Getting into Operational Control of your LLMs

Getting into Operational Control of your LLMs

Have you noticed a change in the collective mood around AI?

Even some of the hard-core evangelists are scaling back their messaging. While the critics remain in full-throated denunciation of the hype and lack of value they think has been thrust upon us over the last few years.

Many billions have now been poured into trying to generate an epoch level disruption in how we work and live. But so far, the stats say that despite full thruster mode, lift-off is still hovering over the launch pad.

Of course, this is as prophesised in Gartner’s evergreen insight that things turn sour before benefits and value broadly flow.

The trough of disillusionment is a well-chosen phrase. The language radiates the deeply emotive state of mind many are currently feeling.

Where Next?

Part of our foundation understanding around new technology is to recognise the inevitability of this phase and what follows.

The map currently shows our collective location in this cycle.

Article content

The reason why this always happens is well captured in the following nugget of market wisdom. It offers a more realistic timeframe in which most organisations grow the competence to extract value from a new technology.

In the short-run, it (any disruptive technology) will be less impactful than hyped.

But In the long-run it will prove more transformational than envisioned

In other words, this change in mood in the market is fully expected. So, if that’s the case, do we also know what usually happens next?

Pretty much.

The fever breaks and people get on with the job of adapting to the impact that a particular technology will eventually have.

Two things happen.

The technology matures in terms of cost, reliability, and ability to customise.  

Meanwhile organisations settle into long-term operational re-modelling rather than posturing about being seen to be first.

The Serious Work Begins

Brainfood Training is ready to help those making the transition:

  1. We are focused on foundation understanding to enable informed decision making
  2. We are enabling leadership and influencer minds to re-imagine customer engagement in an AI-first context as initiation for relevant strategy planning
  3. We will guide you in pursuit of AI Readiness. From issues of Responsible AI that customers and legislators now demand to the topic of this post - Getting into Operational Control of your LLMs

If you are in a customer facing role e.g. Marketing, Sales, Service, the technology you use is now powered by LLMs. They are great and are potentially transformative. But they are also unreliable. And that’s before they start to become common place in complex agentic AI use cases.

So what? That’s a tech team issue.

Not any more! It’s a shared responsibility.

As operation leaders you need to understand LLMs to be effective in your decision making around them. The second reason is your growing accountability for ensuring responsible AI or whatever label is used internally to meet customer expectations and growing demands of legislation. All part of becoming an AI first organisation.

If you are ready to tune into this emerging reality, then a great place to start is getting to grips with what you’ve got in your LLM cupboard and how they are being managed.

Article content

Here's the first of thirteen topics you and your LLM vendors should be discussing on a regular basis. All the questions are figured out for you. All you need to do is invite them in for a discussion.

The full set is available to download plus a bonus briefing on LLMs and one of their key behaviours: hallucinations.

Takeaway

Learning how to use disruptive technology safely and effectively starts by understanding it.  

 

 

 



Wim Rampen

I set organisations in motion

5mo

Hehe.. been there, done that, got the t-shirt, learned something new though, until next time 🤪.

Like
Reply
Graham Hill (Dr G)

Customer base optimisation | Operating model redesign | Change facilitation | AI augmentation of work | Opinions my own

5mo

This problem isn't new Martin Hill-Wilson. Solow's Paradox captured it in the 1980s. It plagues all general purpose technologies as Brynjolfsson describes... https://ide.mit.edu/sites/default/files/publications/jcurve.pdf

To view or add a comment, sign in

More articles by Martin Hill-Wilson

Others also viewed

Explore content categories