(N.B. If you’re interested in marketing applications of AI, Econsultancy’s Supercharged conference takes place in London on May 1, 2018 and is chocked full of case studies and advice on how to build out your data science capability. Speakers come from Ikea, Danske Bank, Just Eat, Age UK, RBS and more)

Purpose

AI for what? How can AI help your organisation? What business problem are you trying to solve?

AI is good at targeting ads, product recommendations, deciding if someone is likely to repay a loan, face and voice recognition, even driving cars. AI is not good at more profound thinking such as creativity and innovation.

The economics of your business can help guide you to areas where AI can add value. Customer acquisition, conversion and retention are good places to start. Consider the whole customer journey. 

Anything that you can automate is a good candidate for AI. Gartner forecasts that by 2020, customers will manage 85% of their relationship with the enterprise without interacting with a human. That will require a lot of automation. AI can help with the necessary personalisation of targeting, content and customer journey. Optimising those three things in real-time is better suited to AI than a human being.

Benchmark your competitors and similar industries for potential AI applications. One of the reasons AI is not more developed is that many organisations are yet to work out what they can do with AI. With AI answers are easy, it’s the questions that tend to be more difficult.

Predictive data 

The increasing power of computers and the greater availability of digital data have fuelled the growth of AI.

First ask yourself if you can access the data you need to fulfil what you want to achieve with AI? Ideally you want a single customer view. Data silos can be a challenge. Data integration tools such as tag management and APIs are improving things but we still have a long way to go. Data is often the largest part of an AI project. Two thirds of the work in data mining projects is typically data preparation.

As with all things IT – rubbish in, rubbish out. Data quality is important. Data needs to be predictive. If you have lots of data variables, by the time you add the last variable it is less likely to add much value to the overall AI solution. The answer may have already been found with the previous data. Data gives diminishing returns, especially if it is all saying the same thing. For example, customer databases in the financial services industry normally contain lots of wealth indicators. Ideally you want different types of predictive data. 

Sustainable competitive advantage and barriers to entry tend to be data rather than AI related. Data is easier to protect than AI. Data is less easy to copy than an AI algorithm. You do not need to own all the data, just enough to put competitors off from copying you.

People 

AI needs people to make it work. Similar to how digital started, develop a centre of excellence. Centralise your AI task team. Then over time integrate AI resources into the business units.

Recruiting and retaining AI people is hard. Good analysts are like gold-dust. You want people who are good with numbers and analysis, communicate well and understand your business. Not an easy combination to find.

Do you use internal or external personnel? It’s often best to pilot AI externally and over time bring proven AI value in-house. External resources may be more flexible and knowledgeable about the many different AI possibilities. Internal resources will work out cheaper and may be better at optimising your AI solutions in the long run.

Process

Building an AI capability is not a short term project. It is a long term process. Changing your organisation’s culture to embrace and leverage AI takes time. 

Initially keep it simple. Walk before you run. Start with quick wins to build corporate confidence. Things change so AI algorithms need to be regularly updated. You need an optimisation process to continually improve them.

ROI is obviously important. The incremental improvement AI brings has to outweigh the total costs of implementing AI. That is why testing is vital. AI is ideally suited to digital businesses that have developed an agile, data driven, test and learn culture.  

Platform

Last but not least, which AI platform? I say last because the choice of platform should only come after the four Ps above. Don’t rush into the mistake many marketers make of buying the latest technology and then wondering why it has not fixed the problem. 

There are many AI platforms available from cloud providers such as Amazon, Microsoft, Google and IBM to many new start-ups. Which of the numerous tools available is best for your organisation will become clear over time. And you do have time. AI is not built in a day. It’s much more important to grow a strong AI capability than finish a few quick and dirty AI projects.

AI is poised to unlock incredible value for marketers. Define your goals, then test and optimise your AI processes. Develop talented people who can capitalise on AI. The next industrial revolution will come when enough smart people start asking the right questions about what AI can do.

More on AI strategy for marketers: