AI Best Practices in the Real World - Dr. Merlin Stone discusses Artificial Intelligence in the Near Future, Part II
We continue our conversation with Dr. Merlin Stone about the proliferation of artificial intelligence in the real world. Does AI's use of algorithms always lead to smarter decisions? The answer may surprise you!
Why More Mobility Is Not Always Better
Peter: We’re back again to discuss how AI can use algorithms to make smarter decisions starting with transport investment. Surely AI can be applied to help political decisions around HS2 (a high speed rail link between London and Manchester now estimated to be needing a budget of $100B+), or the third runway for Heathrow. How many times do these ideas need Parliamentary approval? These cost huge sums of money that could be better spent elsewhere such as National Health Service (NHS).
Dr. Stone: Indeed! Don’t give the NHS any more money. It needs less money and better governing. HS2, you're right. Did you know that HS2 doesn’t even go to Heathrow?
Peter: Last time we talked about the use of big data for people living in large cities and staying at home to be more efficient by taking a later train. That’s just time shifting isn't it?
Dr. Stone: Yes! Time shifting is important, but there is an argument to reduce travel time to improve efficiency overall.
Peter: But if you take 20 minutes off the journey time from Manchester to London, will it really achieve anything?
Dr. Stone: Timesaving, is a different argument. The HS2 case was partly made on travel time savings, but if you identify that business people work on the train, it's not worth debating. Instead put the money saved into housing, which is a huge issue in the UK. I suggest we cut taxes and create a more efficient travel network, not another rail service.
Let’s talk about why we assume that more mobility is a good thing. There's the concept of ‘mobility as a service,’ which is this latest dream the civil servants have — the idea of driverless cars. Nothing beats a bus with 40 people in it! Do you want 40 driverless cars, even if they're all shared, clogging up our roads?
Actually, my answer to London's congestion is to pedestrianize more because intelligence, it's taken us a long time to do what other cities have done — it's been on the cards for about 30 years. I think that's a good return. And retailers have been scared of it as they don't understand how good it is that people who feel relaxed when they’re shopping, and not worried about the amount of traffic congestion, are going to spend more … who should be told to work late in a very congested city — that requires AI.
AI in Prospect and Customer Management ... the Next Frontier
Peter: Let's bring this back to marketing! You and I have done a lot of work on the B2B side, and I haven't seen companies change much in their scant regard for how they store data and keep it up to date. There isn't any data to apply AI to!
Dr. Stone: To me, developments like LinkedIn have been very good, because they've forced companies to say, “Suppose there was all this data with all my buyers there, and it's not my database, but it's the data that's out in the public domain that covers all the buyers that are out there in the market – How can I use that data?” Still very basic, and there’s a lot more that Microsoft could do as the owners of LinkedIn to really turn it into a fully featured B2B database.
The work done by some of our clients in terms of response management is pretty cutting edge, but too much has been done with prospecting and not customer management. It’s always a problem. I'm optimistic about that. The issue though is that this B2B stuff is so much more complex. Whereas with consumer it's personal, maybe one or two people at most; but in B2B, it can easily be 30 or 40 people and the budget cycle could last a year or two. You'd have thought more work would have been done in this area - because of the complexity people will organize the data more and use AI more in future for sure.
But I haven't seen that yet. I’m sure it’s on the target list for the more aware companies like Hyster-Yale; they’ve created a leading edge approach to prospect management and customer management because often most of their sales take place to existing customers. They’re two completely different things, but prospecting is more difficult due to the length of the buying cycle.
Peter: But is this something that proves marketing can really support sales, and even be perceived by the sales teams as a good thing to have?
Dr. Stone: Exactly; because you don't really need so much AI for consumer — if a consumer says they’re interested, then they’re interested. Whereas a giant corporation like Amazon building a new warehouse — the number of people involved in that to grasp all the elements surely becomes a complex diagnostic problem that's similar to diagnosing a human to see whether they've got cancer. So let’s say a company looking at the pattern of browsers on their website, inbound traffic on LinkedIn and a whole variety of sources can say, I can see this coming and also things like planning permissions — someone’s building a warehouse, so we need to sell them a truck that would be the equivalent, but it’s all very manual at the moment.
Manufacturing vs. Services ... AI Can't Replicate Human Touch
Peter: So there’s no linkage between systems currently?
Dr. Stone: None. So what we’d dream up as the ideal system would be quite something: a unified database, or the equivalent in whatever industry, planning it as you would for a physical facility. But there must be a similar situation that could be triggered by industry.
Peter: Or even sensors on the trucks so that you can actually see how a customer is using the truck and possibly see that it isn't very efficient the way they’re doing it; we can replace that with three machines at a fraction of the cost and improve your productivity by x percent.
Dr. Stone: Yes, we are seeing this in some industries.
When I was at IBM we went to visit Caterpillar in 2003 - I said you guys know more about the mining of ore from the ground than anybody because you're selling all the diggers and trucks involved. Surely you should be able to sell excavation as a service. Why are you selling trucks? Ooh, we couldn’t do that — we can sell financial services, but this is a bridge too far. But it's happening in other industries. So, one of the triggers to that is, when you move from selling the asset to selling the service, you must then have the data so you can now see these trucks in operation.
So, for example, when I was in Denmark for IBM, you got food processing companies whose equipment is full of sensors. The client needs it for quality control and you can use it to identify how the clients are using it. Or Vestas, the wind turbine company — their equipment is full of sensors. You'll tell your clients whether the machines are working efficiently, and the sensors are working for two purposes. One is for you to service a machine; the other is for them to decide whether it's worthwhile or not with wind directional data and so on, but you have the database of all of it. So you can advise them what to do, and you can even see a weather front approaching from another country and pre-set the equipment.
Peter: That’s all very well for manufactured products, but what about services?
Dr. Stone: Most business we sell in the UK is services-related: consultancy and a whole range of different service products — all the work going on in the Middle East, that's very different; it's very much human-based, and I can't see AI working as a delivery mechanism there. There's still so much that is hand-crafted; writing proposals — you know the story. You would have thought that AI would be able to turn to anything like proposals because there’s nothing new under the sun.
Peter: Just plug it into the web and find all the research!
Dr. Stone: Absolutely.
Peter: We can dream on about that, and it will become interesting to see how you then actually differentiate between different bids in the future when we’re all delivering the perfect solutions! We’ll need AI Max to then analyze the massively fine detail and show the results in advance of whatever has been proposed … WOW!
Let’s stop here before our heads blow up, and come back to earth in Part III!
Related story: Replacing Unskilled Data Marketers With AI