We wrote a blog post a few weeks ago based on our early work with OpenAI's Operator tools. In short, Operator is supposed to allow AI to function against desktop and web-based apps...sort of like RPA. So, given these tools still require code to integrate them into your tech stack (writing code=bad!), it seems like there is a need to thread the needle between leveraging AI but not building a whole new pile of tech debt.
A couple of strategies we're using to alleviate this risk include:
- Writing code in a modular form that mimics how we expect AI to eventually replace each module. We actually already experienced this when Microsoft released and upgrade version of some AI that we could "throw-out" a bunch of code we were working on.
- Wait and watch. Spend the time to regularly scour the market for vendors that are solving the integration gaps.
- Postpone. Admit that certain solutions are no better than non-AI until integration challenges are solved.
Curious what others think....