3 Characteristics of Tasks Suitable for AI in Lighting Design Companies

Recently, a peer mentioned that after using Anylight to generate lighting renderings, several designers in his company seemed at risk, and many junior positions might become unnecessary.

This statement cannot be denied. I have previously written that AI will inevitably impact the lighting industry, but I think we should be cautious about this idea.

Because a lighting project is not as simple as writing a piece of copy or generating an image.

From the moment a client speaks, to site inspection, to design direction, to quotation, to presentation, to construction and implementation, there are many stages in a project.

In some stages, AI can indeed save you time.

But in other stages, if you let AI take charge, when something goes wrong, AI won't explain to the client on your behalf, nor will it bear the losses for you.

1. Not All Work Is "Design"

In lighting projects, tasks like designing renderings and writing proposals can be complex, but their boundaries are clear.

Once completed, you can basically tell at a glance whether they are usable.

If not suitable, revise. If the direction is wrong, change.

I think it's fine to delegate such tasks to AI.

But there are also tasks that appear to be part of the design process but are not simple execution.

For example, the sequence, tone, and manner of presenting a proposal, as well as client feedback and complaints.

Such judgments cannot be directly determined by AI.

They rely on your understanding of the client, the site, the budget, and the audience for the presentation.

This is where the real difficulty lies in lighting projects.

2. Tasks Suitable for AI Usually Have 3 Characteristics

I believe that tasks suitable for AI in lighting projects generally have the following characteristics:

First, clear boundaries.

For example, organizing client materials into a one-page project background.

Compiling meeting minutes, writing proposal descriptions, design notes, using Anylight.net to generate several different lighting renderings, etc.

For these tasks, you can judge whether the result is good or bad. If AI does a poor job, you know where it falls short.

Second, fast feedback.

For example, a version of a proposal copy, a rendering, a set of presentation titles.

If done today, it can be revised today.

A mistake just wastes a bit of time.

It won't immediately lead the project into a pitfall.

Third, low cost of errors.

If the copy is not smooth, revise it.

If the rendering is not suitable, generate a new batch.

These are all trialable.

So I have always believed that the primary value of AI in lighting projects is not to make final decisions for you.

Instead, it takes over the repetitive, trial-and-error, organizational, and expressive tasks in the early stages.

This allows people to spend less time on low-value execution.

3. What Cannot Be Delegated to AI: Things That Cannot Afford Mistakes

What truly cannot be easily delegated to AI is another type of work.

For example, pre-sales for major clients, how to commit to budgets, whether site conditions can support the design approach, whether construction changes need to be communicated to the client in advance, and during proposal presentations, whether to first discuss the effects or the risks.

If these go wrong, it's not just a matter of revising a PPT.

It could mean the client loses trust in you.

It could mean project delays.

It could mean construction rework.

Or it could mean losing a major deal entirely.

In many lighting projects, the biggest fear is not that the proposal is a bit slow.

The biggest fear is making a wrong judgment early on.

If the direction is wrong, the harder you work later, the more trouble you create.

If commitments are overpromised, you'll be passive no matter what you do later.

If site conditions are not thoroughly understood, the construction team, client, and designers will end up in disputes.

In such situations, AI can help you list items, remind you of risks, and simulate what questions the client might ask.

But whether to proceed, who says what, and how to say it—these must be decided by humans.

AI only presents options and can even help you do many things faster.

But it cannot distinguish priorities for you, nor can it bear the consequences for you.

So, what kind of tasks in lighting projects are suitable for AI?

Those with clear boundaries.

Those with fast feedback.

Those where mistakes can be corrected affordably.

Which tasks should not be left to AI?

On-site risks.

Critical commitments.

Final trade-offs.

Ultimately, AI should be used in the right places.

When used for data sorting, proposal drafts, direction exploration, and communication assistance, it is a tool.

When used for client judgment, on-site risks, and project decision-making, it can become a risk.