Applying strategic thinking to applications of AI in the AI gold rush

CTOs are grappling with the AI gold rush and making choices about participating. Here are some things to consider.

I see a lot of activity with AI — many companies are experimenting with it. Some companies are releasing new products with AI at their core; some of those address something potentially valuable, but most don’t. Even more are using it to augment existing products, again some successfully, but many appear to be just bolting it onto their products. It feels like a gold rush. Some will strike it rich, and many will strike out.

The difference for many will be either one of luck or strategy.

• Luck is a factor - you must be in it to win it, as they say. And if you are executing quickly and trying many things, you will see things work and many things not, and you might learn enough to navigate to success. Of course, it's really that you are one of thousands doing the same thing, so some of you are bound to stumble on something valuable.

• The other factor is strategy. Some will put equal or more energy into determining where to dig. Or work out a mechanism to dig more places - think VCs who back hundreds of startups exploring a space like AI.

Doing strategy appears difficult for several reasons. A lot of what is described as strategy is not strategy but planning. This might seem like splitting hairs, but you don't have a strategy if you skip over forming a strategy and straight into planning actions. You skipped over making choices that shape a theory of how you will succeed and enable you to focus and test your theory. You are back to just executing — another prospector digging in a random spot. So to start being strategic means being conscious of the choices any investment of time and resources represents. That’s easy in concept and hard in reality when so many things conspire to take time away and pressure you into action.

When I see the activity in AI applications today, I see the same mix - those energetically prospecting as part of the gold rush and those who act with intention.

Let’s focus on examples of companies adding AI to their existing products, which is the largest by volume; we can look at pure AI product offerings separately. Let’s start with the positive. In my daily working life, I am experiencing some examples of simply brilliant uses. One highlight is the ‘Magic Switch’ feature in Canva.

Let’s consider some of Canva’s strategic positioning. Canva already excels at taking a lot of manual and non-value-adding thought and labour out of preparing graphics for use in social media. It’s a lower friction option than Photoshop - you can use it without downloading or even purchasing it, as it offers a freemium option. It can be used without much experience in design and thus with a drastically lower learning curve.

For instance, searching for the platform and context you use for the graphic will usually identify the format and canvas size you need. Recently, I wanted to update my Facebook Banner image. It has templates that size or can start you with a black canvas of the correct dimensions. So far, none of this is using AI overtly but demonstrates part of Canva’s value proposition - it has provided more people access to Photoshop’s capabilities with a fraction of the required learning curve.

In my case, I had an existing image that looked the way I wanted but was configured for another social platform. Using the Magic Switch feature, I could quickly adjust it to the Facebook Banner format, and the AI would retain the essential aspects of the composition as it adjusted the dimensions. Maybe less wow factor initially as compared to a lot of applications of generative AI darlings that have captured so much attention of late, but I’d wager this is a function that will get used frequently and strengthens the value proposition of Canva and continues to help it to stay ahead of alternatives.

As I write this post, Grammarly is seamlessly suggesting improvements to my writing using a mix of its core intelligence (which is no doubt a mix of some traditional language analysis algorithms and in-house developed AI and enhancing these capabilities further with integration with the generic generative AI platform capabilities. So, these are examples of companies finding meaningful ways to improve the value of what they already deliver. Let’s move to the bad.

I think it’s safe to say the majority of examples of integration of generative AI capabilities into existing products released over the past months are detrimental additions. Adding gimmicky features makes these products more complicated because they are not sufficiently helpful and add to the clutter of the user experience. These features will not have enough sustained usage to warrant the cognitive load they add to product users, and over time, many will be removed, or the products will fail.

AI has a wide range of capabilities. In recent times, the most accessible examples that have captured the imagination are the text generation and image generation capabilities. They are very easy to generate prompts for and deliver results, providing a surprising leap of quality over what might have been expected. And so it’s these capabilities that have been the hammer for every nail. It’s almost like companies have looked at their products for any output that might contribute to formulating a prompt and then bundled that as a feature in the hope that some of these might solve a problem their users have. In most cases, it’s another button in an already confusing user interface.

Why is this happening? Well, two explanations working in concert explain most instances:

• The barrier to entry was lowered.

• The feature factory is alive and kicking.

The barrier to entry was lowered

One of the reasons for this pattern is the barrier to participating in the AI gold rush was lowered. OpenAI, the FAANG companies, and a range of other startups released AI capabilities available for easy integration via APIs, for which there are point-and-click toolchains, such as Zapier and the like, available for easy usage.

Barriers to entry are always, in a strategic sense, a double-edged sword. When they are high, there is a strategic ‘moat’ - the ones with the technology advantage can compete in ways others can’t easily replicate. When they lower all can participate with less friction using technology that was previously beyond reach and execute their ideas more easily. The trouble is, so can everyone else.

For this reason, it becomes even more vital to consider your organisation's strategy and how different opportunities might complement it. Or else you are hoping you will stumble on a helpful combination.

The feature factory is alive and kicking

7 years ago

John Cutler

popularised the term ‘Feature Factory’ in this post: https://medium.com/@johnpcutler/12-signs-youre-working-in-a-feature-factory-44a5b938d6a2

The term describes the idea that product development had been reduced to delivering features in the hope that the product would continue to be more valuable with each new feature.

The AI gold rush has potentially reawakened the feature factory instincts of even reformed feature factory companies. As I covered in a previous post, it is not completely wrong to try to move quickly and see what sticks as a strategy; this can sometimes work.

Odd contradiction: Focusing only on delivery optimisation is sometimes enough

Even better is when you do this intentionally. For instance, maybe initially, you want to increase the number of people thinking and trying things out in a space, or you have confidence in your organisation’s systematic ability to determine which experiments are working and which are not. A lot of the time, though, it amounts to wasted effort and time competitors may use more effectively.

I continue this thinking in the following post:

Atlassian’s release of AI features might lack strategy.

How is your organisation approaching the AI gold rush? What steps has the organisation taken thus far? Do they resemble frantic prospecting or more deliberate exploration in areas that complement the wider organisational strategy? Share your experiences in the comments.