AI After the Gold Rush
What’s Actually Working in 2026
From inflated expectations to grounded value where AI is genuinely delivering results.
The past few years have seen AI framed as a technological gold rush. Organisations were urged to move fast, scale aggressively, and adopt ever-larger models or risk being left behind. For many SMEs, cultural organisations, and public bodies, this period was characterised by experimentation, uncertainty, and, in some cases, disappointment.
As we enter 2026, the tone has shifted.
The question is no longer “How quickly can we adopt AI?” but “Which AI approaches actually work and why?”
The answer is far more pragmatic, and far more encouraging, than the hype suggested.
This blog reflects on what has emerged from the initial rush, and where AI is now delivering meaningful, sustainable value.
The End of AI Maximalism
One of the clearest lessons from the past two years is that more AI does not necessarily mean better AI.
Large, general-purpose models promised universality, but in practice many organisations found them:
expensive to operate at scale
difficult to control or explain
poorly aligned with niche or domain-specific problems
environmentally costly
hard to integrate into existing workflows
For SMEs and heritage organisations, the economics rarely stacked up. The result has been a gradual but decisive move away from AI maximalism the belief that scale alone produces value.
Instead, 2026 is shaping up as the year of selective, purpose-built AI.
What’s Actually Working
Across sectors, several clear patterns have emerged.
1. Smaller, Targeted Models
Rather than training or licensing massive models, organisations are deploying smaller systems trained on well-curated, relevant datasets. These models are:
cheaper to run
easier to interpret
faster to iterate
better aligned with specific tasks
For SMEs, this has meant focusing on narrow but valuable use cases, measurement, classification, workflow optimisation, rather than generalised “AI assistants”.
2. Hybrid AI Approaches
Some of the most successful deployments combine machine learning with established analytical or rules-based methods.
By embedding domain knowledge directly into the system, whether through physics-informed models, constraints, or expert rules, organisations reduce uncertainty and improve reliability. Hybrid approaches have proven especially effective in:
3D reconstruction and measurement
structural assessment and risk analysis
heritage documentation
industrial and environmental monitoring
This marks a shift away from treating AI as a replacement for understanding, and towards using it as an accelerator of expertise.
3. Explainability as a Design Requirement
Another defining feature of successful AI in 2026 is transparency.
Organisations are increasingly demanding systems that can show:
why a decision was made
how confident the model is
which data influenced the outcome
where uncertainty remains
Explainability has moved from a regulatory checkbox to a practical necessity, particularly in public-facing, safety-critical, or culturally sensitive contexts.
AI that cannot be interrogated is increasingly seen as a liability rather than an asset.
4. XR and 3D with a Clear Purpose
Extended Reality and 3D imaging have also matured. The projects delivering value are not those chasing spectacle, but those solving concrete problems:
field-based 3D capture for inspection and assessment
accessible visualisation for education and engagement
documentation of fragile or inaccessible sites
training and scenario planning using real-world data
Lightweight, mobile-first approaches, often integrated with AI, are proving far more scalable than bespoke, hardware-heavy deployments.
What Hasn’t Worked
Just as instructive are the approaches that have fallen away:
one-size-fits-all AI platforms
tools adopted without internal skills or governance
projects driven by novelty rather than need
opaque models used in high-trust environments
pilots with no plan for long-term operation
In many cases, failure was not due to flawed algorithms, but to unrealistic expectations and weak organisational foundations.
A Shift in Mindset
What defines the post–gold rush era is not a retreat from AI, but a change in mindset.
AI is increasingly treated as:
infrastructure, not spectacle
a machine tool, not a substitute for judgement
something to be measured, documented, and maintained
a component of broader systems, not a standalone solution
For SMEs and cultural organisations, this shift is empowering. It lowers barriers to entry, rewards expertise, and favours thoughtful design over brute-force computation.
Final Thought
The AI gold rush created momentum, but it also created noise.
What’s emerging in 2026 is a quieter, more sustainable phase, one where success is defined by reliability, transparency, and relevance rather than scale. The organisations that are thriving are those that resisted the pressure to adopt everything and instead focused on what genuinely serves their mission.
AI is no longer about chasing the future.
It’s about building tools that work, here, now, and for the long term.
