Experience in the age of AI
There's a common argument about artificial intelligence and the future of work: as AI becomes more capable, years of professional experience will become less valuable.
Someone with limited experience but access to powerful AI tools may soon be able to produce work that previously required a seasoned professional.
There's some truth to that. AI is making many tasks easier, faster and more accessible.
But I believe there's another side to this transformation that deserves more attention.
AI isn't just helping inexperienced people do more. It's also allowing experienced professionals to accomplish things that were previously beyond their practical reach.
That's the change I've experienced personally.
After years working across graphic design, photography, websites and client projects, I now use AI-assisted development tools to build digital systems that would previously have required me to engage additional developers or coordinate several specialists.
The experience I've accumulated hasn't become less relevant.
If anything, I've found more ways to apply it.
AI is narrowing some skill gaps
Research provides useful evidence that AI can help people with less experience perform certain tasks more effectively.
In a study of more than 5,000 customer-support agents (opens in a new tab), researchers found that access to a generative AI assistant increased productivity by approximately 14% on average.
The improvement was considerably greater among less experienced and lower-skilled workers, who achieved gains of around 34%.
A separate Harvard Business School and Boston Consulting Group study (opens in a new tab) found that consultants using AI performed better on certain tasks, with lower-performing participants benefiting particularly strongly.
These findings are significant.
They suggest that AI can make useful knowledge and capabilities available to people who haven't yet accumulated years of experience.
I see that as a positive development.
Learning and professional development have always involved acquiring knowledge, practising skills and learning from others. AI introduces another powerful way to accelerate parts of that process.
But there's an important distinction between producing an answer and knowing whether it's the right answer.
The problem with answers that are almost right
One of the most interesting findings from the Harvard and BCG research concerns what the researchers called the jagged technological frontier.
AI performed well on some tasks but poorly on others, and the boundary between them wasn't always obvious.
For a task deliberately selected outside the AI's capabilities, consultants using AI were 19% less likely to produce the correct solution than those working without it.
That's a useful warning.
AI can produce an answer that sounds convincing, looks complete and follows the requested format—even when important details are wrong.
The same concern appears in software development.
A study by researchers at Stanford University on AI-assisted programming (opens in a new tab) found that participants using an AI assistant produced less secure code in several studied scenarios while showing greater confidence in its security.
And the 2025 Stack Overflow Developer Survey (opens in a new tab) identified AI-generated solutions that are nearly correct, but not quite, as a major frustration among developers.
These findings don't mean AI should be avoided.
They mean that reviewing, questioning and validating AI output is part of using it responsibly.
That's where experience becomes especially valuable.
Experience teaches you what to look for
Before working with modern AI development tools, I'd already spent years dealing with clients, creative projects, websites and business requirements.
Those experiences taught me things that aren't necessarily visible in a technical specification.
A client might describe a particular feature they want, but the real problem could be an inefficient process behind it.
A website might meet every item in a design brief but still be confusing to visitors.
A system might technically function while creating unnecessary work for the people expected to use it.
These aren't always programming problems.
They're problems of understanding people, businesses and how work actually happens.
AI can help implement a proposed solution. But deciding whether that solution addresses the real problem requires investigation and judgement.
That's one of the most valuable things I bring to a project.
What photography and design taught me about software
My background in photography and visual design has had a surprisingly direct influence on how I approach software development.
In photography, technical correctness doesn't automatically produce a good image.
A photograph can be properly exposed, sharply focused and technically sound, yet fail to communicate what the client needs.
You still have to consider composition, context, attention and purpose.
Graphic design works similarly.
An attractive layout isn't necessarily effective if the information hierarchy is wrong or the audience struggles to understand it.
I've found that digital systems require much the same judgement.
An application can contain all the requested features and still be difficult to use.
A dashboard can display accurate information but present it in a way that makes decisions harder.
A website can look modern but load slowly or prevent its owner from updating basic information.
These are the kinds of problems that experience helps me recognise.
When working with AI-generated interfaces, I don't evaluate them only by whether they run.
I also consider whether the layout makes sense, whether the information is understandable and whether the result genuinely helps the intended user.
The ability to build something and the ability to recognise what should be built are different skills.
Both matter.
AI has changed what I can personally deliver
One of the biggest changes in my professional work has been the distance between identifying a business problem and delivering a working solution.
Earlier in my career, I could design websites, develop concepts, understand client requirements and coordinate technical work.
That experience predates AI. I was building websites with HTML, CSS and JavaScript well before AI coding tools existed. AI-assisted development didn't introduce me to website development; it expanded how much of a complete system I can implement myself.
But more complex functionality often required additional development resources.
That meant more coordination, more handovers and sometimes compromises between the original concept and what could realistically be delivered.
Modern development tools have changed that.
Today, I use AI coding agents alongside technologies such as Next.js, React, TypeScript, Sanity and Vercel.
This allows me to take responsibility for a broader range of work, from planning and interface design to implementation, content management, integrations, testing and deployment.
For example, I've worked on systems involving structured content, data imported from spreadsheets, research workflows, automation and internal business processes.
I've also developed a progressive web application designed to make business information and workflows accessible through an app-like interface.
These projects involve different technical disciplines, but they share something important: they begin with understanding what people need to accomplish.
AI assists with the implementation.
My responsibility is to direct the work, evaluate the results and make sure the system serves its intended purpose.
That doesn't make me equivalent to an entire team of specialists in every discipline.
It means I can now cover more of the development process independently than I could before.
For appropriately scoped projects, that can be a meaningful advantage.
AI is an amplifier, not a substitute for judgement
Google's 2025 DORA research (opens in a new tab) described AI as an amplifier of an organisation's existing strengths and weaknesses.
I think that's a useful way to understand its role.
If a project begins with unclear requirements, AI can help produce the wrong solution more quickly.
If the underlying design is confusing, AI can generate additional features without resolving that confusion.
And if nobody checks the implementation carefully, errors can become part of a system that otherwise appears complete.
But when the objectives are clear, the requirements are understood and the results are reviewed critically, AI can help accelerate meaningful work.
I've found that the quality of the instructions matters, but so does the ability to recognise when the output needs improvement.
Sometimes the right decision isn't to ask AI to generate more.
It's to simplify, remove an unnecessary feature or reconsider the original approach.
Those decisions still require human judgement.
Experience also means knowing your limitations
I don't believe using AI responsibly means pretending it can do everything.
There are situations where specialist expertise remains essential.
Security-sensitive applications, complex infrastructure, regulatory requirements and systems involving significant financial or operational risks may require independent review or additional specialists.
Specialist developers and technical professionals remain valuable. AI lets me take greater responsibility for implementation, but part of that responsibility is recognising when a project needs specialist expertise and bringing it in.
AI-generated code still needs testing. Integrations need to be verified. Access controls need to be checked. Software dependencies require maintenance.
And a working demonstration isn't the same as a production-ready system.
Recognising those distinctions is part of professional responsibility.
I also believe transparency matters.
I'm comfortable explaining that AI-assisted tools are part of my development process.
They don't replace the work of understanding requirements, making design decisions, reviewing implementations or accepting responsibility for delivery.
Clients should know how their projects are being developed and what safeguards are appropriate.
Ultimately, the tools are part of the process—not the entire value of the service.
The value of a broader professional background
For many years, professional careers were often described in relatively narrow terms.
Someone was a photographer, graphic designer, web designer, developer or project manager.
Those specialisations still matter, particularly where deep expertise is required.
But AI-assisted tools are making it easier for people with experience across several disciplines to connect those skills in practical ways.
For me, photography contributes an understanding of visual communication.
Graphic design contributes layout, hierarchy and attention to detail.
Years of client work contribute an understanding of expectations, revisions, deadlines and business requirements.
Website development contributes technical knowledge and experience maintaining real systems.
AI-assisted engineering helps bring those capabilities together into working applications.
None of those experiences became irrelevant when I started using AI.
They became more useful in combination.
What this means for businesses
For businesses considering a website, application or internal system, I think there's an important lesson.
Choosing someone to build a solution shouldn't depend entirely on how quickly they can generate code.
It's also worth asking whether they understand the business problem.
Can they question requirements constructively?
Can they identify unnecessary complexity?
Will they consider the people who need to use the system?
Do they understand what happens after launch?
And will they take responsibility for the finished work?
AI may change how much work one person can accomplish, but those questions remain important.
In some cases, a capable independent professional using modern tools may be able to deliver a focused project efficiently.
In others, the complexity and risk will justify a larger specialist team.
The objective should always be to match the approach to the actual requirements.
Experience isn't becoming obsolete. Its role is changing.
I don't think AI will leave every profession unchanged.
Some tasks will become easier to automate. Some skills will become less valuable, while others will become more important.
But I don't believe the ability to generate an answer is the same as understanding a problem.
And I don't believe producing something quickly is the same as producing something useful.
My own experience has shown me that AI can dramatically expand what an individual professional is able to deliver.
But the value still comes from understanding what needs to be done, recognising when something isn't right and taking responsibility for the outcome.
AI changes the tools available to us. Experience helps us decide what to do with them.
That's the combination I'm continuing to build my work around.
If you're exploring a website, business application or digital system and want someone who considers both the technical implementation and the business behind it, I'd be happy to discuss your requirements. Get in touch.