The right start: why aimless experimentation fails
Many companies want to start an AI project but don’t know exactly where to begin. The temptation is great to simply try out ‘anything with AI’ – perhaps an internal experiment, the use of a chatbot or the integration of an existing AI tool. But this is exactly where many companies fail: Without having taken a moment to develop a clear strategy, the whole thing often ends in a lot of effort, little added value and the realisation that artificial intelligence doesn’t work as easily as expected.
Typical mistakes when starting an AI project
Before we talk about how companies can start an AI project properly, it’s worth looking at the most common pitfalls:
❌ No clear objective: many companies deploy AI without knowing exactly what problem they want to solve. This leads to inefficient projects that do not deliver any measurable benefits.
❌ Expecting too much from AI: Artificial intelligence can do a lot – but not everything. Anyone who thinks an AI tool can automatically optimise all business processes will quickly be disappointed.
❌ Wrong choice of tool: Many companies invest in AI software without first checking whether it is really the best solution for their problem. There are often leaner and more effective alternatives.
❌ Detail work too early: AI projects often fail because teams rush into technical or organisational details too early instead of first clarifying whether the project is worthwhile at all.
The 4 key steps to successfully launching an AI project
Starting an AI project often sounds complex – but with the right approach, you can quickly find out whether a project makes sense and how best to implement it. At Dentro, we rely on a structured process that helps companies to make the right decisions quickly and efficiently.
With this structured approach, companies avoid making expensive mistakes and reach a result much more quickly. Because ultimately, it’s not about doing ‘anything with AI’, but about creating real added business value.
Objective: What problem should be solved?
The most common mistake in AI projects is a lack of objectives or objectives that are too vague. Many companies start with the idea without clearly defining what result and added value they expect.
Good objectives are specific:
❌ Bad: ‘We want to use AI for customer service.’
✅ Good: ‘We want to use AI to increase the proportion of automated customer enquiries from 30% to 70% and thus save 20% on support costs.’ (Btw, more on AI in customer service here).
A clear objective ensures that:
- The focus is on real business added value
- The right KPIs are defined to measure success
- No resources are wasted on projects that do not deliver ROI
An AI project should always be thought of from the business side – and only then from the technology side. At Dentro, this is exactly where we start: Together with our customers, we identify the best use cases that really deliver economic benefits.
Testing high-level feasibility: What can AI really achieve?
Before companies release large budgets or get into technical details, a fundamental question should be answered:
Is the problem even solvable with AI – and if so, at what cost?
A pragmatic approach helps here:
- Initial tests with sample data: We use existing data to obtain an initial assessment. Are there recognisable patterns? Is the quality of the data sufficient? With the playgrounds and APIs of large language models (e.g. from OpenAI or Anthropic), this can usually be found out very quickly and easily.
- Identify uncertainties: There are critical points in almost every AI project – be it the availability of data, scalability or regulatory hurdles. A quick analysis helps to recognise and clearly identify these at an early stage.
- Estimating the technical effort: Do we need complex AI models or is simple automation enough? This is where the wheat is often separated from the chaff at an early stage.
Only once the basic feasibility has been confirmed is it worth taking the next step: the economic assessment.
Is it worth it? Evaluate ROI and potential
Not every AI project pays off – and this is precisely where many companies start too late. Before resources are committed, a clear ROI calculation should be carried out. This is about:
- Costs vs. benefits: How high are the development costs and what savings or increases in turnover are realistic?
- Scalability: An AI project is particularly worthwhile if it can be transferred to several business areas or processes.
- Compare alternatives: Is AI really the best solution or are there already simpler ways to solve the problem?
At Dentro, we help companies to create realistic ROI estimates in order to make informed decisions – instead of realising months later that the effort is not worth it.
Define the strategy for implementation
Once it is clear that an AI project is feasible and makes economic sense, the most important decision has to be made: How should it be implemented? There are three options here:
- Use an existing tool – If a suitable solution already exists on the market, this is often the quickest and most cost-effective option.
- Customise an existing tool – If a standard tool is not 100% suitable, it can sometimes be extended with additional features or APIs.
- Develop your own solution – If no suitable solution exists, a customised AI application can be built.
The right choice depends on factors such as cost, flexibility and time-to-market. In general, the selected technology should clearly follow the problem and not the other way round. At Dentro, we provide pragmatic advice and help to find the best option for the respective application.
Conclusion: The perfect start for an AI project
Many companies fail due to vague expectations and aimless experimentation. These mistakes can be avoided with the right approach.
1️⃣ Clarify the objective: What problem should the AI solve?
2️⃣ Carry out a high-level test: Does AI work with the existing data?
3️⃣ Evaluate ROI and feasibility: Is the project economically viable?
4️⃣ Define strategy: Use an existing tool, adapt it or develop a new one?
Companies that go through these steps have a much higher success rate. If you don’t want to spend months experimenting internally, you can quickly gain clarity and successfully implement AI projects with a partner like Dentro.
💡 Dentro tip: The best time to start an AI project is now. The right approach determines whether it will be successful. Feel free to contact us directly!