Why We Must Think Cross-Vertical

Have you ever contemplated to buy a car that is made to exclusively ride between Munich and Hamburg? And another one especially built for trips between Munich and Berlin?

Certainly not. Curiously this is exactly what we are doing when it comes to artificial intelli­gence. So many people in business and even investors who claim to shape the future, look at AI with limited use case driven lenses.

As a consequence, they get what they ask for : AI solutions narrowed down to their current use cases (reminds me of Henry Ford’s quote about faster horses). Next use case, next individual solution.

This is great for AI solution providers, but it is a tremen­dous waste of economic resources. And that’s why we observe a confusing & inflating AI landscape of products and solutions.

It is a ground truth, that we have a limited set of algorithms, suitable to build a limit number of data products.

No doubt, it is still a challenge to decide which business problem can be solved with machine intelligence and which cannot.

But, does this current challenge justify inferring that anomalous custo­mer, machine, network and other behavior is something completely different and must be computed by completely different approaches?

And, thinking within the limits of verticals and specific use cases is dangerous:

What if we predict the next best price based on customer data compromised by a cyber attack? Telemetry data from connected cars or other devices that reflect a synthetic picture created by a bot?

AI solutions exclusively narrowed down to either E-commerce or Internet-of-Things never tackle this dimension. And it is not restricted to a mix-in of cyber defense.

What about beacons in brick-and-mortar stores and a mix-in of sensor readings and customer purchase transactions? We could provide an endless list of examples, but the message is always the same:

Traditional “divide-and-conquer” thinking is not suited to tackle world’s complexity.

Machine intelligence is made to take pictures from enterprise data and we need to mix-in many of them from many different perspectives to see the truth.

If we do not change our perspective from verticals and specific use cases to data, no digital transformation will ever end in a future-proven AI-driven enterprise.

Sketching a problem is easy. What is an appropriate solution?

We suggest starting from data (and its transformations) and proceed upwards to use cases:

• A limited number of algorithms, data operations and data products imply that we should build and use a Lego like toolkit (or AI Operating System) with a predefined set of AI components and connectors to data sources.
• Leverage this toolkit without writing a line of code to build data lenses that are suited to take snapshots of your data: anomalies, forecasts, recommendations, similarities and other data products.
• Let the toolkit automatically connect your data products, form a holistic picture and serve a specific use case. Next use case, same toolkit and same procedure.

At this point, you may argue that you do not know which Lego components to choose to point and click a certain data lens. What if these components (and assemblies of them) tell you that they answer questions like “what will happen?” or “how can we make it happen?”.

Still too complicated? Ok. What about configurable AI blueprints based on these Lego com­ponents, that you can use to find the best price, determine the next unscheduled main­tenance and also identify fraudsters? All based on the sample platform.

We talk about configurable AI templates based on the same technology and designed to address many different business cases and mix-ins of them.

Always the same mechanism: Select an AI blueprint, do some minor adjustments and automatically create an AI solution to face a certain business case. No matter whether the problem is within or cross-vertical and you are prepared with the flexibility to respond to world’s complexity.

With this in mind, we could start thinking about industry-specific or cross-industry blueprint catalogues, sharing or even standardizing blueprints.

Artificial intelligence no longer exposes itself as a plethora of rigid narrow point solu­tions originating from a “divide-and-conquer” mindset fallen out of time.

Originally published at https://www.linkedin.com.