BLOG / TECHNOLOGY & COLLABORATION
AI can build quickly. Who takes responsibility for the result?
AI has made it easier to create websites and software. I use AI too. But a solution used by customers or colleagues needs more than a fast first draft.
The hard question comes before the code.
When someone asks for a new website, the job is rarely just writing code. Who should the site help? What do visitors need to understand? What should they be able to do next? A consultant can ask those questions, prioritize the content and propose a solution that fits the business. AI can help with sketches and alternatives, but it needs someone to provide the right context.
The same applies to internal systems. Before automating a workflow, you need to understand who uses it, where the data comes from and what happens when something goes wrong.
Security and privacy need context.
A contact form can look finished after one AI prompt. Someone still needs to decide who can access the enquiries, where they are stored, how long they are kept and what happens if the form is misused. Norway’s Data Protection Authority reminds businesses to protect personal information about customers and users, including data collected through websites.
If the solution also includes an AI model, there are additional questions about data flows and control. OWASP identifies prompt injection and sensitive information disclosure among the risks in AI applications. That does not make AI unusable. It means the application should be designed and tested with those risks in mind.
Security in ordinary websites and systems also involves access control, validation, updates and the ability to recover. Measures should match the risks of the particular business. That is also the starting point of the Norwegian National Security Authority’s basic cybersecurity principles.
A first draft is not a finished delivery.
AI can provide a useful starting point for copy, design and code. Someone still needs to check that the content is correct, the site works on mobile, it can be used with a keyboard and screen reader, the form delivers messages, and the solution can be maintained. Will it still work as the business changes services or receives more enquiries?
NIST’s AI Risk Management Framework emphasizes mapping, measuring and managing risk throughout the lifecycle. It is a useful reminder that quality takes continued work; it does not appear automatically when the first version goes live.
When is doing it yourself enough?
For a simple idea, an internal prototype or a temporary page, trying AI tools yourself can make sense. When the solution becomes important for sales, operations or customer data, a professional who can ask the right questions and take responsibility for the whole solution becomes more valuable.
The strongest approach often combines both: a good consultant uses AI where it saves time and uses experience for the decisions the tool cannot own: needs, tradeoffs, security, quality and ongoing operation.
My role is to combine fast tools with clear decisions and a responsible result. If you are considering a new website or digital project in Southern Norway, I would be happy to talk about what is actually worth building.