August 08, 2017
Human after all: The missing link between AI and business benefit?
By: Christopher Walker

[Thank you to our collaboration partners for their input to this article, their guidance, real life examples and passion.  Specifically, Barbara Kennedy and Beverley George from HMRC; Sarat Pediredla, Hedgehog Lab and Dr Yifeng Zeng, from Teeside University]

Artificial Intelligence (AI) has come seemingly from nowhere to become one of the most important emerging technologies for all businesses in all industries. Like the mobile internet, the cloud and other transformational technologies before it, AI is disruptive: those businesses that fail to act – or fail to implement AI correctly – will quickly fall behind those that get it right.

In any major transformation, the hardest thing is knowing where to start. For that reason, Accenture recently brought together a panel of experts to discuss the AI projects they’ve already brought to fruition. Participants from HMRC, Hedgehog Lab and Teesside University shared their experiences in robotics, AI and Machine Learning; and in the process provided some useful guiding principles that all businesses can follow. 

What follows below is a summary of those principles: six important steps that all businesses should take when they implement an AI system. Counterintuitively, these steps come back to one overriding principle: the most important element to success in AI is the people who manage the technology. We may well be entering the machine age, but humans remain of prime importance.

The six steps are:

  1. Get the right foundations in place

    First, you must identify a use case where AI can add value to your business. It may sound obvious, but you must know exactly what you want to achieve through AI before moving forward. From there, data is key. As with any intelligent technology, AI is only as good as the pool of data it can draw on. Once you’ve identified your use case, consider what data will be needed to support it. Have you got access to this data, and can it be seamless integrated with the AI technology?


  2. Educate and manage expectations

    If you want your customers and internal users to keep using an AI application, you must get the initial experience exactly right. A bad experience at the outset may well be the difference between success and failure. It’s therefore vital that you take your users with you through the development of any AI application, and make them feel like part of the test team. Build an understanding of iterative technologies within your organisation and only release an application to the customer when you’re sure the experience will be, if not perfect, then certainly compelling.

    “From the pitch sold we had high expectations of delivery and results in a short 3 month period – reality is 12 months on issues are still being addressed and benefits are low.  I think we were a little over ambitious and need to start small, build up confidence and ramp up as knowledge and expectations are understood.” A client breakfast attendee.

  3. Manage the handoffs

    Currently, not every customer interaction can be managed by AI alone and humans must often step in. What’s of concern is that this ‘handoff’ from AI to humans can add friction to the customer experience. For example, having to authenticate with a person, even though you’ve already been through a personalised automated welcome with a chatbot, can be frustrating.

    What’s more, different customers will want, expect or need different experiences – one size won’t fit all. Have a think through the various possibilities of the overall experience, and see if you can create seamless joins between human and technology interactions. You want the outcome to be an enriched experience for your users and a better customer service.


  4. Technology as commodity

    Today’s ‘must-have’ technology is tomorrow’s commodity. To avoid the commoditisation of your AI systems, focus on outcomes rather than technology. Remember: anything that can have a significant business impact is not going to be ‘plug and play’, so ensure you have the people, senior sponsorship and a commitment in place to get the best out of the system.


  5. Measure what’s important, not what’s easy

    Often the easiest things to measure are the least important. In the world of AI, it’s as simple as anything to measure, for example, the number of transactions processed – but when it comes to the crunch how useful is that information? It only tells you what’s happening, not why it’s happening. To ensure your implementation is doing everything you need it to, you should instead measure outcomes – in the case of our example: what value the transactions add. As your delivery model evolves, so too must the way you measure your performance - KPI’s and metrics need to evolve with the technology you are delivering.


  6. Consider governance and ethics from the outset

    Let’s be honest: AI will bring with it certain challenges. These range from its disruptive effect on the workforce, to the ethics of how personal data is used in machine systems. You need to put in place a set of guidelines around how your business will adapt to these challenges, and how you will deploy AI systems in a way that is both good for your business and socially responsible. 

    AI is not a matter of if – it’s a matter of when. Our advice is for businesses to start experimenting with AI applications as soon as possible. First-movers in this space will, after all, be the business leaders of tomorrow.

The IEEE Standards Association have published some useful guidance regarding Ethical Considerations in Artificial Intelligence and Autonomous Systems here

Popular Tags

    More blogs on this topic