Something old, something new
There is nothing new in the world except the history you do not know – Harry S Truman
So, you’re thinking of adopting an AI? That’s a big commitment. Some of you are still busy with that Digital Transformation Programme you kicked off a few (many) years back. Are you sure now’s the time to be taking on a second child? Well, to be frank, it’s too late for second thoughts. Whether you know it or not, your organisation is already on the AI adoption path.
As luck would have it, the approach to adopting any new technology is tried and tested. So many new inventions have entered the world of business over the last few decades. Unfamiliar and mysterious though it can seem, AI is just one more, and the same rules apply.
One step at a time
The adoption of any new approach or technology is a process of continuous evolution, but for ease of understanding, I like to break it down into 5 stages. If we call the thing to be adopted “X“, these stages are as follows:
- Discovery
Your people are trying out X within the workplace, but there’s no clear ownership, and no concrete measures of effectiveness. - Experimentation
With guidance, you are embarking on experiments in using X that measure outcomes to determine where it can be effective. - Adaptation
You are integrating X into workflows supported by evidence of effectiveness and adapting your business to maximise value. - Consolidation
Where value is proven, you are ensuring that the adoption of X is consistent and complete across your organisation. - Maintenance
You have fully adopted X into your processes and products, where valuable, and keep up with advances in the technology.
1 – You are here
If we take these stages and apply them to a well established technology such as office productivity tools, we can easily see that nearly all organisations are at the Maintenance stage and have been for some time. If we look at something like the agile approach to product development, the picture is more mixed. Most organisations are in either the adaptation or consolidation stages, a few are in maintenance, and a not insignificant remainder are still in experimentation or even discovery.
You’ll notice there is no stage zero. The reason is simple. Once a thing exists, someone somewhere in your company will be using it, and sharing it with others. You are, by default, in stage one (the discovery stage) whether you want to be or not. It is therefore imperative that you get to stage two as soon as possible. You need to start experimenting in order to work out where you should be using the technology, and at the very least, take control in areas where you shouldn’t.
With any new technology, there will be a surfeit of “helpful” advice explaining where it’ll make a real difference to your organisation. Usually this advice comes in one of two forms: “‘Tis witchcraft! Avoid at all costs”, and “this is the silver bullet solution to all your problems”. Not surprisingly, neither of these it in any way helpful. Unfortunately, the same is true for much of the more balanced advice. Although it might be carefully considered and correct for the context in which it is given, you have no way to determine whether your context is the same, and no experience on which to judge its veracity. That’s why the first step to adoption involves getting your hand dirty.
2 – Experimenting with purpose
No new technology is without its uses, nor is it free from risk. What you need to start finding out, as soon as possible, is which is which. To do this you need to run some experiments. In AI adoption terms, an experiment involves testing it out in a real scenario. The scenario you choose needs to be one where you can see ways in which it might help, or where others have indicated that they’ve seen success. Preferably both. The trick is to start small and aim to get to an answer as quickly as possible.
The point of the experiment is to compare, so you also need something where any improvement is measurable. If you want to move quickly that measurement also needs to be fairly immediate. Only then can you start to see past the hype and the fear to where the true value lies. Once you find this value, you can expand the experiment and turn it from speculation to investment. Apply what you’ve learned to larger problems and over longer timescales to harvest that value.
A word of warning: Although this will allow you to discover places where AI can deliver immediate improvements, when it comes to productivity what it won’t give you is a prognosis on the long term effects. Sometimes, short term gains can turn into long term issues. More haste, less speed turns out to be true more often than we like. Carpenters will often say “measure twice, cut once” and the same is true here. For some of your experiments, you might want to extend the test for a longer period to make sure the gains you’re seeing are sustainable before committing wider application.
One cautious experiment doesn’t necessarily put you at stage 2, but a deliberate and encouraged approach to testing AI in a variety of potential situations does. Once you’re here, and your experiments are starting to yield fruit, it’s time to consider a move to the next stage.
3 – Adapting to the change
It’s one thing to experiment, but if you don’t act on the outcomes of those experiments you won’t progress in your adoption journey. An organisation considered to be at stage 3 will be able to show evidence of successful experiments leading to real change in operational processes. For example, if use of coding agents has been demonstrated to achieve improved software delivery, there should be examples of those same agents embedded in teams working on real business initiatives.
However, there’s a reason this stage is called adaptation. It’s rarely enough to just plug a new technology into existing ways of working. During those experiments, the participants will have identified new approaches that need to be adopted to make best use of the technology. Hence, successful adoption is as much about adapting to the technology as it is adopting it.
This adaptation starts with the findings from the experiments, but it doesn’t end there. Adaptation is an ongoing activity, and teams will continue to learn and improve how they use AI to maximise the benefits it promises to deliver. It is therefore important to tread carefully. Expand AI adoption in a measured way so that findings can be gathered to make subsequent adoption even more effective.
4 – Consolidating your position
Eventually, your adoption of AI in proven areas of success will become routine. Adaption will have taken place and any further changes to ways of working become minimal. Now is the time to consolidate your position. Once you enter stage 4, the goal is to make sure no stone is left unturned. You have now definitively found the contexts in which working in the new way, with AI embedded, is proven to be better than working without it. This means you should be ensuring that it is adopted in every such context that exists in your business.
Consolidation of your AI capability may take a long time, especially if you’ve found many places where it helps, but the adoption process should be repeatable and reliable. You are now well on your way to stage 5.
5 – Maintenance as usual
This is the endgame as far as technology adoption is concerned. You know you’re in the maintenance stage when there are no new places to use the technology that either haven’t already been exploited or where experiments have shown there’s no benefit case for doing so. We reached that point a long time ago with office software suites – some would say we’ve gone beyond that point, but that’s a different question.
The maintenance stage is all about keeping things up to date. All technologies continue to change, even when well established, and so it’s important to keep up on a regular basis with that change. It would be a shame to fall behind after all that hard work to get to this lofty position. Also, there’s a question of security. They say software doesn’t rust, but the security of it certainly does. If nothing else, maintenance is all about keeping up with the very latest patches.
For AI adoption, no-one is anywhere near this point, and it’s unlikely anyone will be for some time to come. It will, however, be just as important as it has been for other technologies, possibly even more so if some of the more wild suggestions about self-improving AI are to be believed.
So what about AI?
To wrap up, my key point is that adopting AI is no different to any technology adoption. Many are talking about this transition as uniquely different to any that have gone before. I disagree. Regardless of the change, all the same rules apply. The five stages hold true for AI adoption, and should be followed. Don’t allow yourself to be driven to excess by FOMO (fear of missing out), and don’t hide from the change because of FUD (fear, uncertainty and doubt).
Invest wisely in finding the places where benefit lies, and ensure you exploit those uses to the full extent that works for your organisation. Avoid copying others without understanding the context – every situation is different, and every organisation starts from a different point in the journey. If you follow this approach and move through the stages steadily, you should avoid most of the major pitfalls.
Then you can enjoy the view as others make all the mistakes you avoid.