Keep Calm and Be Prepared
Implementing AI tooling in key operational processes are a must for all organisations. Doing so does exposes you to a new type of operational risks that, if not mitigated, may cause significant damages.
In this post Sven explores the nature of those risks and what measures an organisation should take to build resilience. And as it turns out, these measure are easy to implement so that, as long as you are prepared, you can roll out AI without needing to worry.
Carefree AI adoption made easy
A few days back, I had a conversation with one of my close friends on the backdrop of my earlier post that argued for the role of astute craftsmanship in tomorrow's organisations. It gave rise to several concerns that really have kept me thinking since. Essentially, the concerns boil down to: how do you protect your organisation against instant withdrawal from the AI service backends, if or when push comes to shove?
As in most complex cases, there is not a single clear-cut answer to this question. But that does not mean we have to worry. By understanding the potential risks, we can formulate easy to use stratagems that will build the resilience of your organisation. And as it turns out, these tactics will at the same time help you maximise the impact of your AI implementations.
In this article I will investigate the two most important scenarios in depth. These scenarios can create grave operational and even continuity risks when AI is rolled out in mission critical applications. But first lets establish why ignoring AI technology is not the recommended path forward.
The need for adoption
Before diving into the potential risks and how to mitigate them, lets reiterate the most common reasons why AI technology is a must. Most specifically that of the large language models (LLMs), which is crucial to implement for organisations.
Speeding up is about removing the gaps
The first and most obvious reason is found in the potential efficiency gains that are to be had with the strategic implementation of AI technology in existing procedures and processes. Automation of the menial parts is sure to create savings in labour needed. The same effort should lead to an improvement of the challenges that your workforce experiences, as the drudgery is now out of the way and they can focus on the more interesting aspects of their work.
Remedy existing shortages
Most industries have seen a shortage of qualified labour to fill their vacancies adequately. It has been one of the major challenges for over one in four business in the Netherlands for the past years. The expectancy is that this gap between people and vacancies will only grow in the next years. Research published last year by TNO indicates that 50% off Dutch organisations are already struggling to find the people they need or expect that this will happen in the next two years. The costs associated with this challenge are not to be underestimated. There are the direct costs of recruitment and of time spend in the candidate search to begin with. On top of that, there is the opportunity cost due to not being able to execute on all the value that could be captured, given the people to do so. Thirdly, there is the real danger that positions that remain vacant for longer stretches of time take their toll on the people in your employ, leading to risks of long term absences or attrition, both exacerbating the situation at hand. It is obvious that the aforementioned efficiency gains will alleviate this challenge, as part of the efficiency gains will lead to lower need for headcount.
Keeping up with the neighbours
Last, but not least, is the factor that other organisations are also experimenting with the implementation of AI solutions in their processes and products. This will include competitors in the same market as your organisation. Though not all these innovations will resound with your customers, you will need to keep an eye out on what delights your customers so as not to become obsolete in your product or service offerings.
AI is a key geopolitical factor
Lets face it: for us in Europe this is not a theoretical exercise. We do not have truly independent viable cloud infrastructure that is fully under European legislation - and European alone. Let alone that we have massive infrastructure to train and run large language models, the staple of our ongoing AI revolution.
For organisations that want to adopt AI solutions into their operations, this necessitates to work (mostly) with either US or China based companies. Unfortunately, in the current geopolitical climate, threats to continued access to cloud infrastructure and solutions is very real.
Another angle in the geopolitics is that European organisations no longer get access to the latest and greatest versions of the solutions. Last year, Meta decided to no longer release new LLM implementations in the European market, citing pesky privacy and copyright laws, to be exact. It is not hard to fathom the impact a punitive political directive with similar consequence, i.e. no access to the latest and greatest, will have. Not being able to leverage the state of the art, whereas commercial counterparts in the two computing superpowers do, is sure to negatively impact the viability of enterprises being all in on automation with AI. For the European market, to be frank, this would be an existential threat for its ability to viably compete in many domains.
Admittedly, the likelihood of either scenario fully transpiring is at the time of this writing slim. However, as most leaders will be aware, the relative stability of the 1990's up to the 2010's has made place for a more turbulent, potential unstable, global climate. As such, it is good to take these scenarios at heart when planning for long term investments.
AI is cloud infrastructure too.
Radio Kootwijk inspired the Dutch discussion on cloud sovereignty.
To emphasize that this challenge is not unique to the adoption of AI, we would like to refer to this excellent summary of the risks regarding cloud infrastructure more general, by Bert Hubert. These very real concerns and risks have already been discussed in the Dutch parliament. The reason were the wide reaching plans to move most, if not all, Dutch governmental email- , file-, and other common services in to cloud premises. Recently, the decision was made to keep at least email and file services local despite the technical challenges that poses. And this is not a Dutch-disease. Denmark is already embarking on the drastic course to cut out all non-European cloud solutions from governmental agencies, including office applications.
