentic AI safe to ship.

Adopt 8

Trial 14

Assess 6

Hold 2

On the periphery — watchlist

  • World Models — Models that learn predictive simulations of physical/virtual environments — early, but a plausible successor to today's LLM paradigm.
  • AI Coworkers — Persistent, named AI identities with their own accounts and memory, working alongside humans rather than as a tool.
  • Browser-Use Agents — Agents that operate a real browser UI to complete tasks — promising, but reliability and safety guardrails are immature.
  • Quantum-Safe Cryptography — Post-quantum algorithms for long-lived data that must stay confidential past the arrival of cryptographically relevant quantum computers.
  • Context Engineering — The emerging discipline of deliberately designing what goes into a model's context window, as distinct from prompt engineering.
  • Answer Engine Optimisation (AEO) — Optimising content to be surfaced and cited by AI answer engines, the successor discipline to SEO.
  • GPU Neoclouds — Specialised, GPU-only cloud providers undercutting hyperscalers on raw training/inference price.
  • AI SRE / Agentic AIOps — Agents that triage, diagnose and in some cases remediate production incidents with minimal human involvement.
  • Digital Product Passport-style Metadata — Structured, machine-readable provenance metadata attached to software artefacts and models, not just physical goods.
  • Private 5G / Edge Connectivity — Dedicated low-latency wireless networks for campuses and facilities, increasingly paired with edge inference.
  • Spatial Computing for Ops — AR/VR interfaces for remote collaboration and operations, well beyond the consumer headset hype cycle.
  • Agentic Supply/Ops Planning — Agents coordinating multi-step operational planning across internal systems — an early but fast-moving category.

Editor's note

If I had to defend five bets to a CFO tomorrow, these are them: Agentic Workflows and Autonomous Coding Agents because they compound engineering leverage faster than any hire; Sovereign & Regional Cloud because it is becoming a deal-breaker in enterprise procurement; Threat Modeling for ML/LLM Systems because the first serious agent-related breach is a matter of when, not if; and Rust for Systems & Data because the talent and tooling have finally caught up to the hype.

Healthcare & Life Sciences

Where clinical care, biomedical research and health-tech platforms are converging.

Edition 2026.1 · published 2026-10-01

Themes this edition

  • The clinician gets an AI co-pilot, not an AI replacement — Ambient scribes and diagnostic-support AI are being adopted fastest where they reduce administrative burden, not where they touch final clinical judgement.
  • The lab goes autonomous — Self-driving labs and generative biology are compressing discovery timelines that used to take years into months — the bottleneck is shifting from experiments to validation.
  • Interoperability is finally non-negotiable — FHIR-native data platforms and federated learning are responses to the same pressure: health data must move safely between systems without ever leaving the system of record.

Adopt 8

Trial 14

Assess 7

Hold 1

On the periphery — watchlist

  • Agentic Clinical Trial Matching — Agents that continuously match eligible patients to open trials across a health system's records.
  • AI-Native Patient Engagement — Conversational agents handling intake, triage-adjacent questions and follow-up outside clinical judgement.
  • Context Engineering for Clinical Prompts — Deliberate design of what clinical context is fed into a model, distinct from general prompt engineering.
  • Quantum-Safe Cryptography for Health Records — Post-quantum encryption for health records that must remain confidential for decades.
  • Spatial Computing for Surgical Planning — AR/VR visualisation of patient anatomy for pre-operative planning and training.
  • Answer Engine Optimisation for Patient Information — Ensuring authoritative health content is correctly surfaced and cited by AI answer engines.
  • Private 5G for Hospital Campuses — Dedicated low-latency wireless networks supporting real-time device and robotics coordination.
  • AI-Powered Pharmacovigilance — Automated signal detection across adverse-event reports at a scale manual review cannot match.
  • Digital Biomarkers from Wearables — Continuous physiological signals used as endpoints or early indicators in research and care.
  • Embodied AI for Hospital Logistics — Autonomous robots handling supply delivery and routine logistics within care facilities.
  • AI SRE for Clinical Systems Uptime — Agentic operations tooling applied to the uniquely high-stakes uptime requirements of clinical IT.
  • Generative Models for Clinical Trial Design — Early-stage use of generative AI to optimise trial protocol design itself.

