Data science faces a technological debt issue. Its nature, however, is different from that of IT …

• As in IT, tech debt is often perceived as the result of
  implementing poor-quality code or outdated architecture &
  technology. In data science, source-code is only a part of the
  problem.
• There is tech debt in data science when scope boundaries and
  assumptions aren’t defined and documented clearly and precisely.
  Modelling processes without documentation are even more problematic
  sources of technical debt.
• In AI, tech debt can sometimes be attributed to the use of
  state-of-the-art solutions. The desire to be many steps ahead in data
  science also could cause technical debt.
• Without monitoring, maintaining, and improving the model on a
  regular basis, you will experience it as well.

In AI, it doesn’t matter whether you’re ahead (more innovative) or behind (less innovative). It does matter if it’s really worth it and logical.