How much time do employees in your organization spend on work that could actually be automated? Typing data from one system to another, processing emails, preparing quotes, or updating customer information in the CRM. These are often small, repetitive tasks that, when added up, take a lot of time.
1. Traditional automation: fixed rules, predictable results
Traditional automation has been around for years. In this type of automation, software carries out predefined instructions. When a specific event occurs, a predetermined action is automatically triggered.
For example:
- Send a notification when inventory falls below a certain level.
- Automatically add a new customer to a mailing list.
- Forward an order to the appropriate department.
- Synchronize data from the CRM with another system.
This type of automation works according to the If This, Then That principle: if this happens, then do that.
As long as the conditions are clear and the process is predictable, this approach is highly efficient. The solution is often relatively simple to build, inexpensive to use, and easy to monitor.
There is, however, one limitation: traditional automation follows the rules you’ve set in advance. When a situation deviates from what you’ve anticipated, the process can get stuck or produce the wrong result.
2. AI Makes Automation More Flexible
Not all business processes follow fixed rules. Business information is often hidden in unstructured data, such as emails, documents, and meeting minutes. The same question can be phrased in different ways, and documents do not always contain the same information in the same place. This makes it difficult to process this data using traditional, rule-based automation.
That’s where AI can make a difference.
Whereas traditional automation follows predefined rules, AI can interpret information, recognize patterns, and use context to perform a task.
Here’s an example: A customer sends an email with a question about an order. A traditional workflow might forward a message based on specific keywords. But if the customer uses different wording, the system might not recognize the question.
An AI application can analyze the content of the email, assess the customer’s intent, and identify relevant data. That information can then be used in an automated workflow—for example, to update a CRM record or provide an employee with the correct information.
So AI does not necessarily replace automation. Often, AI makes it possible to automate even the less predictable steps of a process.
3. Practical Use Cases
The combination of automation and AI offers opportunities in many business processes.
Processing Emails
Analyze incoming emails, identify their purpose, and assign the appropriate action. An LLM scans the (unstructured) text and extracts data from it (such as a customer number, an invoice amount, a date, etc.). AI can first interpret and structure the content so that a workflow can handle further processing.
Populate CRM data
If information is missing from the CRM system, such as an address or VAT number, you can automatically send a notification when that field is left blank. With AI, you can track down the missing information in available sources, such as emails or documents, and generate a suggestion that is reviewed by an employee when necessary. This helps keep your data more complete and reliable.
Talk to your data
Allow employees to ask questions in plain language about data in, for example, documents and the CRM system. Instead of searching through various sources themselves, they can, for example, ask, “What do you know about customer X?” The quality of the answer naturally depends on the available data and its quality. Of course, it remains important to determine which steps can be automated and where human oversight (human in the loop) is still required.
4. Automation with or without AI?
AI isn’t automatically the best choice. The technology is capable of a lot, but it also introduces additional complexity. Consider the use of AI models, processing costs, integration with existing systems, and verification of results. A simple rule-based workflow is often the better solution when a process is predictable. AI becomes more interesting when interpretation or recognition is required. A good rule to follow: if you can describe the task simply using clear rules, start with traditional automation. If the system first needs to understand what’s written, what’s meant, or what’s shown in an image, then investigate whether AI offers added value.
5. Don't start with the technology; start with the problem
Successful automation doesn't start with the question of which tool to use. It starts with the question of where your organization is wasting time, making mistakes, or missing opportunities.
A pragmatic approach consists of four steps:
- Identify repetitive work. Which tasks recur frequently? Where do employees enter the same data multiple times? Which processes cause delays or frustration?
- Map out the process. Examine the steps involved and the systems used. Automating a process that is inherently inefficient does not automatically eliminate that inefficiency.
- Choose the simplest, most appropriate solution. Break the problem down into parts and choose the right solution for each part. Use AI only where necessary. Take costs, reliability, privacy, and security into account.
- Measure the results. Don’t just ask whether the automation works. Also measure how much time it saves, whether the number of errors decreases, and whether employees and customers actually benefit from it.
Start with a single, well-defined process. If the approach works, you can expand it further. It’s likely that the same patterns exist elsewhere in your organization. Perhaps it’s not just the sales team that wants information from your emails, but the marketing team as well.
Conclusion: Automate Wisely
Automation ensures that tasks are carried out more efficiently. AI makes it possible to process information that is less structured and requires more interpretation. The greatest added value is created when the two work together. AI interprets information, and a workflow carries out the next steps. This allows you to automate processes that previously had to be done largely by hand. But technology is only part of the solution. A clear process, reliable data, and explicit agreements regarding oversight and accountability remain essential.
At Understanding Data, we take a pragmatic approach. We start with a specific business problem, choose the technology that fits the need, and build processes step by step that save time and create value. The most important lesson: you don’t have to solve everything with AI. The best solution is the one that addresses the problem efficiently and reliably.