Imagine it’s machine learning and nobody goes there. The foreseeable end of business software.

Last week, on the occasion of the 20th anniversary (!) of the TopSoft business software trade fair, I had the opportunity to say a few words about the future of business software. I briefly explained why I consider machine learning to be a fundamental paradigm shift in technological development and why the digital industry itself will be caught up in the maelstrom of digital transformation as a result. A few further thoughts on this.

(Reading time: 5 minutes)

Paradigm shift

In his work,Ray Kurzweil ported the concept of “technological paradigm shifts”. In very simplified terms, this means that technological progress on one track is optimized as long as it is either no longer physically possible or no longer makes economic sense.

Once this point is reached, a paradigm shift takes place. This means switching to a different technology or type of technology and further optimizing on this. With the foreseeable end of Moore’s Law, we are at such a point. The transistor density in computer chips can soon no longer be increased any further. And new concepts such as the quantum computer or three-dimensional chips are being traded.

Software paradigm shift

If we look around in the software world, there is an astonishing amount of software that still works in exactly the same way as it did 20 years ago. An application that has to be installed, parameterized, operated and maintained. And a corresponding business model from the manufacturer that is based on it. In addition, the vast majority of such software incurs considerable direct costs.

I believe we are on the verge of a fundamental paradigm shift in this area, in which machine learning will play a major role. Here’s why:

Machine Learning

When I started working with machine learning 2 years ago, I thought that these concepts and the algorithms behind them were “rocket science”. I thought you had to be a math genius, so to speak, to be able to work with it. Gradually, I learned that machine learning is actually relatively simple. In very simplified terms, it’s about finding an equation that makes sense with the available data. And solving this equation or applying it to new data in order to make a classification.

There is now quite a lot of academic research into machine learning and the “mother field” of artificial intelligence in general. And yes, that is rocket science. In addition, there is GAFA (Tesla), which uses considerable resources to bring machine learning into its products and do things that amaze us: Facial recognition, predictions, autopilots, etc.

In addition, and therefore in 80% of software applications, machine learning is virtually non-existent. The common opinion is that it is so complicated and expensive to bring machine learning into software applications that it makes no sense.

However, anyone who has spent a little more time with it will immediately recognize that there are many smaller problems in the software environment that can be solved very well with machine learning in conventional software. However, since machine learning has this “magic sauce aura” surrounding it, most CTOs don’t even consider it at first.

Better today than tomorrow

Software manufacturers, and business software in particular, should get to grips with it sooner rather than later. Because machine learning suddenly makes completely new concepts possible.

“User experience should primarily be measured by the total amount of time spent using an application to perform a specific task.

”
If we look at today’s software, it is still far too complicated and time-consuming to use. We have become accustomed to this because we have developed an idea and a habit of how software should look and function. But if we let go of these ideas, we have to admit that the user experience is pretty disastrous.

Remove unnecessary work

Machine learning can primarily reduce unnecessary work in business software. By unnecessary work, I mean, for example, clicking on confirmations, searching for the same address in ERP systems, selecting printers. I know it sounds childish and unspectacular, but just look over the shoulders of your colleagues in accounting as they work. 50% of the time they spend with software, they do exactly these things.

So if we have software that reduces the workload by 50%, that’s not only great for the users, it’s also a blatant economic advantage. Quite a “no-brainer”.

It is not entirely clear why manufacturers are not investing more in this area. When I spoke to a representative from the product development department of a large ERP manufacturer a few weeks ago, he said dryly: “As long as we sell licenses per user, I don’t think we have any real interest in massively reducing user effort.”

New business models

If we completely break away from the idea of how software should be and use the possibilities of machine learning to solve specific customer problems, completely new fields open up. Whereas previously the customer was provided with the tool to develop the work result, machine learning has the potential to provide the work result directly.

With our start-up Accounto, we do exactly that for small businesses. Whereas it was previously necessary to purchase or rent an accounting program in order to do bookkeeping, Accounto customers only have to do what is relevant to their business: approve payments and supplier invoices, write customer invoices. Everything else is done by the software without their input and with the same or higher quality.

The really dangerous thing for traditional software providers is that the costs for traditional software will not differ greatly from machine learning-supported software over a long period of time. Because anyone who thinks they have to develop everything themselves is very much mistaken. There is already a great deal available that can be used. Providers who are unable to respond to this trend will end up with “old-school software”, which will cost the customer at least twice as much in terms of personnel costs, if not more.

I expect this to bring about major changes in the industries. What this type of technology does is that it can merge two or more industries in the medium term and thus replace them. This is possible because new processes and methods can be implemented with the new possibilities offered by software.

A first example of Accounto, and this is why we are in this market, is accounting and administration for small companies. What was previously provided by a tax consultant, the customer and a software provider in a complementary manner, we now implement with software.

Accounto’s long-term mission is to bring machine learning technology to the business software sector. We have chosen accounting and administration for small companies as our first field because it seems to us that many parameters are just right for implementing this on a broad scale.

Machine learning has its own rules

We also urgently need this breadth to advance machine learning, as we envision it in business processes. The more data we have, the faster our technology advances. The faster we can use it to solve much more complex problems faster, more cheaply and with higher quality.

The ideas (and the euphoria) in this regard are limitless, as the market constellation as it is found in the small customer accounting sector can often be found. The hope is that the concept can simply be copied to another vertical at some point.

Small steps

My advice to software manufacturers is therefore to radically focus on the customer and use machine learning to reduce the need for interaction with the software. You have to start with the banal, unspectacular things. The users will thank you for it and so will the customers/decision-makers.

Competitiveness

I think the necessary interaction time with a software will become an important evaluation criterion in the future. We compared various ERPs last year and the differences are huge. If I add up these times with full cost rates, this easily results in differences of up to 150% of full-time equivalents for a 150-person company.

The use of machine learning puts you as a manufacturer right at the forefront. Get involved in machine learning now. It’s high time.