What do we do about AI in 2024 at ACME Co?
Actionable advice to get started and why you should, even if you think maybe you shouldn't
As 2024 ramps up, many of the conversations I’m having revolve around the same theme. “All this AI talk is very nice/interesting/scary/overblown. But what should my company/organisation do with AI in 2024?” Specific advice is very sector and company life stage specific but I believe there are broad themes that are applicable to pretty much everyone.
Dive in and skill up
Absent a crystal ball, I don’t know where AI will be, in any particular industry this time next year. (And neither does anyone else BTW. If folks claim that they do, I’d see that as a very strong signal to broaden the circle of people you are seeking advice from!) Not knowing is not in itself a negative thing, it’s the reality of a space that is moving fast, has significant geopolitical complexity and a not insignificant number of ‘supply chain’ constraints (a non exhaustive list includes skilled labour, GPU capacity, data availability, legal challenges and shifting regulation).
However given the broad potential for impact and the billions of dollars invested and looking for returns, I am betting that it’s unlikely that your industry and your company won’t see some change, and most likely change that will accelerate over the next decade. A change I lived through and hence like to reference is the emergence of internet banking in the financial services industry. Some readers will not remember passbooks and bank branches with absurdly short opening hours but let me tell you, it was a drag. And one my parents occasionally had to organise their busy lives around because there was no alternative. Almost inconceivable then as a small child clutching my Kashin that ‘few second’ digital transfers would be at your fingertips on an interface that fits comfortable in your back pocket. This time scale of change (perhaps circa 15 years from idea to ubiquitous) could easily be repeated with generative AI driven change where it is likely to be ‘much more termites than tornados’ for many business models.
So, use 2024 to get over the overwhelm and learn enough about the traditional and emerging techniques and capabilities in data and AI to be able to confidently make knowledge-based decisions and investments in 2025 and beyond.
And don’t inadvertently limit your scope of learning, as the potential impact of using and creating data and AI products is broad. Just some of the areas I would encourage you and your team to consider learning more about are:
algorithms - Although deeply mathematical, there is nothing magical about the algorithms at the heart of traditional and emerging AI and 2023 saw a lot of accessible explainer articles being widely read and shared. These won’t make you an algorithmic wizard but they will help you understand likely capabilities and limitations of these new AI tools. Find two or three and form a reading group to test your understanding.
working with data - If you want to use AI in your business there is no shortcut to understanding the data that you have available, very likely capturing more of it and potentially storing multiple copies of it in several formats and places. Here’s one place where lived experience can really help you out if you go looking for it. Folks have been cleaning, structuring, moving and processing data for many decades. Lots of the techniques used have changed but in ways that are (important and) evolutionary rather than revolutionary. Perhaps start by putting aside time to speak with folk in your org about the biggest challenges they’ve encountered working with data in the last 24 months. It will give you a good grounding in some of AI’s ‘immovable objects’ in a familiar context.
human interface design - Banish from your mind the idea that AI products are all about the algorithm, just plug into Open AI and away you go. As with any durably successful product or service, you have to step back and understand the end to end. Nowhere is this more true than in the bit when humans - your users - enter the picture. Both the probabilistic nature of algorithmically generated output and the growing awareness and distrust of AI augmented systems make the interaction a complex beast to understand. Best practice is still emerging so be curious and begin with an open mind to think about how workflows evolve as pieces of them are automated and semi-automated with AI augmentation. Consider creepy lines and how to avoid them. And think about where you are accidentally taking away opportunities for customer connection.
procurement - You may well not have any intention of building your own AI based systems but that doesn’t let you off the hook in terms of learning about data and AI. Take a deep dive into your current procurement processes. Do you have questions specifically focussed on the collection and use of data (both yours and that of the vendor)? Do you have a clear policy on copyright (in the model and/or of the generated output)? How does the overlap of AI supply chain and modern slavery potentially impact your ESG-driven procurement commitments? What about climate targets? (AI can take a lot of both water and electricity)
security - Do you have data that may now make you a honeypot for hackers? If you were to make use of, perhaps a RAG augmented Generative AI use case for customer support, would you open a new attack vector and what plans can you put in place to mitigate? If a third party gained access to your Open AI API tokens, how soon would you find out and what level of audit would you be able to perform over the unsanctioned activity?
legal & regulatory - Do you operate in more than one jurisdiction? Will you be able to meet all current and emerging regulation in an auditable fashion or will you pick one set of regs to follow? What is your position on data sovereignty? Will you be able to persist with storing and processing all your data in one country or will regulation drive you towards a distributed processing and storage approach? (with potentially deep implications on your application architecture), How is your patent portfolio looking? Are you thinking of applying AI in a ‘high risk’ use case (e.g. facial recognition)? Will that remain legal in all your markets?
trust & authenticity - If 2024 is the year that junk content overwhelms the internet, what risks and opportunities exist for you in that transition? Will there be a market position for more artisan approaches? How will social norms shift to accomodate AI generated content? What will the etiquette be for human-bot conversations?
Consider giving AI literacy (or whatever you chose to call it) a standing space on your leadership, executive and board meetings. Make the learning directed with in-depth content revision and discussion and independent expert speakers who can lead a deep dive into particular areas.
Choose a place to invest and get down to the brass tacks
However if you are like most humans and every busy company I’ve met, desk research and listening to the experience of others will only take you so far. So I would also choose an area for your organisation to gain practical experience with AI in 2024 and fund a program of work.
Make the cash / opportunity cost big enough to hurt (otherwise, let’s be honest, the exec & board won’t maintain focus over 12 months of failures) but don’t bet the company on it.
Nothing teaches any of us like the actual practicalities of real problems and the unpleasant feeling of seeing invested cash seeping away. So invest to learn by doing … and make sure that ‘what we’ve learnt’ is genuinely seen as a valuable payback from the investment.
Why?
Because then you will start 2025 from a place of confidence and with actual lessons learnt.
Lessen your dependence on the advice of vendors, who are often lovely people but are always selling a product. Which definitely isn’t to say not to buy, just to emphasise the ‘purchaser confidence’ that comes with being a truely informed buyer.
Maybe all that you will learn is that AI today is ill suited to your industry and can’t provide a commercially viable business advantage at this stage. But you’ll know what to keep an eye on, when / under what circumstances to revisit your decision and just think of all the money and opportunity cost you will save by your informed decision not to play in the AI playground in 2025. That’s a lot of knowledge gained from even the ‘worst case’ outcome! . Given the lack of corporate spending discipline that can surround hype cycles, even that knowledge could well be company defining.
Maybe you will build a smashing new product and change the future trajectory of your whole industry!
You will definitely learn a lot about market dynamics, big vendor sweet spots and achilles heels, (they aren’t all good at everything), how to negotiate data contracts that don’t suck, how to attract and retain talent (possibly by doing the opposite of what you tried in 2024) and how to get your questions answered in plain language that doesn’t seek to distract attention from some of the less robust or less salubrious parts of the AI building blocks.
Here’s to your learning and success.
Till next time
This weekend I shall be tackling the slightly awe inspiring prospect of building ~90 lineal metres of retaining wall for a spontaneously created terracing below our top dam. I’m excited about the orchard I’ll establish in winter 2024, hopefully that will keep me going through all the required rock shuffling.