What you will know after reading this article:
Value and impact created by AI vs Software.
Different risk profiles when using AI vs Software, and communication to investors and customers.
Behaviour of AI vs Software, and changes to expectations and investments.
Our ecosystem is changing fast, and your ability as a leader to clearly distinguish between the powers of emerging technologies is important for the success of your business. Poor investments into new ventures can threaten your competitiveness and waste valuable resources. If you invest into an AI venture, but you treat it as a software venture, then you are doing it wrong.
Despite the huge business potential of AI technologies, many AI ventures are poorly executed and miss significant business opportunities. There are many reasons for this poor execution, e.g., ill-prepared culture and strategy, insufficient access to talent, and poor data and infrastructure preparedness. However, in this article I focus on what I have experienced being the biggest AI misconception when advising CEOs, CxOs, boards and government ministers: AI technology is viewed and treated as if is Software technology.
Value and impact created by AI vs Software
If you don’t know the difference of the power of AI vs Software technology, you will likely make decisions with bad consequences for your business. Distinguishing between the power of these technologies is important for several reasons. For one, as a leader, you need to know what value you obtain when you pursue a Software vs AI venture - in other words, what magnitude and type of business problem can you expect to solve with AI vs Software? Both are very different. Generally speaking, an AI technology can perform a task that today could only be performed by a human, e.g. a switch board that understands the intended request by a caller and connects the caller to the right department. In addition, AI technology can often perform these tasks quicker and more accurate than humans and therefore have a bigger impact on your business. This translates directly into major efficiency gains and higher customer satisfaction, and frees your workforce from performing repetitive and tedious tasks to more engaging tasks that add more value. Taking the switch board example, traditional software helps a switch board employee to look up a person in a directory, but not to perform this process automatically. Furthermore, software will not reliably recognise the intention of the caller. For full business impact, AI technology requires indeed a very different way of technical and business development that differs from traditional software practice (as this discussion would stretch beyond this article, I will address this in a future article).
The value that you obtain from AI technology is not only much bigger than that of software but also applicable in areas which traditional software can not reach. Many companies that are now valued at 1+ trillion USD all use AI and are now market leaders, such as Amazon, Apple, Google, Baidu, Alibaba, and Tesla. Market leadership with the help of AI technologies is about to happen in many industries, such as health, logistics, manufacturing, energy, and retail. While traditional value created from software is robust and useful, it does not have the same scale and reach, and therefore has less market impact as AI. AI powered tools will be used to create value in all parts of the company (and certainly not only in the IT department). This is also a reason why the market size of business using AI technology is many-fold larger than the market size for software technology.
Different risk profiles when using AI vs Software, and communication to investors and customers
A second area of consideration when considering the difference between AI vs Software is the risk to your company's productivity and brand. As the saying goes, with great power comes great responsibility. The behaviour of AI technologies are not predictable. The behaviour of most AI systems today is likely more predictable than that of a human. Overall, the benefits of most today's AI technologies are considered to outweigh its risks. This has been recognised by every major institution in the world, from the European commission, OECD, UN, China KP, and US Government. Indeed many major companies using AI clearly stated in their annual report the risks so that investors and customers know what they are in for. Here is an example from the annual report from Google / Alphabet.
"[[N]] ew products and services, including those containing or containing artificial products Intelligence and Machine Learning can create or reinforce new ethical, technological, legal and other challenges that may negatively impact our brands and the demand for our products and services, and adversely affect our sales and operating results."
Behaviour of AI vs Software, and changes to expectations and investments
A third distinguishing factor is that an AI system is not deterministic or "error free". A software system is expected to do exactly 100% what it is designed to do without change in behaviour. When you press "save" on your word processing tool on your laptop, you expect that document to be saved 100% of the time.
This is not the logic of an AI system. An AI system behaves more like a human, here are two considerations. First, an AI system makes predictable "mistakes". An example is an AI system that classifies scanned documents into "driving license" and "car registration" for a government authority. On average, the AI system classifies 90% of the documents correctly into each category and 10% of the time it does these classifications incorrectly. So, if you had a software and you press "classify document", you would expect it to be a 100% correct, but for this type of task these outcomes cannot be guaranteed. This logic is similar to self driving cars regarding accidents or safety. My rule of thumb is that if an AI system is better than a human in terms of making these "mistakes", then it might be worth considering applying the AI system (possibly in combination with a human).
Second, AI systems learn over time - every new observation will make the system behave differently as the AI system learns. And every time the system learns, it could potentially make a different decision. In the classification example above, it might mean that the AI system over time can learn certain new patterns in scanned copies of "driving licenses" and therefore is better able to classify "driving licenses" with higher accuracy.
Knowing that AI systems are non deterministic and make predictable mistakes will influence your expectations in terms of what an AI-powered product development or AI venture may look like in your organisation. The structure, flow and application is different to Software technology, and you can also set expectation in your organisation.
Making all the above actionable for a leader
Let us consider a scenario where a company leadership makes a significant business investment and takes the above into account. Leadership decides to invest 10% of the annual development budget into a business challenge that can be solved by an AI technology. Using AI technology requires different approaches, skills and infrastructure than using software. Hence, the company seeks AI specific competencies both for business and technology, develop an AI strategy and an initial roadmap (e.g. buy or build AI or find a good partner with a strong AI profile). The brand and reputation of the company is also protected as a risk assessment was conducted by a competent AI business team. As the AI strategy is aligned with the company strategy, and AI risks have been communicated effectively to the board and constituents of the company. The pitfalls of treating an AI venture as a Software venture are avoided, and the company has a good start into becoming an AI first company ready for entering the era of AI and Data Driven business.
In light of the above, Artificial Intelligence and Software technologies can be a great complement. From the outset, a leader needs to clearly understand the different requirements in terms of competence and infrastructure, and the expectation on the business, strategy, and culture. In short,
View and treat AI differently to traditional software – you are more likely to adopt AI technology quicker and more successfully.
In this article, I laid out a few of my insights on the value and business impact, risks, and behaviours of AI vs Software technologies. It is worth noting that other technology areas are often confused with AI technologies. If we extend the title of this article, then it would look something like this:
AI ≠ Analytics ≠ Business Intelligence ≠ Robotics Process Automation ≠ Data Science
If you like me to elaborate on any of these difference, please let me know in the comment section - more articles will follow. I am looking forward to your thoughts and comments.
Written by Christian Guttmann.