Editor's note

If I had to defend five bets to a chief medical officer tomorrow: Ambient Clinical Documentation AI because it is the clearest, safest, fastest ROI available today; Generative Biology and Self-Driving Labs because together they are compressing discovery timelines more than any single technology has in a generation; Federated Learning because it finally offers a credible path through the data-sharing deadlock; and Evaluation-Driven Clinical AI because, in this industry more than any other, you cannot responsibly ship what you have not rigorously measured.

Financial Services

Where banking, payments, insurance and wealth management are re-platforming for an agentic, real-time world.

Edition 2026.1 · published 2026-10-01

Themes this edition

  • Explainability stops being a nice-to-have — As AI moves into credit, underwriting and AML decisions, explainable-by-design models and evaluation suites are becoming the condition of regulatory approval, not an afterthought.
  • Real time becomes the default, not the premium tier — Real-time payments rails and agentic fraud investigation are compressing decision windows from days to seconds across the transaction lifecycle.
  • The core finally gets rebuilt — Composable core banking and autonomous coding agents are, for the first time, making large-scale legacy core modernisation a realistic multi-year plan rather than a perpetual deferral.

Adopt 6

Trial 15

Assess 8

Hold 1

On the periphery — watchlist

  • Agentic Underwriting Assistants — Agents assembling full underwriting case files across multiple data sources for human sign-off.
  • Embedded Finance Orchestration — Platforms letting non-financial brands embed banking, lending and payments natively into their products.
  • AI-Native Insurance Claims Triage — Multimodal models assessing claims evidence (photos, documents) for faster, fairer triage.
  • Context Engineering for Risk Prompts — Deliberate design of what data and context is fed into risk and compliance models.
  • Answer Engine Optimisation for Financial Content — Ensuring regulated, accurate financial guidance is correctly surfaced by AI answer engines.
  • Central Bank Digital Currencies (CBDC) Rails — Infrastructure readiness for potential retail and wholesale CBDC rollouts across markets.
  • GPU Neoclouds for Risk Model Training — Specialised GPU providers offering lower-cost training capacity for internal risk model development.
  • AI SRE / Agentic AIOps for Core Systems — Agents triaging and in some cases remediating incidents in payment and core banking systems.
  • Behavioural Biometrics for Continuous Authentication — Passive, continuous identity verification based on interaction patterns rather than one-time checks.
  • Climate & ESG Risk Modelling AI — Models quantifying climate transition risk exposure across lending and investment portfolios.
  • AI-Assisted Regulatory Horizon Scanning — Automated tracking and summarisation of incoming regulatory change across jurisdictions.
  • Decentralised Identity for KYC — Reusable, user-controlled digital identity credentials reducing repeated KYC friction.

Editor's note

If I had to defend five bets to a chief risk officer tomorrow: Explainable AI for Regulatory Compliance, because every other AI bet in this list depends on it being solved first; Agentic Fraud Investigation, because the ROI case is immediate and the risk profile is manageable; Composable Core Banking, because it is finally making legacy modernisation tractable rather than perpetually deferred; Sovereign & Regional Cloud, because data-residency regulation is tightening faster than most technology roadmaps assume; and AI-Powered AML, because false-positive fatigue is now a genuine operational risk in its own right.

Manufacturing & Industrial

Where the shop floor, the supply chain and the engineering desk are being re-wired by AI and connectivity.

Edition 2026.1 · published 2026-10-01

Themes this edition

  • The factory gets an agentic layer — Agentic shop-floor operations and supply/production planning agents are moving from pilot to production fastest where downtime cost makes the ROI case undeniable.
  • OT security stops being an afterthought — Threat modeling for ICS/OT systems and non-human identity governance are catching up to a decade of converged IT/OT risk that regulation is now forcing into the open.
  • The digital twin becomes the engineering desk — Industrial digital twins and generative design are converging into a single simulated environment where new products and processes are validated before a single part is machined.

