Still a Lot of Talk About GenAI

It's been another week where there's been many conversations about generative AI, but production use cases still feel a little way off.  A lot of businesses are still trying to digest the myriad of options that are presented to them. For instance, I heard of one business with potential access to around 200 different GenAI technologies.  I suspect some of these will not offer huge benefits and will never see the light of day, but it all creates a huge landscape of options that a business still needs to evaluate.  Choice paralysis might be an issue too.  Then you have questions about accessing these technologies, who should have access, what data can they use as a prompt, and who’s checking outputs? These are very important questions for enterprise use of GenAI, especially when you also need to comply with the recent EU AI Act.

From my recent conversations, it also seems apparent that most businesses want a broad range of employees to have access to GenAI, not just Data Scientists and coders, which is correct.  To truly grab hold of the many advantages GenAI can bring, it needs to be democratised across the organisation, not just limited to a small silo of specialist employees.

To me, it’s clear that an orchestration layer is vital here to manage all of this:

• Model access
• Data access
• Collaboration
• Governance
• Adding LLMs to existing data flows/projects
• Deployment

Maybe more?

And despite what I said in my first sentence, there are enterprises out there that have a lot of this figured out.  And hats off to them. Taking bit of a risk, pushing their capabilities.  And whilst their first use cases have no doubt gone through rigorous testing, approvals and monitoring, they will no doubt only represent an initial version of what will likely see several iterations in the future as the technology and their understanding continue to advance.