FAQs: The Future of Transfer Pricing – Benchmarks with AI


After our successful interactive transfer pricing webinar : The future of Transfer Pricing – Benchmarks with AI, this blog post answers the questions asked by our attendees that we could not get to during the session.

What is the success rate of the AI Review Assistant performing the web review and the final result?

The review comments made by the AI Review Assistant has a 90% accuracy rate from our observations. The user would want to review a company further. This could be either accepting a company, checking a company for a website availability or simply because they have a hunch.

Are you able to add qualitative screens to the benchmarks?

Yes, the user is able to add ratios (e.g. intangible to total assets, number of FTEs etc) in the step before the AI Review Assistant. The AI Review Assistant will not check the companies that do not meet the criteria of the ratios as they are excluded from the search strategy.

How can the search process be made transparent to tax authorities so that they can assess the validity of the study result?

In the final report, the user can manually add a note that an AI feature was used to help perform the benchmark study.

Does the AI Review Assistant contain its own database of companies or does it need to connect to a existing database, for eg. Orbis?

The AI Review Assistant is not inherently connected to any database. It uses the information from the database that you have licensed (such as Orbis), after it is uploaded, to develop the suggestion comment. It is possible to use other databases besides Orbis.

Is the business description from the database, a combination of website/database or web data only? What is your priority and how do you ensure the website is the most updated source?

The AI Review Assistant is capable of designing a review comment based on solely the trade description (from the database), the company’s website or on a combination of both the trade description and company website. At the current stage, the AI Review Assistant is not “googling” nor verifying the website. This is a manual task that needs to be done by the TP specialist. Hence our emphasis that the user further qualify the AI Review Assistant’s output.

Would the tool also become applicable to royalty and interest benchmarks (where I have to review agreements)?

Royalty and interest benchmarks is not currently in our product roadmap. We will keep you updated on any developments in the coming months.

Can it be also used for loan or bond benchmarking?

The TPbenchmark tool is built for goods & services benchmarks (TNMM). It is not possible to perform a loan/royalty search in TPbenchmark.