These are the obstacles stopping you from industrialising AI

| minute read

How to escape the curse of the never-deployed Proof of Concept

If you think the hype around AI is too big to be believed, you’re not entirely wrong.  

Yes, AI has the potential to transform many industries – but how often is this potential realised? Some estimates put the failure rate of AI projects somewhere between 83% and 92%.  

That means that as few as 1 in 10 AI projects ever get off the ground.  

It is possible to avoid the curse of the never-deployed Proof of Concept and successfully industrialise AI in your workforce. It could even unlock new areas of potential for your business.  

The key is using your human critical thinking and leadership skills to overcome four key obstacles.  

1. Knowing how much to invest – and for what value  

In every sector there are dozens, if not hundreds of potential use cases for AI. And it isn’t always easy to know which ones will give you a strong return on investment.  

To support your decision-making, we’ve identified four archetypal categories of AI use based on the value they provide to businesses. We’ve even mapped the impact of these categories across different industries, identifying the specific use cases with the highest impact and shortest lead time to maturity. 

Using this information, we recommend investing 70%-80% of your AI spending in use cases that can be deployed within 18 months, reserving 20%-30% for exploratory use. 

Find out the top use cases for your industry at a glance in our CaseCompass guides

 

CaseCompass - Tailored AI strategies for your industry

2. Managing data  

It takes a huge amount of information to train AI models, and this often starts out as a disorganised mass of unstructured data.  

Before any AI implementation, companies should integrate the processing of unstructured data into their data management system, including tasks such as data cleansing to remove any erroneous or inconsistent information.  

This is not just a best practice, but a business imperative: poor quality data costs large companies over $12 million per year

In addition to data cleansing, leaders should also enrich their existing datasets by using them as the basis for synthetic data. Synthetic data reproduces the qualities of your existing datasets while offering additional benefits such as protecting customer privacy.  

AI can even assist you in improving its own training data quality. For example, you can use AI to spot discrepancies or to ascertain whether it meets completion requirements for certain classifications. 

3. Choosing an AI solution with growth potential  

Many generative AI models, such as OpenAI, offer ‘foundation models’ that are already trained on huge amounts of public data. This allows you to get up and running fast, but don’t rush into choosing your model. 

Select the right AI model based on its performance, scalability, compatibility, and compliance.  

You should also select the model that best fits your current culture and maturity but also leaves room for growth. For example, if you already have the capability to embed your own data into a GenAI model – essentially personalising its insights – factor this into your decision-making. If not, consider your timeline for personalisation and how easily your chosen AI model can accommodate this. 

4. Getting your people on board   

At the end of the day, AI cannot implement itself. It would get nowhere without your strategy behind it – and your people using it at the front lines of your business. If you can’t get buy-in from your employees, your AI solution is dead in the water. 

It is essential, therefore, to prioritise use cases that respond to real pain points or business problems experienced by your employees and customers. If they don’t need it, they won’t use it, no matter how flashy it is.  

The other barrier to AI uptake is expertise. Train a core team of AI ambassadors in each function or department of your organisation. These people will be the interface between their colleagues and the AI tech experts developing your AI solutions, ensuring that they respond to real business needs and employees’ skill levels. 

Finally, define and execute a strong AI talent strategy to ensure you have a strong base of AI AI skills within your organisation. Don’t just rely on hiring here: use upskilling and partner initiatives to bring the skills you need in-house. 

Target your AI strategy with our data insights 

The hype around AI might make you feel like you need to start implementing it everywhere, all at once. But, as the obstacles above show, that’s not true. You can’t apply a scattershot approach, or just go for the flashiest implementation – not if you want real results. 

Instead, you need to target your investments at high-impact areas and create the infrastructure for AI to succeed. You’ve got the leadership skills and critical thinking to ensure AI’s success: AI is nothing without you. 

For the data to support your decision-making, consult our report: Navigating the AI Era

 

Download the report

Search

artificial-intelligence

Related content

Responsible artificial intelligence

As organisations race to seize AI’s benefits, prioritising responsibility is key. Embracing responsible AI practices is not just about staying ahead but building a sustainable competitive advantage. 

AI and cybersecurity

In today’s digital age, traditional cybersecurity measures are no longer sufficient. Cyber threats are evolving rapidly, and adopting innovative solutions is essential to protect your business. Discover how AI is revolutionizing cybersecurity and giving you a strategic edge. 

Digital Banking Experience Report 2023 The AI-enabled banking era

Banks must leverage their trust capital if they are not to lose market share to tech giants broadening their offer into financial services. Our Digital Banking Experience Report 2023 outlines the key trends globally shaping banking in the hyper-connected era.