Artificial intelligence is already changing the way designers create, communicate and deliver ideas across architecture, interiors and lighting design.
We asked an experienced lighting designer to share their thoughts on where AI is heading, where it can support designers, and why human creativity and judgement remain essential.
AI in Lighting Design
Many architects and interior designers are already using generative AI tools for 3D visuals and conceptual rendering. By doing so, they can quickly build a library of ideas and options to discuss with their client in the time it might traditionally take to create just a small handful.
The same can be said for lighting.
The ability to produce a series of early visual mock-ups in a matter of minutes, rather than hours, is a huge time saver. At this stage, it’s more about nailing the aesthetic and refining the scope and brief, rather than creating a picture-perfect scene.
Mood boards can be rapidly assembled with an array of lighting elements and textures to support your proposals. Gone are the days of trawling the depths of Pinterest, a designer’s worst nightmare of a time-sink.
When dealing with accuracy, one common downside of generating AI imagery based on a specific scene is often the tool’s willingness to “hallucinate”, imagining changes to a building’s architecture or room space beyond the user’s specific requests.
This is not ideal when wanting to show a client how their project will look in a variety of illuminated conditions. However, recent models have started to minimise this issue, and guardrails can be put in place with strict instructions and rulesets to guide the model to only make changes as requested.
For this critical reason, at present, whilst a great assistant, the accuracy of AI can always be improved.
Furthermore, the output of an AI tool is no greater than the sum of the user’s inputs. You only get out what you put in.
The quality of the tool’s potential suggestions and assistance in the design process is wholly dependent on the volume, quality and variance of training data it was given.
Most importantly, you are still the designer.
You are the judge and arbiter of what good lighting design looks like. AI can’t replace that. In fact, it needs you for that very reason.
It is here to augment us, to enhance our work production capabilities, but we still call the shots, even down to the information it is trained on.
You are training it not just on best practice, but also on what you consider good lighting design. Style, mood and aesthetics are very human, subjective qualities.
Over the next five years, I am hopeful for greater performance in the technical side of lighting design, supporting lighting calculations, automating CAD layouts and take-offs.
Some early signs are there, but in my experience, it needs a greater level of consistency to be more reliable in these areas.
There is an inherent level of forgiveness the user can offer when it comes to conceptual hallucination.
Technical drawings and calculations must be accurate. There is a much smaller margin of error and greater ramifications.
That said, to err is human, and just as a team leader checks their team’s workings, AI outputs should always be quality checked.
AI in Lighting Control & Automation
Beyond smart home automation, AI has been pitched for monitoring both energy use and user wellbeing in construction projects.
Over the past few years, the idea of dynamically responsive buildings using AI and machine learning for energy efficiency and occupant wellbeing has been put forward.
Lighting control companies like Helvar have published whitepapers on lighting and artificial intelligence.
AI-augmented daylight monitoring and dynamic human-centric lighting tools from companies like Kumux look to optimise and bridge the gap between the control system and luminaires.
AI in Lighting Project Delivery & Workflow
There is a popular AI conversation:
“I want AI to do the dishes and the laundry, so I can have fun making art — not AI making art so I can do the dishes and the laundry.”
The truth is that this is becoming a possibility.
Whilst most attention is paid to AI-generated imagery, agents and large language models can be utilised for menial and repetitive administrative work.
In lighting design, this covers tasks like:
- data extraction for luminaire schedules and specifications
- comparison of technical submittals from contractors
- finding key guidance data from lighting regulation documents
Tools like Google’s NotebookLM can behave like highly specialised databases based on the subject matter you upload.
By only referencing these specific files, it minimises the chances of hallucinations.
Streamlining these pattern-matching tasks frees up the designer to focus on the more challenging and creative aspects of lighting design.
Ethical, Creative & Human Considerations
Manual sketching, slow render cycles, painstaking revisions, hours spent iterating.
Still admirable, still respected, but time consuming.
Meanwhile, someone using paid AI tools can generate dozens of variations in minutes, revise on demand, produce entire concept suites before lunch and test multiple directions without pausing to breathe.
Time, once a neutral resource, has become something you can purchase.
And as in any pay-to-win system, those who don’t pay are not merely slower; they are systematically disadvantaged.
The job market has already begun to adjust, even if not without pushback.
The real question is whether, if we continue down this path, people who avoid or resist AI tools will find themselves more expensive, less competitive and harder to hire.
It isn’t framed as punishment; it simply manifests as preference.
Faster becomes better. Better becomes safer. Safer becomes indispensable.
Once that loop begins, it compounds: increased productivity leads to more opportunities, which leads to more income, which funds more capable tools, which accelerates productivity further.
Creative professionals relying purely on traditional tools are not less talented; they are simply operating inside a time regime that could potentially no longer align with future market expectations should AI achieve mass adoption.
One continued concern is the rate at which these AI tools develop. It gives its audience so little time to truly master them.
The digital race to stay on top is creating an endless loop within the early stages of AI tools’ lifecycles. Before users can truly master their intricacies, these tools are often made obsolete by a rival, or sometimes even by their own development teams as newer models are released.
In such an evolving landscape, a market leader one week can be old news the next.
Instead of this endless pursuit to outdo and outperform, more attention needs to be spent on approachability, consistency and, most importantly, reliability.
With each advancement, newer models can sometimes appear to experience a step back rather than a step forward.
Accusations of “nerfing” (reducing a model’s capabilities), changes to behaviour and responses, broken functionality or sudden increases in errors can all be by-products of chasing the competition.
These changes are often carried out via AI itself to “save time”, even if the development team then spends the next few months hot-fixing.
None of this instils confidence in the userbase.
As with any new system, greater care needs to be taken in training, understanding and intuitive design from the ground up, in order for general mass adoption and implementation of AI-driven infrastructure.
Final Thoughts
AI will continue to influence the way the lighting industry designs, collaborates and delivers projects.
But technology alone does not create great lighting.
The future will depend on balancing intelligent tools with the creativity, experience and judgement of the people using them.
