Do you have a good reason to implement AI?

Staff Writer
By Staff Writer, ITWeb
Johannesburg, 02 Sept 2019
Johan Steyn, portfolio lead at IQBusiness.
Johan Steyn, portfolio lead at IQBusiness.

With all the hype around the fourth industrial revolution, many organisations are scrambling to implement artificial intelligence (AI) and machine learning (ML) within their operations and offerings. However, many are failing. 

Where should businesses start, and what are the steps they should take to ensure a successful implementation?

According to Johan Steyn, chair: Special Interest Group – AI and Robotics at IITPSA, and portfolio lead at IQBusiness, a good place to start with is any area that directly impacts the way you service your customers. 

Most businesses have have lots of technical debt and legacy systems, with a workforce that might not be ready or even suited for the introduction of new disruptive technologies, he explains.

And while many may wish to have AI and ML underscoring their whole business from day one, it is crucial to ensure that the right foundations are laid before introducing these technologies.

Steyn says there are several important steps to take before embarking on an AI journey. 

1. Understand the problem you need to solve

First and foremost, there needs to be alignment with the business strategy, and an understanding of the problems that need to be addressed. AI or ML might not be the answer.

2. Have a good reason to implement AI

Next, the business needs to understand its AI ambitions. Sometimes the reason for embarking on this journey is the classical fear of missing out and the mindset that if others are doing it, so we better do it, too.

”Other times, a divisional executive will want to tell the board that he or she is implementing AI. I call this the 'AI tick-box' exercise. These initiatives cost money, divert unnecessary time and are often doomed to fail.”

He says there are three good reasons for introducing any new technological capability to your business - decreasing your business cost base, lowering your business risk exposure and improving customer experience.

3. Assess your AI maturity

“Map out your organisations’ main process areas," advises Steyn. "These could include customer service, finance, operations, human capital management, or service management. Then, as part of the matrix, map the maturity per process area in order, from manual processing, isolated automation with individual tools, tactical automation utilising a variety of tools, and end-to-end strategic automation.”

Steyn says this matrix should give the business a high-level view of areas most ready for an AI initiative. 

4. Keep change management in mind

There are also several pitfalls to keep in mind, he says. Firstly, change management. Employees may feel insecure about their future when the business starts talking about the introduction of AI and robotics. “You need to take them by the hand on a journey of discovery. Rather speak about co-botics: the fact that this technology should enhance our jobs, rather than replace us.”

5. Know your regulatory requirements and labour relations

“If you work in a highly regulated industry, such as banking and financial services, you may be constrained to all the potential benefits that AI may bring to your business operation," notes Steyn.

"You may also have an unionised workforce and will have to plan for the strategy and messaging to your staff and unions.”

6.  Plan for workforce upskilling

“Intelligent augmentation is key to the journey. A well-formulated plan regarding the impact of AI on the current way work is done, on how AI will change the way work will happen in the future, and the skills needed. New roles need to be introduced if the business does not have them already, such as data scientists and AI engineers. Businesses who are working in a market where future skills are limited, need to consider a hybrid model of upskilling staff while utilising the expertise of a third-party vendor.”

7. Have a solid data strategy

The final piece is data. “It starts and ends with data, as this is the lifeblood that AI and ML live on. Are you harvesting enough and suitable data from your clients (if you have their permission and adhere to regulation), and from your internal business operations? Behind every AI strategy is a data strategy.”

Once these foundations are in place, the business needs to consider whether it should build from scratch, or buy a solution from a vendor, says Steyn.

 “The next step is aiming for the initial proof of concept and minimal viable product. In the spirit of the agile process, you need to start small, fail fast and learn quickly. You need to build momentum to ensure your mandate is maintained, your current and future funding are secure and that the organisation sees value early on.”