AI has been hyped up as one of the most game-changing technologies seen in years, and as a result, there's huge pressure on businesses of all sizes to get involved.
Indeed, global spending on artificial intelligence solutions is set to reach more than $50 billion in 2020, and then more than double again over the next four years, reaching $110 billion by 2024.
However, as is often the case with hyped technologies, many firms may find there's a big gap between their expectations and what they achieve in reality. Most firms report some failures in their AI deployments, with one in four experiencing failure rates of up to 50%.
The challenges with going from POC to production
Often, the problems come down to difficulties in turning small-scale proof-of-concept (POC) initiatives into wider solutions that can be rolled out across the business. According to Accenture, more than three-quarters of C-suite executives (76%) have found it difficult to scale AI across the business.
However, 75% agree they risk going out of business within five years unless they can scale up their AI initiatives effectively. Therefore, it's clear that the stakes are high and businesses can’t afford to do nothing to improve their use of AI.
But why are so many firms finding it difficult to translate POC efforts into full production? There are often a number of factors contributing to this, from a lack of clear direction to a poor company culture.
Businesses may often struggle with the technical challenges involved in scaling AI. Going from pilot schemes with relatively small data sets to company-wide efforts can often be difficult for firms without expertise or experience in handling the volumes of data this demands.
5 steps to make large-scale AI a success
So how can firms avoid these pitfalls and ensure their AI scaling efforts stand the best chance of success? Here are a few best practices that every firm should bear in mind.
1. Be clear about your intentions
A common problem is that firms are unsure about exactly what they want to achieve with AI - instead simply feeling they need to get on board quickly, whether or not they have a clear plan for how it’ll help meet their wider goals. To avoid falling victim to this, it's vital firms understand how AI will align with their wider business strategy.
Accenture calls this "intentional AI", and it requires several steps. This includes identifying the base problems firms want AI to solve, educating themselves on the capabilities and limitations of AI technology, and being realistic about the timescales involved. AI projects can take years to bear fruit, so patience is a virtue.
2. Get organized
Another key step is having an effective structure for the scaling up of AI, both in terms of who within the organization takes ownership of a project and how it’ll actually be executed.
McKinsey notes that successful businesses tend to divide the workload between a central analytics 'hub', usually led by a chief data officer or chief analytics officer, and 'spokes', typically within business units or specific geographies.
Hubs will be in charge of issues such as data management and government, managing systems and hiring talent, while spokes will be responsible for areas such as end-user training, workflow redesign and impact tracking.
3. Sort the signal from the noise
No AI project can work effectively without access to data, but ensuring you're feeding the right information into the system is often harder than firms may think, especially once they scale out to company-wide levels.
Not all data is created equal, so it's vital you're able to pick out the useful sources - the signal - from the surrounding noise. Start by identifying just a few key data points that’ll be most relevant to your business strategy and filter out anything else.
4. Invest in talent
One of the biggest challenges for any firm is the shortage of talent available. Indeed, more than half of firms (51%) say they don't have the right skills and experience in-house to make their AI strategies a reality.
With so much competition for talented workers, it’ll usually be expensive to bring in these resources. Therefore, it's vital you have a clear plan for training and upskilling your internal IT teams.
According to McKinsey, only 35% of top-performing companies and 10% of others say they have an active continuous learning program on AI for employees, so this is often a clear area for improvement.
5. Drive adoption
Finally, no AI project can be a success unless end-users within the relevant business units understand how to use it effectively. Yet firms perform poorly in this regard, with just 36% of high-performing companies and 8% of others saying their frontline employees use AI insights in real time to enable daily decision making.
Focus your efforts on comprehensive training programs to educate employees about how these tools work and what their benefits are, and consider offering incentives for those who adopt the tools successfully.