AI or automation? Which path to take?

This question is not just a technological one, but primarily a business one. Why not get swept up in the AI fad at all costs and instead look for smart solutions?

The reality of AI projects: 90% end up in the pilot phase

Artificial intelligence is undoubtedly a fascinating technology. But in practice, we face a harsh reality: nine out of ten AI projects never make it past the pilot phase. Implementation typically takes 18 months or more, and the results often fall short of initial expectations. This is not because the technology doesn't work, but because a crucial requirement for its success - good data - is forgotten.

The data paradox of modern manufacturing

We live in an era where we have more data than ever before, yet we use only 2-5% of it to make real decisions. It's like having a huge library in which we can only find a few books.

Data problems can be divided into three types: missing data, bad data, and isolated data. If we look more closely, we find that, for example, 60% of machines have no connectivity, 80% of process data remains isolated in separate systems, and 90% of quality checks are still done manually.

When intuition fails: the story of predictive maintenance

Imagine a situation: a manufacturing company decides to implement an AI-based predictive maintenance system. Expectations are high - the system is supposed to predict machine breakdowns before they occur and thus save downtime costs. After a few months of work and a significant investment, sobering news arrives. The model achieves only 15% prediction accuracy.

What happened? Machine data was collected only once per hour, the context of the manufacturing operations was missing, and the quality of the data collected was low. In this case, even the most advanced AI cannot produce a reliable predictive model. It's a classic case of "garbage in, garbage out" - garbage in, garbage out.

Digital maturity: a mirror of your readiness

Before embarking on any digitalisation project, it's crucial to understand where your business stands in terms of digital maturity. According to our survey of Czech companies, there are four basic levels:

• Level 0 (35% of companies).

• Level 1 (40% of companies): Basic digitalisation, isolated systems without connectivity

• Level 2 (20% of companies): Partial integration of systems, e.g. linking MES and ERP

• Level 3 (5% of companies): Fully interconnected infrastructure, real-time data

For companies at levels 0 and 1, the road to AI is still long and automation is a logical first step.

Automation: the underrated hero of digital transformation

When we talk about digitalisation, we often think straight about AI and skip the intermediate step that can deliver a faster and more certain return on investment - process automation. Automation is the ideal choice when:

• Your processes are based on clearly defined rules and logic ("when X occurs, do Y")

• You need to optimize the flow of data between systems that you currently connect manually

• Your teams are spending too much time on routine tasks that take them away from higher value-added work

The results of well-executed automation can be impressive: 60-70% reduction in data processing time, 90% reduction in human error, and significant speed-up in decision-making processes.

AI: when you have something to build on

Artificial intelligence is not a self-sustaining technology that solves all problems at once. Its real power only becomes apparent when you have a good foundation. AI delivers the most value in situations where:

• You need to uncover complex patterns in data that would be difficult for a human to capture

• You work with large volumes of unstructured data

• You want to predict future trends based on historical data with many variables.

But the key is that all of these scenarios require good quality, well-structured data in sufficient quantities. Without it, even the most advanced AI will be powerless.

From automation to AI step by step

Instead of jumping on the AI bandwagon in haste, it pays to proceed systematically and smartly:

1. Audit your data - find out what data you have, what's missing and what's of poor quality.

2. Build a basic data infrastructure - connect systems, ensure real-time data collection.

3. Automate routine processes - digitize paper processes, eliminate manual data transfers.

4. Create a single view of data - ensure all departments are working with the same data.

5. Only then consider AI - with a good foundation, implementation will be faster and more successful.

And what kind of timeframe to expect with this approach? Allow 2-4 weeks for data auditing, 4-8 weeks to build the underlying infrastructure, 6-12 weeks for iterative solution development, and 4-6 weeks for deployment to operations.

So the entire process will take 4-8 months, but the results will be lasting.

Success story: from Excel to predictive maintenance

One of our clients, a medium-sized manufacturing company, wanted to implement predictive machine maintenance using artificial intelligence. However, during the initial analysis, we found that their data was scattered in Excel files, and some machines didn't even have any sensors. Maintenance planning was based on the experience of the engineers.

Instead of implementing AI immediately, we started by automating data collection, installing basic sensors, and creating a centralized maintenance scheduling system. Only after six months, when they had enough good data, did we start working on a predictive model.

The result?

One year into the project, their system can predict potential failures with over 75% accuracy, downtime has been reduced by 30% and maintenance costs have dropped by 25%.

The key to success was a gradual build, not a big leap straight to AI.

Set realistic metrics for success

Whether you choose automation or AI, it's important to have clearly defined metrics for success. These should include both technical metrics (accuracy, reliability), business metrics (ROI, cost reduction, productivity gains) and operational aspects (employee adoption, speed of information acquisition).

Only with clear metrics will you know if your investment is delivering the expected value, and you can correct course early if results differ from expectations.

AI without data is like a car without fuel

Artificial intelligence has huge potential to transform your business, but without a good data foundation, it's like a luxury car without fuel - it looks great, but it won't get you anywhere.

For many Czech companies, the smarter move today is to focus on automation and building data infrastructure.

Once you have these basics in place, you can start thinking about more advanced AI-based solutions. This incremental approach may not be as sexy as jumping straight to AI, but it is much more effective in terms of ROI and long-term success.

Remember, digital transformation is not a sprint, but a marathon. Each step along the way should deliver tangible value and build a solid foundation for further progress. It certainly helps to have a trusted partner along the way who can assess your options with an experienced eye and guide you through the entire process, from data housekeeping to advanced AI solutions.