Gardening and a review of AI content – a match made in heaven
So spring is here. The cherry blossoms are blooming and the Red Robins are… well… red. It also means, as I had a weekend with no significant rain forecast, it was time to get the garden ready for summer! Two days of spading bamboo, jet washing patios, de-weeding front and back gardens and mowing lawns. After my first 90 minutes of working which were delivered with a can of Monster and some great drum and bass (which is now my favourite music to make spreadsheets to AND edge my lawn), I remembered I had a Blinkist account, so thought I’d expand my mind while I worked.
Putting it on random, after listening to the first Blink about leadership, it moved onto a topic which is close to everyone’s heart at the moment, Artificial Intelligence. I listened intently while blasting lichen from my paving, but as it went on I realised something. This was….. Well…. Not good. It was 10+ minutes of buzzword bingo which quite easily could have been ‘generated’ by an earlier version of ChatGPT. It gave me nothing of note or value for me to take forwards. This got me thinking. How much good quality content is there actually out there for senior management and above about AI and how it can impact business? I’m fortunate that I have been a ‘do-er’ in the field of data and analytics before moving into the Commercial side of consultancy life, and I work at a consultancy which does this, so I know enough to know what is technically feasible and sound. However, by no means am I an expert (I understand the difference between supervised and unsupervised learning, but couldn’t train a model using these myself, if that makes sense). So, over the next day and a half, I randomly listened to a mix of content that is out there, to try and see what was good and what was bad, to establish key takeaways which I have summarised below to help you.
DISCLAIMER: I did not do this in a systematic fashion. I put some keywords in, and listened to the content that came up at random across a number of platforms (YouTube, Blinkist and others), skipping anything that was unrelated. I was working with my hands at the time, so I didn’t write down the names of the good or bad; this is more about the general learnings from the content out there at the moment.
Beware the 'Snake oil' salesman
This is a bit of a sensationalist header to start with, but there were a number of ‘What C-suite people need to know about AI’ themes of content which won’t really help you much. They were awash with grandiose language about ‘transformative capabilities’ or ‘iterative delivery’ without explaining what, where or how this could be true. When listening to this type of thing, I always came back to the Einstein quote of ‘If you can’t explain it simply, you don’t understand it well enough’. AI is new (or at least relatively new), and knowledge is a sliding scale from nothing to everything, but if the person you are listening to can’t make it clear to you (in language you understand) how, what and why AI will help with productivity, its time to move on.
Be clear about the outcome you want to impact
The first practical learning may seem obvious, but having a clear focus on the outcome you want your AI to influence is massively important going into any programme around this area. Though AI is great at finding patterns, using it for broad-themed explorative analysis can be a challenge. Anything it produces will need very strong and deep evaluation from a business standpoint and to turn into action. Focussing on a particular outcome ensures better value from the work being undertaken.
The people are as, if not more, important than the AI itself
Aside from many of the pieces delving into what AI means for the job market, all of the good articles came back to a key component of your organisation’s AI success being driven by having relevantly skilled people and people who have a strong knowledge of your chosen area for improvement involved in the process. The skill and ability of your data engineers and data scientists on their own will not be enough to make the project a success. A number of pieces of content also point out that putting tools like GenAI in the hands of your experts in a given field gives a greater impact, both from productivity and output standpoints, than putting it in the hands of those less experienced and qualified. Experts will always be able to pick out and improve the outputs of AI to provide a better outcome.
Begin iteratively with the data you have, work towards transformative
One of the largest challenges within an organisation aiming to bring AI into the heart of its success-driving strategies, is ensuring all of your data is structured correctly. This covers everything from having the relevant consent for its usage, it having the correct granularity and structure to perform the analyses that you want, through to having the architecture and the ability to perform actions from it in a timely manner. Getting all of your data past this hurdle (which is kind of like the ‘Great Filter’ for AI) can be both time consuming and hugely expensive to do at once across customer service data, digital data, transactional data and into all of your engagement tools and solutions. As a result, don’t think ‘transformative’ has to be the required outcome from the very beginning, and reduce the work needed to create iterative value now. Look at your use cases and your most mature and available datasets to find things that you can do today, which also start to get these datasets ready for your larger, and more transformative, tasks in the future.
Don’t be afraid to 'build on the shoulders of giants'
To paraphrase an ex-colleague of mine, ‘Why build an NLP model for yourself, when you can build on the shoulder of giants by using their models’. Tailoring an existing model and training it on your data for your outcome can save multiple iterations (and their associated time and cost) of development.
So that was a selection of the best learnings I found from a weekend of garden labour, getting a tan and listening to the great, good and the shockingly misguided musings on AI across various platforms. If you want to talk about any of these please do get in contact, or if you want to dive into greater detail on the how, where and why of Artificial Intelligence, let me know and we can arrange a time with one of the bigger brains than I at Station10.