Five Steps to AI Success

Five steps to AI success

Tech giants such as Microsoft, Amazon and Google have artificial intelligence (AI) at their very core and harness its potential routinely. Non-tech companies are now beginning to understand the opportunities that AI presents to them chiefly from a productivity perspective. Few are achieving anywhere near the ultimate potential of AI to significantly increase consumer demand through personalised and enhanced products and services. This article outlines the key steps to realising the potential and overcoming common barriers based on our experience of helping clients successfully navigate the AI field.

Today’s AI focus: Stimulating consumer demand

90 per cent of our everyday experience with AI is derived from tech companies. We constantly receive personalised offers based on our past buying experience and other data. Business and tech media focus on tech companies that are increasingly entering traditional markets, such as finance and health care, and are competing with entrenched competitors through their use of AI. For example, with the launch of Apple Card, Apple will compete against established credit card providers. Amazon competes with traditional retailers, media and logistics companies. Their superior data, understanding of consumer behaviour and offers designed with precision and foresight make it increasingly harder for traditional competitors to maintain their position in the market.

Non-tech companies are using AI with a focus on achieving productivity enhancements. This includes automation of routine tasks, augmenting employees’ capabilities and freeing them up to focus on more stimulating and higher value-adding work. For example, insurance companies aim to automate their claims processes and banks are working towards the automation of standard activities such as account opening and mortgage applications.

The real prize from AI derives from stimulating consumer demand with precise, personalised offers. According to the PwC report, “Sizing the prize: What’s the real value of AI for your business and how can you capitalise?”, the uplift from product enhancements and subsequent shifts in consumer demand, behaviour and consumption emanating from AI will outstrip the productivity gains. Of the total estimated $15.7 trillion of additional GDP worldwide in 2030 generated through the use of AI, increased consumption will account for 60%, or $9.4 trillion. Productivity gains are estimated to account for $ 6.3 trillion over the same period.

Lack of trust in AI and regulatory ambiguity are key reasons for companies not realising the potential of AI

Despite the benefits, the PwC 2019 Global Risk, Internal Audit and Compliance Survey shows, that over half (52 per cent) of respondents “do not plan to use in the next two years” or are “unsure” if they will use AI. This is principally driven by a lack of trust in AI, as well as unrealistic expectations and lack of regulatory clarity:

• AI is not 100% perfect. People trust algorithms until they see them in action and make mistakes. Instances include Tesla crashes or flash stock-price drops driven by automated trading algorithms. AI does not have to be perfect, just better than humans acting on their own, to provide significant, positive bottom line benefits. For this reason, we maintain that human intelligence plus artificial intelligence is the optimal combination.
• AI is not capable of performing magic on antiquated, messy data. Tech companies have quality data. Most other companies do not. As a result of various acquisitions and legacy systems, their data is often less than perfect. There is a misconception, particularly with machine learning (ML), that it will solve all data problems, despite years of neglect. ML needs good data, a high-quality investment and may take 12-18 months before delivering a positive return. Companies should not be disappointed that their data is not at the same level as major tech companies. That is the norm.
• Regulations around AI are unclear: Regulation is currently unclear with regard to AI-related risks, for example in relation to financial transaction monitoring and the use of AI in clinical trials. In  most countries, including Switzerland, there is no clear responsibility in terms of the development and enforcement of AI-related regulations.
• AI is sometimes perceived as a black box: Many AI algorithms are beyond the comprehension of most people, and some AI vendors refuse to reveal how their programs work in order to protect their intellectual property. In both cases, when AI takes a decision, its end users will not know how this decision has come about. 

As a result of such issues, AI projects are often relegated to the proof-of-concept (PoC) stage, siloed in IT departments and lack the investment necessary to maximise returns. 

Closing the gap: 5 steps to AI success

We recommend five key steps in order to move away from focussing on AI failures to realising its potential for protecting and growing your business.

1. Make AI a leadership responsibility
   To move beyond the raft of PoC AI projects and gain the scale necessary for optimising the business benefits of AI requires a top-down approach. Only data-driven companies are able to realise the full benefits of AI. To be a data-driven company requires more than technology. It requires visionary thinking, creating a culture focussed on attracting and developing talent, by building an exciting, consumer driven brand. It also requires the investment to build an optimal data infrastructure. And it needs leaders with the ability to understand and drive AI projects.
2. Determine Your AI investment path
   Work out what AI means for your business from a scan of your industry’s technological developments and competitive pressures, the operational and business growth opportunities that AI can address. Prioritise your potential investments against key criteria including your business goals, your appetite and readiness for change. You will also need to determine if your strategic objective for AI is to transform your business or to disrupt your sector and if you are an early adopter, fast follower or follower in regard to AI implementation.
3. Nurture your AI talent and culture
   While the cost of AI technology is likely to be reduced significantly, the supply of data and how it is used are set to become the primary asset. For this reason, as well as the acute shortage of AI experts, you will need to invest in the talent and culture within your organisation. Training and continuous learning are required to close the skills gaps that hold back AI progress.
4. Recognise and address the fears and risks relating to AI
   Across most organisations the biggest fear is that people will lose their jobs. The adoption of AI will mean that some posts will inevitably become redundant, but others will be created by the shifts in productivity and consumer demand emanating from AI, and through the value chain of AI itself. Communicating how AI can improve the status quo, such as increasing capacity or accomplishing tasks that weren’t possible before, will help generate support. The more AI is fused into your organisation, the more governance and risk management is required to ensure the minimisation of risks including avoiding algorithmic bias, misuse of data and exposing private data.
5. Get inside the black box
   If people don’t know how AI comes up with its outcomes and insights, they seldom trust the technology. Work with your AI experts to get inside the black box. For example, we recently worked with a major doctors network to understand the algorithms around a decision support system that assists them with therapy selection for their patients. In order to improve the adoption of the new technology, and the trust of key members of the firm, we reworked the inputs and user interface. For example, in the minority of cases where the doctor's recommendation differed from the AI-suggested therapy, an onscreen explanation was provided. Such features improved the doctors’ comfort and adoption levels with the program, because it reinforced that the system only recommends the therapy and that they themselves select the appropriate therapy for each patient.