A feature that took a small team months to build can now be approximated by a competitor in days. That is uncomfortable for any founder whose pitch rests on a clever interface. The answer is not to panic about copying. It is to be clear about which layers of the business are hard to copy and to protect those first.
The first layer is the technical contribution itself. Where a venture has solved a genuine technical problem, such as a new algorithm, a new sensing method or a new way of reducing energy use, a patent can protect that solution in exchange for disclosing it. The timing rule is the one founders break most often: an invention generally has to be filed before it is publicly disclosed, and a conference talk, a demo video or a pitch deck posted online can count as disclosure. Long experience with patent families across areas from cybersecurity algorithms to energy optimisation tends to teach founders this early.
The second layer is know-how and data. Process knowledge, training data and tuning methods are usually better kept as trade secrets than patented, but a secret only counts if the company behaves as though it is one: access limited, agreements signed, and documents marked. The third layer is the slow-building one, which is quality assurance, certification and customer trust. In regulated or safety-sensitive sectors, a competitor can copy a feature and still be years from passing the same certification.
Sequence matters as much as selection. A venture should validate its business model first, using customer discovery and a business model canvas, and decide what to protect afterward. Filing for a patent on a product that no customer will buy is an expensive way to learn the lesson.
Two housekeeping checks are worth doing early. Confirm who owns the work: code and inventions created by employees, contractors and university collaborators can belong to someone other than the company unless agreements say otherwise. And take advice on how intellectual property law treats material produced with the help of AI tools, since the position varies by jurisdiction and is still developing. None of this replaces a qualified IP lawyer.
A useful exercise: write one page on what you would still own if a competitor had your interface tomorrow. If the page is short, that is the work for the next quarter.