Adopt 5

Trial 15

Assess 9

Hold 1

On the periphery — watchlist

  • Embodied AI & Mobile Robotics — Autonomous mobile robots handling material movement and routine physical tasks on the floor.
  • Context Engineering for Technician Copilots — Deliberate design of what equipment and process context is fed into technician-facing AI.
  • AI SRE / Agentic AIOps for Industrial Control Networks — Agents triaging and in some cases remediating incidents in converged IT/OT network environments.
  • Spatial Computing for Remote Assistance — AR-guided remote expert assistance for maintenance and commissioning tasks.
  • GPU Neoclouds for Simulation Workloads — Specialised GPU providers offering lower-cost capacity for digital-twin and simulation compute.
  • Quantum-Safe Cryptography for Industrial Control Systems — Post-quantum encryption planning for long-lived industrial control and safety systems.
  • Agentic Energy & Utility Optimisation — Agents continuously optimising plant-level energy consumption against cost and carbon targets.
  • Answer Engine Optimisation for Technical Content — Ensuring authoritative technical and parts documentation is correctly surfaced by AI answer engines.
  • 3D-Printed Production at Scale — Additive manufacturing moving from prototyping into genuine low-to-mid volume production use cases.
  • Self-Driving Labs for Materials Science — Automated, closed-loop experimentation accelerating new materials discovery for industrial use.
  • Decentralised Supplier Identity & Provenance — Verifiable, portable supplier and component provenance credentials across supply networks.
  • AI-Native Capacity & Scenario Planning — Generative scenario modelling for multi-site capacity and investment decisions.

Editor's note

If I had to defend five bets to a plant operations director tomorrow: Agentic Shop-Floor Operations because the downtime-cost math makes the ROI case undeniable; Industrial Digital Twins because they have quietly become the default environment for any serious process change; Digital Product Passport because extended-producer-responsibility regulation is turning it into a market-access requirement, not a sustainability nice-to-have; Threat Modeling for OT/ICS Systems because converged IT/OT risk is the single most under-addressed exposure in most plants today; and Generative Design because it is producing manufacturable designs that quite simply outperform what human-first engineering was producing a few years ago.

Retail & Consumer

Where shopping, merchandising and the store itself are being rebuilt around agents and answer engines.

Edition 2026.1 · published 2026-10-01

Themes this edition

  • The customer is increasingly an agent, not a person — Agentic commerce and shopping assistants are starting to browse, compare and even check out on a shopper's behalf — optimising for AI answer engines is becoming as important as optimising for search.
  • Personalisation needs a rigorous evaluation discipline — As recommendation and merchandising decisions move to reasoning models, evaluation-driven development is the only way to avoid silently degrading customer experience at scale.
  • The store becomes a real-time, edge-native system — Unified commerce, edge AI at the point of sale and private connectivity are converging so that online and in-store inventory, pricing and promotions move in lockstep.

Adopt 7

Trial 15

Assess 6

Hold 2

On the periphery — watchlist

  • Context Engineering for Shopping Assistants — Deliberate design of what customer, catalogue and order context is fed into shopping-agent prompts.
  • AI Coworkers for Merchandising Teams — Persistent AI identities handling recurring pricing, assortment and promotion analysis tasks.
  • Browser-Use Shopping Agents — Third-party agents that operate a real browser to compare prices and complete purchases on a shopper's behalf.
  • GPU Neoclouds for Personalisation Model Training — Specialised GPU providers offering lower-cost capacity for recommendation-model training.
  • AI-Native Visual Merchandising — Generative tools designing in-store and digital visual merchandising layouts automatically.
  • Embodied AI for Fulfilment Centres — Autonomous robots handling picking, packing and routine fulfilment-centre logistics.
  • Decentralised Loyalty & Identity — Portable, customer-controlled loyalty and identity credentials usable across retailers.
  • Quantum-Safe Cryptography for Payment Data — Post-quantum encryption planning for long-lived stored payment and customer data.
  • AI SRE / Agentic AIOps for Commerce Platforms — Agents triaging and in some cases remediating incidents during peak-traffic commerce events.
  • Spatial Computing for Store Design — AR/VR tools for planning and testing in-store layouts and planograms before physical rollout.
  • Hyper-Local Dynamic Pricing — Real-time, store- or micro-market-level pricing driven by local demand and inventory signals.
  • Social Commerce Agent Integrations — Shopping agents embedded directly inside social and messaging platforms, not just storefronts.

Editor's note

If I had to defend five bets to a chief commerce officer tomorrow: Agentic Customer Service because the resolution-rate gains are already proven at scale; Answer Engine Optimisation because the shift from search boxes to AI answer engines is happening faster than most commerce teams have budgeted for; Retail Media Network Infrastructure because it is quietly becoming one of the highest-margin lines on the P&L; Agent & Tool Protocols for Agentic Commerce because the standards race for how shopping agents interact with your storefront will determine who stays visible in an agentic shopping world; and Generative Product Content because it has made catalogue-scale localisation genuinely affordable for the first time.

The same technology, across industries

30 cross-cutting themes appear in more than one industry — often in different rings.