How long an Intelligent Document Processing (IDP) project takes
The road to a finished implementation of a solution for AI-based document processing is not as long as many assume. In general, an Intelligent Document Processing solution can be deployed productively in the company within a few weeks and thus save time and money very quickly. We show how this works.
Four steps to the goal: All of the following time specifications refer to the pure duration and not to the required person days. Of course, the figures are guidelines - the exact times required always depend on the complexity of the individual projects.
1. Individual automation concept
after approx. 3 days to 2 weeks
At the start of the project, elementary questions about the design of the individual process solution are first clarified: which documents are to be processed, what information is to be extracted, and what is to be done with this information. Based on the results, an individual automation concept is created for this process.
For less complex projects, the concept can already be created during the preliminary discussions. For more complex extraction or classification goals, multi-layered process flows or special requirements, the conception phase takes correspondingly longer.
2. Trained AI models
after approx. 2 to 4 weeks
The core of the solution are the AI model(s). They take care of the classification and the extraction of the specific information. The models are trained for this purpose - the AI thus learns to recognize and output exactly the desired information and, if necessary, even to assess it. This usually requires training data - i.e., sample documents that can be used to train the AI models. In this step, the degree of automation achieved later can already be measured using representative examples.
Training a new AI model usually takes two to four weeks, depending on the complexity of the extraction task and the document formats.
For standard documents, such as invoices, existing AI models can be used - at most, a slight adaptation (a few days) is then required.
Thus, the software reliably recognizes the desired information (in general, recognition rates are between > 96% and > 99%).
3. Integrated process solution
after approx. 4 to 8 weeks
Once the AI models have been trained, they can be integrated into the software. This creates a holistic process solution. The exact scope of functions is individual - usually includes functionalities for document input, workflows, review and export of documents.
In many cases, the software can be configured based on existing modules and integration options. This usually takes one to two days. For individual workflows or integrations, the effort is higher. If the solution is to be operated on-premise, additional time must be planned for installation.
Image: Review interface for AI document processing of contracts.
4. Operation and optimization
from approx. week 4 to 8
Of course, further optimization is carried out after commissioning. The finished solution can either be used as a cloud service or operated on-premise. In both cases, corrections that users make to the extraction are used for further optimization of the AI models.
As a specialist in Intelligent Document Processing in practical use, kinisto will be happy to advise you on the subject and on your specific case!