AI, Where to get started | Issue 5

Welcome to Issue 5 of unfolding artificial intelligence, we are delighted to have you as a reader of this newsletter.

This issue we have a theme around getting started in implementation, part of our AI for Leaders approach.

We explore what to consider, and where to begin and provide some guidance as to how to start, starting…

Paul, unfold:ai co-founder (and newsletter editor)

p.s As this is a new newsletter, we would really appreciate it if you spread the word and did a social share. The icons are at the top of the newsletter.

Implementing Artificial Intelligence

Getting started in moving beyond personal exploration of AI and into the everyday workspace is going to be complex. It is a change programme, and an innovation programme combined. Neither make for straight forward activities.

Here are things to consider.

It’s all about the people

Firstly there is getting over the fear, uncertainty and doubt questions. Will it end the world? What does it mean to my job? To everyone’s jobs?

Then there is finding the explorers in your organisation who want to innovate and lead change.

Then there is giving them time, opportunity and space to learn these new skills and techniques.

For context, even if there was 100,000 skilled Prompt Engineers 3.0 with 1.8million medium sized businesses in the UK, that’s less than 1 hour of resource, per business per week.

A safe place to try

Innovation and change needs a safe place to try. To prove its worth it needs to have impact. This is a CEO moment. High impact and safe to fail, are not going to coincide in most businesses.

Unless that business is broken, in which case rolling the dice to save it has been tried many times in digital transformations as a last ditch attempt, they fail. Don’t do that, maybe implement AI into some of the ‘C-suite’ decision making to help out.

Pick a safe place, a safe task, a mundane task, a task that has repetitive volume that no one likes doing, or is done poorly or needs predictive outcomes.

Pick a solution that is mainly done outside of the enterprise systems, excel based, user systems. Avoiding system change is good for exploring.

How about customer feedback sentiment analysis?

Measure where you are now

Document, and measure the current process. You need to know effort to do (realistic, take a measurement), you need to know accuracy rates, success and fail rates and also frequency. Frequency may be low because its hard to do.

The AI unlock might be in making it able to be done when before it was not possible.

Start

Start. Use what you know, use what you have, and where you are.

Measure it. Write it down.

Be critical. Is it working? Has it increased value for everyone, for good ?

End the experiment

Know when to call it, this is an experiment. It is not implementation. Do not implement experiments into customer production, even if it is tempting. Just take a moment to industrialise. Remember Pak N Save, good idea, given to humans… who might, just might do crazy things with it.

Start a chat with your data

If you have chatGPT 4, and a plug in or two. You can ‘chat with your data’ which allows for an interesting way of surfacing value which may be lost in the depths of those marketing PDF’s, product data sets etc.

The next step from ‘one on one’ chats with data items is to ingest (or add a data corpus) your enterprise data into a safe repository and then layer a chatbot on top which has a private view of this data, but draws on the LLM beyond it to provide the language processing.

It’s a fairly straightforward use case and is very implementable even for a smaller enterprise. We have put one on www.unfoldai.co.uk/articles which uses one of these new generation of chatBots. Look for the chat Icon bottom right (the usual place for them).

We would recommend building one to sit internally on your systems to help with data discovery.

The next step beyond chatBots, would be to look into how to deploy some autonomous working agents. This is a form of process discovery and automation.

Remember Content Moderation

If you saw the news about Pak ‘N’ Save. There are a couple of lessons to unpack (pun intended).

The intent for a helpful, resource saving chat assistant than can recommend recipes based on items in the household is a good idea. Not putting any guard rails or moderation into the flow. Not a good idea, someone was bound to misuse it and send that to the press.

There is guidance out there on how to do this. If you are playing with chatGPT on a broader level and think you may need an approach to moderation techniques, this article is a good starting point.

At unfold we would classify it as advanced prompt level. So don’t be disheartened by wondering how it is working.

Google Deepmind and artists

Google has teamed up with artists to explore alternative images and visualisations on several of the key themes. It’s beautiful work, it’s also available for use.

The art conveys AI's potential to transform society while promoting inclusive development. From biodiversity to neuroscience, the visuals make complex AI applications accessible. This novel approach to demystifying AI technology can spark inspiration about its thoughtful integration into our lives. When people can envision positive AI futures, we take the first step toward building them.

Quick things to check out

Thoughts on regulation on AI in the UK

Aitubo, art and asset generation for game making

A prompt to try in chatGPT (and other LLMs)

acting as a social selling agent on instagram, suggest to me 10 ways I could photoshoot some converse chuck original trainers, describe the shoot idea, the tagline, and the body copy,  be creative but practical in your suggestions 

Inside Premium this issue…

• A summary of AMD’s Research into IT Leaders and AI

• The Peak of Inflated expectations,

• Advanced chatGPT used for Risk Identification,

• Sentiment Analysis, a quick start guide,

Subscribe to Premium to read the rest.