The AI in Renessai

Everyone has their own mental models about what AI means. To add to that, the meanings of words change over time. For example, in relation to AI, many have recently started to talk about “agents”. Not too long ago, AI was synonymous with the Large Language Model (LLM). And a few years back, the synonym was machine learning. But do not worry, this article is not here to provide a new, definitive definition of AI.

What is actually important is that you are not doing AI just for the sake of AI – and when you decide to use it, it will be done in a way that is effective. In this article, we will discuss how we approach creating effective AI systems that deliver a tangible return on investment for our clients.

AI must be viewed holistically

AI encompasses a collection of techniques; some of them are old but gold, while some more recent like LLMs. Building effective, production-scale AI solutions requires domain knowledge, a good definition of the problem to be solved, and expertise in both experimenting with and selecting the right mix of AI techniques. Plus, you have to stay up to date with the latest research and development without forgetting what is valuable about the past.

To give you an overview of simplified high-level AI techniques, check out the table below. It outlines their characteristics, use cases, and limitations. The table spans from traditional mathematical approaches to generative AI techniques, showing that each has its own appropriate uses. It is crucial to know when to use these techniques and when not to.

As this table illustrates, a rule-based system defined by human domain experts may be a perfectly adequate solution in a simple static environment. For example, a simple rule might block an IP address if your website receives an unusually high number of requests in a short period, to prevent potential malicious attacks. When the complexity increases and the environment evolves, we need to move to machine learning techniques. For instance, predicting churning customers can be achieved by learning relevant patterns from large amounts of multivariate customer data. When dealing with conversational interfaces, unstructured data, like text documents, and in content creation, generative AI techniques can be particularly effective.

During hype cycles, the market pressure to apply the latest techniques, like LLMs, to every problem can feel overwhelming. At Renessai, we help our clients resist the temptation to chase after the latest flashy algorithm or AI technique. We do this by guiding them on a journey toward greater organisational AI maturity, beginning with aligning all levels of the organisation on a holistic AI approach. In our world, the focus should always be on solving real business problems with the most appropriate solutions. It is important to appreciate the complexity of the problem while striving for a simple solution. The value created is what ultimately matters — not the specific AI technique used.

AI can be used to automate and augment

To find the best solutions, we need to understand and build on the strengths of both AI and humans. While AI has superhuman capabilities for quickly learning patterns, memorising, and interpolating within large historical datasets, humans can adapt to uncertain situations, achieve creative outcomes, and sometimes invent something new outside the historical data box.

When the problem to be solved is operational, involves complex data points, and requires large volumes of outputs around the clock, AI can be used for automation, doing things on behalf of humans. For instance, recommending products in a personalised way for each online shopper can be automated far better than any human could, with a machine learning-based AI solution, resulting in a double-digit revenue increase. On the other hand, when it is time for complex decision-making involving reasoning, creativity or empathy, AI is better applied to augmenting humans. For example, AI can be used to recognize industry trends and customer behaviours, enhancing domain experts and human decision-making in corporate strategy work.

While many aim for end-to-end automation with AI, often a human-centric augmented approach yields superior results. Being able to clearly assess when to automate and when to augment with AI is a prerequisite for delivering outsized returns when adding AI to any system.

AI is not a product

The recent generative AI boom has many investors anxiously waiting for the killer AI product that will justify the billions poured into developing this one type AI technology. However, AI is not a standalone product; it is a feature. For AI to be valuable, it needs to be part of some other product, service, or process. In other words, it is not about the AI but what the AI can enable. It is also important to realise that a valuable product, service, or process might not need AI at all. Thus, there will not be any killer AI products, but there will be killer products, platforms, and ecosystems enabled and enhanced by AI features.

When AI is viewed as a feature, it is also easier to avoid the trap of hastily building big, flashy AI solutions that often result in annoying and disappointing user experiences. Instead, small, subtle, and task-specific AI features, often running constantly in the background, may be what people actually need. These features are also easier to evaluate objectively, especially since evaluation-driven development should be a present in all AI work.

Think about it: simply prompting LLMs does not free you from the need to invest in good design and finding the right product-market fit. And do not be duped into skipping evaluating AI features and their usage contexts from multiple ethical perspectives; for example, by considering both intended and unintended uses, and how different users and those affected by the outputs might be impacted. Thanks to recent AI advancements, we could say a new renaissance in product design and evaluation is here.

AI (and data) does not need a separate strategy

There was a time when IT was viewed as merely a support function, keeping business and IT as separate silos without alignment or a shared strategy. As Bharadwaj et al. (2013) reminded us more than a decade ago, “the time is right to rethink the role of IT strategy, from that of a functional-level strategy — aligned but essentially always subordinate to business strategy — to one that reflects a fusion between IT strategy and business strategy”. So-called “digital strategies” were just this; a recognition that you can no longer separate the business and technical strategy since anything a company does nowadays is executed to some degree through digital infrastructure.

It is not different with AI and for this reason, we work with our clients to embed AI into the overall business / digital strategy. AI development goals need to align with, support, and drive broader business objectives. The success of AI projects should be measured by relevant business impact KPIs, not just technical metrics. Quite often, these metrics are not even that different, since we usually aim to optimise trained AI models for a specific business goal. Beyond business alignment, AI development must also be aligned with IT, since AI requires a whole lot of traditional IT infrastructure and software around it to be truly valuable.

However, in the short term, we acknowledge that kick-starting an AI transformation phase might require a temporary AI strategy. But the long-term goal should be to fully integrate AI into the overall business strategy, so that eventually there will not even be a need to explicitly talk about “AI” anymore. AI will become as omnipresent as the internet has.

Conclusion

The perspectives on AI we have discussed have been shaped over decades of practical work by our senior consultants, working with clients across various industries and continents through multiple technology and AI shifts. Some of these viewpoints may seem obvious, but they can easily be forgotten amid all the exciting AI news that comes out daily.

At Renessai, we take a strategic, no-nonsense holistic approach to AI, which has led to numerous AI solutions in production delivering measurable business value. While there are quick wins to be had, you also have to dig through the salt mines before you can reach the gold mines. AI hype cycles come and go, but there are always outlier businesses in the market that thrive regardless of the current trends. It all starts from the strategy.

1: Bharadwaj AS, Sawy OAE, Pavlou PA, et al. (2013) Digital Business Strategy: Toward a Next Generation of Insights. Management Information Systems Quarterly 37(2): 471–482.