As software once ate the world, it appears that AI is increasingly consuming traditional software. As technology to run our lives and businesses has become the norm, more organisations are exploring how AI creates new opportunity for innovation, can improve performance and drive growth and scalability. Unfortunately, the growing prevalence of AI projects also brings with it an abundance of failed ventures. VentureBeat estimates that 87% of AI projects fail to reach production. Project failure is commonly caused by a lack of clearly defined business objectives, vague strategy, a lack of understanding of how AI will work within the business or inability to scale prototypes. Before leaping into AI adoption, you have to understand and identify if AI is right for your business. In this article, we will discuss some of the key questions and considerations business leaders must address when starting their AI journey.
What is your business objective?
In our opinion, inarguably the most important question for any business looking to take their first steps in AI. Why?
AI models are built for a specific purpose to solve a specific problem. It is not a magic wand that will solve all your business problems. Therefore, when considering if AI is right for your business, the first action is to define a clear and specific objective to be addressed. This involves pinpointing business challenges, pain-points, bottlenecks or even opportunities that clearly establish the problem that needs to be solved.
The keyword here is specific. Vague, general descriptions such as “grow revenue” or “improve customer satisfaction” - whilst valued business outcomes – do not translate to a problem that can be solved using artificial intelligence. If you struggle to be specific from the outset, an effective technique is to start vague and repeatedly ask ‘how?’. For example, by asking ‘how?’ with regards to “grow revenue”, you will find yourself exploring different aspects of this business objective that may lead to distinct challenges, such as reduce customer churn, grow sales, attract new customers or grow share of wallet – each of which will require its own unique approach when it comes to AI. Each of these objectives can be iterated upon to become more specific and quantifiable – e.g., grow share of wallet using relevant product recommendations based on historical purchase behaviour and the behaviour of customers within the same segment - allowing for the business priorities to emerge.
Another effective technique is to reframe business challenges as prediction problems. At its core, artificial intelligence involves taking data as an input, performing analysis and generating an output or prediction that is valuable because it informs decision making, either across an automated pipeline or for a human. Therefore, determining the types of insights or predictions could help you make better decisions is another approach to identifying opportunities for AI in your business.
Reflecting on your business challenges, needs and distilling them into specific objectives is the starting point of any innovation journey and is fundamental to evaluating whether AI is right for your business. The trick is to not become distracted by the technology, but rather focus on what you’re the expert on: your business.
Is AI the best solution to the problem?
With all the hype surrounding AI, you could be forgiven for forming the opinion that AI is the panacea to all business problems. In fact, AI may not always be the most relevant or appropriate solution.
An artificial intelligence system is one that learns and evolves over time through exposure to data, to perform a task that would traditionally require human-level intelligence. Creating such a system can be a long and complex process of data acquisition and curation, model training and tweaking, then automating these processes to enable scaling: an endeavour that requires time, money, skilled data scientists and engineers.
The fact is that not all problems need this level of data transformation or intelligence applied to solve them. Many business problems can be addressed perfectly well through more traditional techniques such as software or RPA. In some instances, AI may be a viable solution for part of a problem but will also need to work in conjunction with other technologies to achieve the desired outcomes. This further highlights the importance of defining objectives as opposed to being preoccupied with technology that solves it.
What data do you have that is relevant to solving your problem?
When first getting started with AI, it is critical to understand the importance of data.
Analysing data to generate predictions is a core principle of artificial intelligence.
Therefore, what can be achieved using AI will heavily depend on access to high volumes of high-quality data, that is relevant to the predictions or insights you are looking to extract . If you have poor quality data, you are only going to generate poor quality insights: garbage in, garbage out.
What is the business case?
Whilst the plan, build and implementation of AI solutions poses a series of technical challenges that require technical minds, the decision to adopt AI is fundamentally a business one. Depending on your organisation, this decision may require a business case to understand organisational impact and validate investment. The typical approach to developing a business case consists of determining the cost and impact of a problem, assessing the performance, risk, investment and ROI of potential solutions, then evaluating what if any action should be taken. Although this process is fundamentally the same for AI projects, there are some considerations that can make a business case more complicated.
The main consideration for many business leaders is often financially focused: understanding and quantifying how AI can reduce costs by optimising existing processes or generate new revenue by creating new value-add products and services. However, there are considerations that should influence the business case for AI.
Firstly, don’t discount non-financial benefits or treat them as a means to a financial end.
AI has the potential to create a host of benefits such as improved productivity and efficiency, reduced waste, differentiation, competitive advantage, improved customer satisfaction, new products and services that solve unaddressed problems, scalability and long-term growth. These are legitimate benefits that deserve earnest recognition for the value they create that are correlated with financial success, but not necessarily causal.
Secondly, consider adopting a return on portfolio (ROP) approach to AI projects rather than a traditional ROI approach. ROP considers the collective function and value created by a set of AI projects, some of which may be instrumental in testing and validating ideas or creating new opportunities in the future. By adopting an ROP approach you take a holistic view that allows you to validate investment in worthy project ideas that might have been difficult justifying if evaluated in isolation. The collective value is worth more than the sum of its individual parts.
Finally, consider the potential cost and risk of inaction. It’s tempting to become engrossed in dotting all the i’s, crossing all the t’s, trying to understand the art of the possible or falling into the trap of thinking that prolonged speculation is a worthy substitute for action.
Which makes sense: AI is vast, complex, uncertain and constantly changing: all elements that lend themselves to analysis paralysis. But sometimes a bad decision made quickly is better than a good decision made too late. Many brands and businesses have relegated themselves to the graveyard because they were unwilling or unable to take action when necessary, whilst their competitors took prompt action.
What is your openness to innovation and risk?
Most AI projects involve some degree of innovation and investment. Innovation is often uncertain, which creates risk. This risk should be manageable and justifiable: if the objectives are well-defined, solution approaches have been evaluated correctly and considered, the data is fit-for-purpose and the value potential is worthwhile.
Risk is a highly subjective concept and can mean different things to different people.
For some, poor performance, project failure and loss of investment will be the greatest risk. For others, investing vast amounts of money in third party services that result in no intellectual property (IP) creation and having certain services (or a business model) reliant on the continued provision of these services or survival of vendors, might be the most significant risk.
Risk is a fundamental part of the AI journey, one that can be mitigated but not avoided entirely. What is important is to accept risk as part of your journey, have an informed understanding of the risk associated with different approaches to adopting AI and selecting the approach and partners that are right for you.
Want us to help you answer some of the above questions and start your own AI project? Contact us.
Written by Clayton Black and Joe Duszynski Lewis