On today’s competitive market it is important to analyse and optimise product performance. The finite element method is a very common tool to do so when dealing with stress analysis. There are many different commercial finite element software available on the market. They range from complex ones needing special skills for proper use to more easy to use aimed at design engineers with little experience of the finite element method. The latest versions of Autodesk Inventor include a stress analysis tool that falls under the second category (easy to use). The analysis tool of Inventor is only able to perform a few types of analysis, and the user control is limited. The question is how well this tool works. What are its strengths and weaknesses? This thesis aimed to answer those questions. This was done by analysing a few “benchmarks”. Some of the analysed parts are just made up for this thesis, but most of them are provided by Water Jet Sweden AB, and are typical parts from a water jet cutting machine. These benchmarks has been thoroughly tested using both modal and stress analysis. The results from Inventor have then been compared with analysis results from I-deas, a well known commercial finite element software. The aim of these tests is not to see how close to reality Inventor gets, but if it gets a result close to a more advanced FE-tool. Depending on the benchmark different results is presented. For the larger parts and for the beams both stress and modal analysis are made, and the first six modes are presented. The deflection/deformation is shown for most benchmarks and equivalent/von Mises stress is shown for all. It is shown that the analysis tool in Inventor works fine when it comes to small and/or simple parts. Furthermore it is very easy to use compared with I-deas, although the solving time sometimes is rather long for big parts. A few problems do however exist. The main problem is that if the difference between the smallest and largest dimension in the part is big, the part cannot be meshed. This problem occurred mainly on benchmarks including long beams. The conclusion of this work is that as long as one can mesh the part fairly good results are attained. It may take a long time to solve, but one can trust the result.