Blog overview

Part 5: The Difficult Path Towards a Webservice

How webservices can fail, even in a native
digital world

KI-UW0

Uwe Weinreich, the author of this blog, usually coaches teams and managers on topics related to strategy, innovation and digital transfor­mation. Now he is seeking a direct confron­tation with Artificial Intelligence.

The outcome is uncertain.

Stay informed via Twitter or linkedIn.

Already published:

1. AI and Me – Diary of an experiment

2. Maths, Technology, Embarrassement

3. Learning in the deep blue sea - Azure

4. Experimenting to the Bitter End

5. The difficult path towards a webservice

Service is key to a business that moves more and more in a digital direction. In the early days of my career, Jürgen Fuchs, chief representative at CSC Ploenzke AG at that time, once told me while talking about how to acquire more customers:

Don't sell to your customers - make it easy for them to buy from you.

I couldn't forget this sentence, although I have to admit that when I heard it I thought by myself: easily said, old man!

Meanwhile, I am as old as he was then and I have to say, it is true and it is a good reminder that not a massive marketing budget is key to success. Ok, it might help. But it is always much better to create a convincing value proposition that motivates a boisterous eagerness to buy your offer. Sure, equipped with a good idea and a abundant budget, that works best.

To make it easy for customers means to be open-minded and offer a service that customers realy need. Nowadays good services don't have to rely on humans any more. Machines start to communicate with each other and offer each other services without any human assistance. Connected via internet and related protocols they form the Internet of Things (IoT) which in an industrial context is also called Industrial Internet of Things (IIoT). Today, there are already far more machines communication via internet than human beeings!

To make such a network of computers work, well-functioning interfaces, so called Application Programming Interfaces (API) are required. Various architechtures like REST or SOAP define standards.

A webservice should alway be faster and more reliablend zuverlässiger seinHow does a webservice work? A calling instance (computer, machine or even person) sends a request to a server and authorises it with an API key. The server processes the request and sends back the answer that was extracted from the data. It is the same procedure as to order a drink on a sunny terrace and having it delivered by the waiter, but much, much faster and more reliable.

That should be enough to outline the technical aspects. Much more interesting is the strategical perspective. APIs open the doors in many directions and allow for automated collaboration.

An API is the collaboration model of a company, moulded in program code.

That can be of enormous importance. How an API is designed, whom companies gain access and certain rights, determins substantially how the company is integrated into an eco-system of value production. This API-based definition of the compyny's role in a network is more fundamental than any contract can be. This is why manager should spend due dilligence on designing webservices and APIs.

Webservices and artificial intelligence

Attentive readers might ask, "now, what has this to do with artificial intelligence?" Good question. In principle nothing. Services and APIs exist independently from AI. Nevertheless, those technologies combined can unfold tremendous power:

Now, let's go for our own service

These questions of monetarisations are not discussed in our Azure course. The next task is to create a service that can be called via internet. It shall calculate the probability of having diabetes based on the data that is sent to the service.

Webservice-Erstellung ist automatisiertWith Azure it is really possible to solve this task with a few clicks. Creating a service is an integrated part of Azure that can be started vrom any experiment design.

That was the case when I did it some days ago. Today, when I tried to continue where I stopped, I got confused. It was easy to log into Azure and to find the project. There was also a link to the AI workspace, but unfortunately it didn't take me directly to the previousy fnished work but another log-in page and then - nothing! An empty list.

azure13

It tool half an hour to find out that I need to click on the directory symbol in the header and to choose the appropriate server first. Dear Microsofties, that could be done better. Would you please keep servie in mind and to make it easy for your customers? In the end you want me to buy from you, don't you?

The next step would habe been to consume this service with Excel. That sounds like a terrific service because you wouldn't need a seperate web interface. But I failed once again. The connection requires a certain add-in and Office only gave me an error message that it would be impossible to buy this add-in without any hint how to solve this problem. Annoying.

Luckily, our server administrator was able to solve this problem. And indeed, now I can test with excel whether a data set of a person indicates that he or she has diabetes (column L = 1) and the probability (column M). Finaly.

azure-14

 

The next task was much easier although it looks pretty technical: a web-service coded in python 3.6. That's it:

azure-15

Looks cool, doesn't it? To be honest, I haven't written a single line of code. The system did it for me and I only had to copy the result. And it worked!

A journey from learning to service to learning

What do we have now? The machine has learned to predict diabetes based on the provided data. No, that is actually not the fact. "Predictive" sounds cool nowadays but the truth is no prediction but only an automated data analysis and interpretation of the outcome. Anyway, the result is not too bad, but it shouldn't be overestimated.

What else do we have? The insight that services and not at least the user experience regarding the usability and navigation in complex systems like Azure still doesn't meet the ideal of Jürgen Fuchs, that it should be easy to use.

See you soon.

 

⬅ previous blog entry  

published: July 19, 2018, © Uwe Weinreich

new comment


privacy