Gen AI Development: Overcoming Common Pitfalls

A recent McKinsey article says that the firm’s experience working with more than 150 firms in developing AI programs in the past few years has revealed two common problems that almost always surface:

–  Failure to innovate: Process constraints, lack of focus, and cycles of rework that quash innovation. Teams that could be solving valuable problems are stuck re-creating experiments or waiting on compliance teams, who themselves are struggling to keep up with the pace of development. In our experience, roughly 30 to 50 percent of a team’s “innovation” time with gen AI is spent on making the solution compliant or waiting for their organizations’ compliance requirements to solidify and be practical. Teams work on problems that don’t matter, duplicate work, and create one-off solutions that can’t be reused and often fail to unlock real value.

– Failure to scale: Risk concerns and cost overruns that choke off scale. For the few solutions that show real value potential, enterprises largely fail to cross the chasm from prototype to production. Security and risk concerns (including reputational risk) when scaling gen AI applications are handled individually and become too large and expensive to overcome.

The article says these problems often occur sequentially, as companies move from pilot programs to the operational stage and they can quickly derail entire Gen AI programs.  It also says that successful programs share three things in common.

– First, a secure and compliant self-service portal, which provides a single access point to validated Gen AI products. This permits developers to incorporate preexisting application patterns and begin work on their specific solution immediately.

– Second, an open architecture approach that allows developers to integrate and easily swap out reusable modular components.

– Finally, the implementation of automated guardrails in the Gen AI platform to mitigate risk, manage ongoing compliance, and provide cost transparency.