Implementing AI in the enterprise(1)

廖原德
廖原德
Nov 7 · 2 min read

As an AI engineer, I may encounter some problems which can’t easily be dealt with by my data analyze skill but lean much on domain experts’ professions. The following is some of my personal observations

At the beginning of implementing AI in the enterprise, corresponding colleagues will get nervous and don’t know how to work with AI Technician even if they want to show their best cooperation. Thus, listening their needs is the first thing. Only if they felt that we understood their problems, we could start to provide technical solutions.

In my personal experiences, thorough discussion is necessary, as standard technical solutions may not satisfied them. But there is one thing worth to remarked, almost every discussion is inefficacious, for the lack of understanding of the profession of each other’s fields. Hence, the most efficient approach is the means which looks like compromising — execute the proposal which is congruent with domain experts’ opinion “first”. Fully support them to develop solutions by AI even if the solutions are imperfect in technical perspective. After the feedback came out, support them to correct the solution, which means to give up the dominance of the discussion, only providing the suggestions to help solve problems. Surely, things will get tackled at the end with slow work speed.

You might suspect that this is so inefficient, does real enterprise work in this way? In fact, this is the only way to solve a difficult business problem by AI without severe conflicts between wide-range interdisciplinary persons. At last, you will be surprised that how much they appreciate your help and the power of AI, even though you might think you didn’t do much during the process. It’s what I saw in the enterprise when people implement AI technique.