Ambitious management teams are constantly looking forward and considering how to grow their business. Artificial intelligence (AI) is likely to form a big part of this, no matter the size or sector of a firm. Phil Harvey, Microsoft Cloud Solution Architect, speaks with Inflexion portfolio companies about how AI can help.

There has been a large explosion in data in recent years. What is the upshot?

Data gives us new ways of knowing, and new things to know. This is a fundamental change in our understanding of the world. Data makes software smarter; it’s how machines learn.

This data comes from multiple sources – it might come from customer behaviour, operational processes, telemetry embedded into products, or employee feedback – but in all of these cases it creates a vital feedback loop that allows us to continually optimise our businesses.

The transformative power of AI is that it then allows us to use this wealth of data to answer questions that previously would not have been practical to address, or in many cases very costly. Phil shares with us example of a cruise ship company, who used predictive analytics to forecast water usage, and therefore avoid costly overproduction – saving $200,000 per ship per year.

Accessing these transformative capabilities is getting easier – not only do we have access to scalable, on-demand infrastructure in the cloud, accessible to even the smallest businesses, but also frameworks like TensorFlow (originally developed by Google) and PyTorch (originally developed by Facebook) are freely available as open source, and there are a wealth of pre-trained models available on platforms like Microsoft Azure to simplify and accelerate development.

Ethical considerations
  

Just because we can do it, doesn’t mean we should – ethical considerations are key to responsible use of AI.

Phil shared Microsoft’s AI principles based on Fairness, Safety and Privacy, alongside Transparency and Accountability. Issues around bias in AI and accountability have been well-publicised in recent years, with cases ranging from Amazon’s CV screening algorithm learning from historic (human) bias, to Microsoft’s own ‘Tay’ chatbot which had to be taken down after 16 hours having been targeted by ‘trolls’ teaching it offensive behaviour.

Explainability is another key consideration with increasing automation of decisions – “sometimes when we’re developing models, we need to prefer explainability over accuracy,” Phil comments – emphasising that this is not just important in terms of being able to ‘sense-check’ output of models, but also to build trust with all stakeholders.

Data scientists

“As recently stated by our CEO Satya Nadella, at Microsoft we believe the ability to use AI to convert data into insight and action will be the core currency of business in the future,” says Phil – but he cautions it requires an investment in skills in order to take full advantage.

“Data scientists are in short supply, and it can be challenging to find good ones,” Phil admits. He offers three suggestions:

1. Develop them internally. By doing this, you have talent which already understands the business and the data powering it, and so can comfortably rise to the challenge of learning and growing this.  Crucially, they can explain the impact the AI is having in a way directly applicable to the business, rather than in an abstract way which a third-party supplier may be limited to.
2. Test the scientist against the business problems instead of against the technical problems. Ask the scientists how they’d apply the technology against the business problems, because most data scientists know the tech, but a good, competent one will apply it well, too.
3. Consider ‘try before you buy’ data scientists. One of Microsoft’s partners runs a an intense five-week “boot camp” which applies data science to your problem before tackling your own business problem. It’s been hugely successful.”

Above all, Phil advises, “Focus on where your core business value is, since then you’ll find areas where AI can help.”