Many organisations have adopted AI, only to be taken aback by how much material suddenly appears. There is a huge amount of text, code and plans coming out, and people do not always know how to assess the results critically. In the worst case, evaluating the quality is pushed onto someone else, or even a larger group of people.
That is why I want to write about what expertise actually means now that AI is doing an increasing share of the work. I believe AI does not make expertise less important. Instead, it shifts where the value of expertise is created. I’ll first look at what this means in practice and then, at the end, what it means for you if you’re considering your next job as a software developer.
Convincing does not necessarily mean good
AI produces text with perfect grammar and a neat structure. For us humans, this can easily create the impression that the content itself must also be good. When AI is used simply through a chat interface without proper context, the result is superficial at best. It lacks the understanding that a company or an expert has.
That is why challenging what AI produces is the expert’s job, and it is difficult precisely because the output looks so good. Sometimes it looks so convincing that people don’t dare to edit it. We need to be patient with people’s learning curves here, because this is a new skill. But the principle is still clear: you need to understand what you share with others and stand behind it. Ownership does not transfer to AI just because AI produced the content.
Everyone has AI, but not everyone has a competitive advantage
Today, anyone with a credit card can access top-level AI. That means AI alone does not give anyone a competitive advantage. Without an expert and without a company’s own data, it produces generic work, and a competitor can get the same generic work with the same credit card.
The expert’s job is to enrich the work AI produces. They bring in their own expertise and their understanding of the company, challenge the solutions AI suggests and bring the company’s competitive advantage into the way AI is used. A competitive advantage emerges when a person and AI together can produce a better result than anyone else.
This is also how you grow into an expert. When you analyse and challenge AI-generated content and ask whether it is genuinely relevant and whether it offers more value than a generic answer, you are practising exactly the skill that is needed.
Customers still buy results
Customers have always wanted results and been willing to pay for them. That has not changed. It is the supplier’s responsibility to decide how those results are produced.
What has changed is the emphasis in what we expect from experts. Soft skills were already important because people want to work with people who are pleasant, communicate clearly and consider others. Alongside these skills, the ability to challenge AI is becoming increasingly important. Customers should not be given raw AI-generated material to deal with. Every delivery must carry a human stamp.
The same shift can be seen in contracts. If agentic AI is actively used in development, it creates significant token costs. These costs are small compared with the productivity gained, but a purely hourly billing model does not currently serve the customer well. If the supplier has to cover these costs, there is little incentive to make full use of AI, and the customer misses out on the benefits. That is why hourly billing plus token costs is currently a sensible way to buy. With an informed customer and a good supplier, it is possible to go further and agree directly on the result and its value.
Value is shifting from outputs to systems
This is the idea that I believe will make the biggest difference. I see the role of an expert developing in three stages.
In the first stage, the expert is responsible for every output produced by AI. Their value comes from ensuring that the end result is better than generic AI output. This stage cannot be skipped.
In the second stage, the expert has built their own knowledge and expertise into an agentic system so well that the system can independently produce expert-level work. Their value is then already baked into the system. The work changes: instead of improving an individual result, they improve the system itself. This is something to agree on together: how do we want to work?
In software development, for example, this could mean an experienced developer documenting the principles they use to accept or reject changes in code reviews and incorporating them into the team’s AI system. The same applies not only to software development but to all expert work.
In the third stage, the value begins to multiply. The world is constantly changing and people are constantly learning more, so the expert’s work does not end when some of their expertise has been transferred to the system. As they teach the system to become better and better, their impact grows. They can also expand the area they cover and start teaching the next agentic system. Experts who can build systems at this level are hugely valuable to any organisation.
What does this mean for a software developer?
If you are considering your next job, it is worth asking about more than which AI tools the organisation uses. Almost everyone has the same tools. Instead, ask whether your work will give you the opportunity to practise what sets an expert apart: challenging AI-generated outputs, bringing your own expertise into the way AI is used and, ultimately, building systems that produce expert-level work.
I don’t think the expertise of software developers is losing its value. I think its value is multiplying. That requires ownership and a willingness to keep learning. At Gofore, we want to be a place where that is possible.