Evolving business models of service providers
Even if the global situation calms down and geopolitical tensions lessen, there is the final spectre that looms in the massive and complete adoption of LLM solutions. The working hypothesis in the AI industry is that this will be a 'winner takes all' race. In other words, the major player vie for market dominance, as part of the fly wheel to ever faster iterate on their models. The market dominance, for now, is not focussed on monetization, but rather usage. As model usage is part of the learning fuel needed to enable the next level of model complexity.
When looking at the massive amounts of energy that are already needed to feed the power-hungry searches, it becomes clear that the massive investments being made by the major players in this field are funneled into subsidizing the prices. Ultimately, the institutional investors do this, knowing that when the victor is declared, prices can, will and must go up. Even if it turns out that we do not end up with a monopoly but with a duopoly, organisations still need to prepare for massive increases in costs. Here too, we have already seen prime examples of what that can mean for dependant outfits that have gone all in.
Keep calm and be prepared
With the potential impact of these risks in mind, the question beckons what organisations need to do to prevent sizeable impact if one of these scenario's materializes. Fortunately, there are several good actions implementable that will shield your organisation. My advice is to implement the following measures:
1. Minimize vendor lock-in.
Modular building by my son.
When you formulate and execute a multi-vendor cloud IT strategy you are building your first line of defence against price raises or even worse, gouging. When executed properly it will prevent hard vendor lock-in and solution mobility, so you are able to migrate complete solutions with limited effort if confronted by unreasonable new prices. This will not shield you of a scenario where one of the dominant powers (again: USA and China) implement technology access restrictions on industries or geographies. But this strategy for the basis on which the following stratagems build and those will help construct that line of defence.
2. Own your competitive edge.
Keep the keys to your castle in your own hand.
All organisations differentiate themselves from their competition. This differentiation can be achieved through organisational or operational excellence. For a lot of organisations this edge is achieved through the specific application of technology that gets developed in house: their intellectual property or IP. When you own this IP you are and will remain in control. In doing this, it is key that you keep adhering to the strategy formulated in the first measure and prevent vendor lock-in with third parties, but especially parties in differing jurisdictions for your own base of operation.
3. Become Beyond Agile
Agility on steroids is needed in order to have an organisation that is nimble and reactive. Craft mastery is needed to be in control, mitigate issues and pivot quickly and effectively. Remove red tape structures and empower your craftsmen that show true agency. I touched upon the revival of craftsmanship in my previous blog post. The underpinning of the next generation of successful organisations will be the astute craftsman. Currently we are working on fleshing out in detail how this will impact organisations. In particular will we look into what guidance can be given to implement this complex task successfully.
4. Adopt AI deliberately
Investigate where AI adoption can help you solve real business problems or yield actual value. Choose, execute and evaluate. And whatever you do: prevent that everything, everywhere at the same time is AI-ified. Uncontrolled adoption will result in significant cost increases in the not too distant future, where not all these costs can be justified in savings or additional revenues. Most organisations have already experienced something similar when migrating to the cloud. This has even made some organisations decide that their best and most viable option was to bring their infrastructure and platforms back to dedicated owned hardware in dedicated server rooms.
5. Don't believe the hype
Lets be clear: LLMs are transformative in the way we will create results. But as with all new technologies, there is a lot of hype and even snake oil being peddled. The best guard against the hype is to have some grizzled veterans under your fast-dial, ideally from within your own ranks and otherwise in the form of an independent advisor that can help you separate rhyme and reason.
Conclusions
It goes without saying that adopting AI solutions now, is a must for organisations. There are three major reasons that make this imperative:
It produces significant efficiency gains that directly translate in hard cash.
It alleviates personnel shortages that are encountered across all industries.
Customers expect a growing level of AI innovation from organisations they interact with.
With these great boons come risks that will be new for organisations. Even more so that with recent cloud migrations, the means of production are now no longer local, but rather, and completely virtual. With geopolitical tensions rising, access is all but guaranteed. And even when access remains, there should be an expectancy that the major AI players will raise the prices significantly in the next few years. If the expected 'winner takes all' scenario comes to bear, these costs may become debilitating for a lot of operations.
Organisations do well therefore, to take measures now, so that they can reap the fruits of AI, while not becoming incapacitated when ease of access is lost. We outlined five strategies that are easy to implement, that will help you increase the resilience of your organisation:
Adopt a multi IT vendor strategy and transform your IT landscape accordingly.
Identify what sets you apart from the competition and own that technology.
Transform to make craftsmanship front and centre of your organisation.
Prioritise material impact over shininess when prioritizing AI implementations.
Remember: not all that glitters is gold, so separate hype from the valuable.
If any of the points above made you wonder how this would work for your organisation: get in touch. In all cases, we at Hephaestus AI can work with you and your teams to map out your situation. Leveraging our experience and if so needed that of trusted partners, we can also help you materialise these bespoke solutions in your operation.
Implementing AI will be an exciting journey for all organisations. And like all journeys, it is most enjoyable when one embarks well prepared. So, Keep Calm and Be Prepared. Happy sailing!