5 Steps To Success In Artificial Intelligence

How can businesses avoid failing in their approach to AI

This July, a survey of 2,473 global organisations by the International Data Corporation found that 50 per cent saw artificial intelligence as a top priority. Around 25 per cent had successfully implemented a company wide AI strategy, while over 60 per cent had altered their business model in some way to do so. Despite this enthusiasm, a quarter of respondents reported that half of their AI projects had failed.

It comes as no surprise that many organisations experience teething problems when trying to integrate artificial intelligence into their operations. Artificially intelligent technology costs money, can require specialist skills, and relies on the collection of quality data. But, nonetheless, it has become an important business tool with an inexhaustible list of benefits. So, what steps should organisations take to ensure AI success?

1) Good data governance

In order to benefit from AI, data scientists need accurate, up to date, structured information. Artificially intelligent software is only as good as the data it is fed. If an organisation keeps masses of data that is outdated or inaccurate, their algorithms will be of little use. Dirty data, dirty insights. It’s that simple.

2) Pull in project management

Hilary Mason, Cloudera’s General Manager of Machine Learning, believes that AI success goes beyond data to the ability of an organisation to set and achieve goals. Machine learning, says Mason, should be applied “on a sustained basis” in a way that is best for the organisation. This means pulling in people from across the business and pooling their knowledge so that AI applications truly meet the needs of clients and customers.

In the DevOps movement, project managers and data scientists collaborate to make the most efficient, effective software. DeepOps, a new approach, has its foundations in DevOps but focuses on deep learning and data insights. In each case, tech is created and utilised with input from across different teams. This avoids confusion and encourages working relationships that are optimised for cooperation.

3) Invest in skilled staff

Before employees can build and implement AI, they must be equipped with the right skills. As well as hiring talented employees, organisations should consider reskilling their existing workforce. The availability of massive open online courses (MOOCs) has made reskilling much easier, as have coding bootcamp providers. This avoids the myriad of costs involved in hiring a new employee, such as training and decreased productivity.

4) Set a budget

Not so long ago, homegrown AI was only possible in the tech teams of huge companies like Google, Apple, and Microsoft. Today, artificial intelligence is more accessible through Software-as-a-Service and ready made platforms. 

When built from scratch, however, AI projects can easily come with a six figure price tag. This is largely due to the various stages of development, beginning with discovery and analysis. The next phase is creating a prototype, testing it, and then releasing a minimum viable product (MVP) to gauge the reaction of customers. The last stage is the final product release. There are ways to make the project cheaper, including the use of clean data, skilled staff and a coherent, organisation-wide approach.

5) Be realistic

The final and perhaps most important thing that businesses should do when adopting AI in any of its guises is to be realistic. In other words, what is the problem that needs to be solved, and how can artificial intelligence solve it? What capacity does the organisation have to build a coherent AI strategy? Is it worth the cost? Unless it significantly improves the organisation and better serves clients and customers, pouring time, money, and effort into an AI project makes no sense. This is another reason why merging tech teams with project management teams is so important. There’s no point building an AI strategy just for the sake of it.

Companies are already making the mistake of starting AI projects that they can’t finish. Most of the time, this is because they underestimate how much it will cost. This isn’t just about money – building AI more or less from scratch involves specific skills, a lengthy development process, and buy-in from stakeholders and shareholders. AI projects have to answer questions that can’t be solved by another, simpler method… If not, they are far from a smart move.

Stay in the know about artificial intelligence by signing up for our weekly insights here