Tag Archives: analytics

Moving Your Artificial Intelligence Strategy Forward

The artificial intelligence strategy buzz is everywhere. But now we need to move beyond rule based personalization and predictive analytics based recommendations. That kind of AI has been going on for 20 years if not more. Calling every analytics based project AI using a lot of jargon is definitely fashionable. But it also adversely affects the maturity of your artificial intelligence strategy.

artificial intelligence strategy hand-1571842_1920

Our Artificial intelligence strategy should be about traditional statistical approaches plus much more. That “much more” comes from creating conditions that allow your system to learn from data on its own. This learning includes making decisions based on information about both good and bad outcomes. For example, your system monitors commerce data and notes that a purchase is a good thing to have – “a win”. It also figures out that this “win” is positively influenced by offering promotions. And it then figures out when to offer someone a promotion that results in a purchase – a win. And it keeps updating its strategy based on “wins” and “losses”.

In short the ability of your system – not you manually – to figure this out, and learn by itself is already the current frontier in artificial intelligence strategy.  We can sometimes help move this process quicker through “supervised learning”. And our goal should be to constantly expand the boundaries of what’s possible.

So here are 3 tips that I hope will help you advance your artificial intelligence strategy:

1. Always be feeding

Not all good outcomes are actually good – especially outcomes that are limited by channel or product. For example, in the previous paragraph I noted that a purchase on the website or through a chatbot is a “win”. What if I told you that this specific type of customer has an overall negative lifetime value unless other conditions are also satisfied? In addition, what if the product that the customer is browsing is not a good fit for them? In that case even if a sale is good for us in the short term, should we continue to push that product to the customer?

Thinking about the big picture helps you improve your artificial intelligence strategy by constantly feeding updated definitions of what good outcomes are. It can also help integrate the pockets of AI that mushroom everywhere. As this field evolves, narrow applications are natural to implement – such as an NLP based chatbot for customer service. However, the incentives in the artificial intelligence strategy roadmap should be aligned to the goals of constant “expansion of context”. It’s like feeding insights into a central superpower, however menacing that may sound.

2. Always be rewarding

Your artificial intelligence strategy will depend on how well you provide the notion of rewards to your algorithms. After all, to decide what to do, the algorithm must know what happened the last time it did something. And that means that the rewards and penalties should be easy to come by.

Contrast this with how we implemented analytics in the past – we ran regressions, figured out what would influence an outcome, and baked those into rules for our processes to follow. And then we repeated the process often. But now you must build your data and technology strategy around the artificial intelligence strategy. How would you transmit those rewards and penalties to your algorithms?  For example, automatically re-balancing a wealth portfolio based on market triggers is good, but we must also let the AI algorithm know whether this was a “win” or “loss”. That way it can learn and improve. Otherwise it’s just old wine in a new bottle.

3. Always be connecting

As I have outlined in my book “Connected! How #platforms of today will become apps of tomorrow“, the customer experience spectrum is way bigger than our own products and channels. Every company today must be connected to others that serve the same customers. For example, we should think of feeding our algorithms with data – with customer consent of course – about the customer’s overall identity, intent and interactions. This data can come from interaction in retail, fitness, travel etc. And we should be building two way connections between companies to help customers make better decisions.

If we don’t think of connecting with partners to enable cross-industry customer journeys, our artificial intelligence strategy will only partially succeed. We cannot be building jazzy gizmos and expect to survive when competitors (especially the new entrants) are creating customer ecosystems and expanding their product offerings. In fact most innovations today bring together customer interactions across industries. Customers receive tremendous value from these cross-industry customer journeys. We should become better partners to customers too.

In summary

AI is an evolving field but the business landscape is changing at an even faster pace. Your artificial intelligence strategy should look to get beyond its own buzz and complexity. I hope these 3 simple tips provided some perspective so you can give your artificial intelligence strategy a boost. The first step is to not lose sight of defining the right customer experiences and journeys in an increasingly connected world. This definition of purpose and outcomes will automatically help you get ahead of the quest for data as well.

PS:

  1. Image by Geralt. Also I apologize in advance to AI experts for my simplistic treatment of this topic but some good old grounding was due.
  2. Please download the principles of how to define cross-industry customer experiences and personas here, or just get them for free in your inbox.