Investing in Data and AI — When and Why
In a previous 3-part series, we discussed the main hurdles limiting the value that businesses get from data. In particular, we highlighted the importance of the timing of investment into establishing data infrastructure, recruiting data professionals, etc. The reasons are clear. Any significant returns on investment that you make on data are likely to come from their use in conjunction with AI techniques such as the various ML algorithms in your products. This is when your Data Science/ML teams come in, to explore the data, build the ML models, and package them up into what I call AI services that your products can leverage. Before that can happen, you should ideally have started collecting the right data with your product that already has product-market fit.
In this short article, we zoom in on the timing aspect of investment in big data and AI, and as part of that, the benefits of powering your products with them. The points that follow are especially relevant to businesses that offer digital products to facilitate transactions in two or multi-sided marketplaces. From a scalability perspective, AI services can offer such businesses a potentially more sustainable way of maintaining and improving products to engage and retain users in the long run. This is due to the nature of AI-powered products which improve automatically and continuously as more data of the right kind flow in. On the defensibility front, while it is not inherent to data itself, the use of thoughtfully acquired data in the AI services that power your product can create hard-to-replicate user experiences. The lifecycle extension of digital products is also another possibility with AI services.
By “scalability” we refer to a situation where a business is able to multiply its revenue with the minimal incremental cost
On the timing of investment, we use the start-up J curve as the backdrop. The J curve outlines the six phases that entrepreneurs go through as they try to build out sustainable business models. The curve starts with the create and the release phases, where teams and money come together to turn ideas into products and release them into the market. During the morph phase, the products and business models are iterated upon based on customer feedback, metrics and so on. If product-market fit can be achieved within this phase, the companies move on to optimise their business models and grow while keeping cost down and revenue high during the model and the scale phases. These phases are shown in blue in the diagram below.
The best time for entrepreneurs and companies to double down on their investment in platform, data and AI is the period towards the end of morph and during model phases. In the world of “doing things that don’t scale”, this is often the best time to “repay” some of the debts that have been incurred in the name of speed so that your infrastructure, for instance, does not crumble as you stack more on top of it, e.g., more features, more users. While many seasoned entrepreneurs get this point, what some people often miss is the fact that this period is also perfect for putting money into (1) data infrastructure to capture, process and store big data and make them accessible, and (2) data science professionals to explore the growing volume of data and build out AI services that products can leverage. This period which is ripe for investment into all things data is marked as (A) in the diagram above.
It is important to call out that there are AI services that can be developed or utilised in products that will immediately be beneficial during the release and morph phases. These vanilla AI services or tools require next to nothing in terms of data about your users and how they behave in your marketplace. Often, the purpose of these off-the-shelf solutions is to save you the cost/effort associated with human intervention in making your product works smarter and reducing bad frictions. For instance, let us assume your product involves users uploading content. Instead of requiring the users to provide information about the content such as languages or tags, they can be predicted and offered to the users as suggestion. Not surprisingly, such capabilities are unlikely to offer you any competitive advantage in the long run. This is when you need investment during period (A). As an example, once you have started capturing what your users search for, the content they click on and the other actions in your marketplace, you can utilise AI to understand their preferences or intent. The derived user information can in turn be used by other AI services in your product to better serve those users and potentially others.
Some things simply don’t automate. For everything else, build systems; they’re a lot easier to scale quickly than humans.
By now you must be wondering, when will your investment during period (A) starts bearing fruits? The short answer is after product-market fit has been found and it can happen throughout the scaling period, which is indicated as (B) in the diagram. How soon you can reap the rewards can depend on the nature of the problems you are solving with big data and AI and the solutions that your teams have picked. As more and more parts of your product become powered by AI, especially the ones that are built on proprietary data, your business will be able to scale on top of that. This period of scaling with big data and AI is marked as (C) in the diagram above. This period essentially offers your business an alternative path to the harvest phase in the J curve. Do not get me wrong. Your business will still grow and you will probably do fine without big data and AI. However, the often neglected cost associated with retaining existing users and engaging new markets without products that can improve automatically and continuously with more data will become a major drag on your ROI.
Before we end, let us look at the ROI in big data and AI services through two other difference angles. The first angle is through the lens of the typical product lifecycle that involves products going through the introduction, growth, maturity and decline phases. AI services give you the opportunities to extend the lifecycle of your product by allowing you to improve existing features or introduce new ones in ways that otherwise would not be possible, as shown in the diagram above in red.
The second angle, as depicted in the diagram above, looks at the role of big data and AI services as a “dampening factor” to the cost associated with capturing new users and servicing existing users who tend to expect more from your products as time passes. This is especially true if your business operates in a two or multi-sided marketplace and has to deal with constantly growing content pool and match-making the different sides. Your ability to grow and scale efficiently is predicated on big data and AI. Due to the nature of how AI services work, by feeding off increasingly growing pool of data about your users and how they use your product, you will get to reduce the reliance on heuristic-based solutions and solutions that require excessive humans intervention in order to solve your users’ problems.