When AI marketing is more artificial than intelligence

Credit: freedomnaruk / Shutterstock.com

The bar for what counts as artificial intelligence is continually rising. Technologies should lose their AI status when they become routine (optical character recognition is no longer recognized as an example of AI for this reason). However, for marketers it’s tempting to go in the opposite direction and rebrand all kinds of routine technologies as AI in order to make them sound more exciting and newsworthy.

AI was a key theme at this year’s Mobile World Congress. One AI-themed announcement was “aia”, from telecoms IT giant Amdocs, which claims to enable the “self-driving telco”. I must admit to being quite impressed by the strapline, but I wonder how much of what aia does is really AI and how much is just routine technologies rebranded.

Aia integrates IBM’s Watson AI technology with Amdocs’ systems for managing telecom business processes. The idea is that these processes (more than 50 of them) can dynamically adapt and optimize themselves based on self-learning and AI. Amdocs cites a few examples in its press release and marketing collateral, including:

  • Suggesting a new data plan when the customer is about to exceed their bundle
  • Providing a real-time response to a poor network experience (e.g. offering compensation)
  • Suggesting optimal pricing for new products
  • Managing conversations with customers, including next best action and next best offer
  • Improving predictions of customer churn and network failures
  • Detecting cyber-security anomalies.

These are all challenging problems which today are handled by existing advanced, but nevertheless routine, technologies. If aia is really a breakthrough, it must do more than simply repackage the things which today we call business intelligence algorithms, closed-loop marketing campaigns, chatbots, A/B testing, social media listening, fraud detection and other IT systems under a sexy new AI banner.

For me, the key question is whether a human brain is needed: are the advanced algorithms used to solve these business problems programmed with human-inspired (albeit complex) business rules, or are the machines actually thinking for themselves and working out what to do in any given situation?

For example, let’s suppose your customer is running low on their data balance while watching YouTube on their mobile phone – this might be an ideal time to offer that customer a new plan with YouTube data included. Today’s advanced real-time telco IT systems can do this, but they typically provide only the “central nervous system”, not the “brain”. The IT systems collect data in real time (customer is running low on data; customer is watching YouTube), and execute an appropriate response (send marketing campaign for YouTube inclusive data plan), just as the body’s central nervous system gathers data through the senses and sends instructions from the brain about how to respond. However, the function of the brain itself is usually handled offline by a real person, designing a business rule, algorithm or campaign which is triggered in real-time by network events and customer behaviors.

Real AI is about no longer needing a human to provide the brains of the operation. The IT systems would spot the opportunity for the YouTube promotion and execute a personalized marketing campaign without anyone having set this up beforehand. Indeed, aia invites us to “imagine a world where you automatically know which product to launch before you’ve even considered it,” implying that the machines could even create the YouTube data plan automatically if it didn’t already exist in the product catalogue.

This sounds really impressive, but my experience tells me that it’s not realistic today. Telco IT systems and business processes are incredibly complex, and business process definition (let alone automation) is usually the most complicated part of any IT transformation project. What’s more, consumers (not to mention regulators) are extremely sensitive when IT systems go wrong or mishandle data. Are we really ready for the driverless telco?

The analogy is apt – driving a car is a well understood discipline which many people can do very competently already. Training machines to do a well-understood task (even if incredibly sophisticated) is possible. However, few telcos would claim to have really got their IT systems running smoothly, especially for the new digital era. I’d be extremely surprised if they are ready to let the machines take over business critical IT operations just yet.

That’s not to say Amdocs is on the wrong track – applying AI techniques to telco business processes is a very worthwhile goal. I don’t doubt that aia is worth a look, even if it doesn’t comply with the definition of true artificial intelligence. But marketers can easily get divorced from reality, unrealistically extrapolating the potential of new technology (and thereby undermining its credibility) in order to ride the latest wave of hype. As a marketer, I dare say that I have been guilty of this, myself. The real world is so much more difficult than marketing, and it’s only when I spent six months working on an actual IT transformation project that I began to appreciate the huge gulf between the concept demo and a real operational system.

Maybe I’m wrong, and we’ll soon see proof of real-world use cases for AI-enabled business support systems. But my sense is that the AI tag had been used rather inaccurately in some of this year’s MWC announcements – unless AI actually stands for “Artificially Inflated”?

Written by Andy Tiller. This article was first published on LinkedIn.

Be the first to comment

What do you think?

This site uses Akismet to reduce spam. Learn how your comment data is processed.