Scaling generative AI
I just read a McKinsey paper on a technical perspective for scaling generative AI. The author’s main argument is that generative AI will disrupt a company’s data platform ...
…. and that companies have held back on adopting it, even though there are promising results across industries.
Sadly, the article itself is a little hard to read as it tries to focus on the two major areas of interest at the same time:
using data platforms to support generative AI use cases across the companies
using generative AI to improve the data platforms themself
While both of these areas of interest center around generative AI, it is best to separate them because different parties inside a company usually own them. Use cases are driven by product management and product teams and are sometimes supported by data teams. While the data platform is almost always owned entirely by a data team, often without product management.
Nevertheless, the tips to scale gen AI use cases have merit; here they are:
Improve the source data for your use case by investing in good data pipelines.
Improve your source data by investing in labeling and manual/ semi-automatic data improvement.
Upgrade your data orchestration for end-to-end visibility, and invest in it like in any product.
Start to treat your gen AI products like real products, including all the software development best practices your company already knows.
