Feedback from the front line of generative AI
I have had many recent conversations with companies interested in the potential for AI to support their enterprise. It has been intriguing to see the differences in thinking.
Broadly speaking, there are three camps out there.
The first camp is those folk who have heard about generative AI and ChatGPT but have yet to have hands-on experience. It is easy to forget that there are many people in this world who, quite frankly, could not give a damn about technology! These people will most likely have heard of it and may have even sat through some consultant presentation on the subject, but they will have yet to use it or see it in the flesh (so to speak).
I love these people. The response from them when they see generative AI in action is usually extraordinary.
I have seen very senior bankers go from looking at their phones to having almost religious epiphanies (St Paul on the road to Damascus) as they see the technology first-hand. In a few short minutes, they go from nothing to genuinely understanding its potential and then mentally applying the possibilities to themselves, their role and their business. It is the beauty of the technology. Its accessibility. I would challenge anyone who sees it not to start hypothesizing about its potential.
A generation of young people in education have done precisely that. They have heard about it, tried it, and, without formal training, embraced it as part of their learning. There is simple evidence of its prevalent usage among this demographic. There was a decline in use in early summer. The doom-mongers said it was evidence of the generative AI trend being over, before it was pointed out this coincided with the end of term.
Just as a quick aside, ChatGPT is excellent if you are demoing generative AI! Whoever designed ChatGPT’s interface really understands how to create magic moments. The way the interface looks and the way results are drip-fed onto the screen is marvellous theatre. Bards’ quick dumping does not have quite the same impact.
The next cohort of people are those who have been mulling over how to use the technology. They have gone through the excitement of seeing it and thinking about the possibilities, and they are now considering the options. Most people that I have spoken to in this group have realised that, as well as opportunities, generative AI opens a Pandora’s box of questions about data, security, and regulation. Having played with it, they will have experienced hallucinations and discovered the downsides. They can see options, but realise a wrong move could be career-limiting.
What this cohort needs is help sorting the wood from the trees.
I have written previously about companies such as ABN Amro, Citi, and Swift who have identified potential use cases and experimented with them. Without exception, these have been internal use cases rather than customer-facing.
Each of these companies will have explored various opportunities and settled on the one that gives the biggest “bang for their bucks”.
This approach is proving to be an excellent way of testing hypotheses, understanding the mechanics, and building a knowledge base around regulation and security implications.
There is no substitute for getting your hands dirty!
The final group are those who have been through small-scale experiments and are not considering rolling out programmes. This group is in the minority – don’t forget that ChatGPT 3.5 is less than a year old.
Companies that have been thinking about neuro-linguistic programming (NLP), AI, and conversational interfaces have had a head start. However, even with these businesses, their machinations have been very IT-led and often articulated through clumsy, non-performant chatbots. However, a few enlightened enterprises have gone through this discovery, run some trials, and have started to plot out their strategy. More often than not, they have concluded that the enterprise needs a holistic approach to make the most of the opportunity.
With them, I have been opining about the endpoint being a corporate digital twin. This twin being a comprehensive digital representation of the entire organization – mirroring people, processes, and the systems that make it work. Essentially, all its knowledge, history, and memories in one accessible place.
Small-scale tests often demonstrate the power of natural language interfaces to enable access to data in an entirely human way. But they also show that models can be trained on internal data. Often, that internal data is found to be sub-standard. Using generative AI, they have found that improving it is much faster and less painful, creating a virtuous loop.
A digital twin emerges as the data is enhanced and the model is trained. As projects move to programmes, the twin will encompass the entire business with every piece of data, decision, and action taken within and translate it into knowledge. This invaluable asset is the lifeblood of any digital twin brain. By creating a comprehensive model of an organisation, everyone – from top-tier management to entry-level employees – can interact with, contribute to, and gain insight from this knowledge base.
A couple of final points. If you need help to get started, experience matters. It is impressive how fast crypto experts have become generative AI experts, so be careful of the snake oil salespeople.
Ultimately, this is about human experience, so I recommend you don’t let IT lead the way. But without support regarding how it works and how it could work with a business, there is a danger of quickly ending up in a cul-de-sac.
About the author
He is a passionate customer advocate and champion and a successful entrepreneur.