To speed up your workflows, our Hybrid Processing infrastructure leverages the best of human intelligence and deep learning. The platform takes on repetitive data labeling, data monitoring, and data extraction tasks and leaves only the challenging cases to your team.
With Hybrid Processing, tasks can be automated step-by-step. Only in special cases, where complex problems arise, the model passes the decision to humans. The feedback from these manual cases are stored and used to continuously improve the model - which ultimately reduces the human workload.
If you have a process for which accuracy is more important than full automation and where human-AI-interaction is possible, then Hybrid Processing is the optimal solution.
There is a broad range of applications for Hybrid Processing.
Here are examples including different data types that can be applied across industries.
We built a model that automatically scans nudity in media content and gets feedback from humans to keep learning and improving. These tedious tasks requires a large amount of time and classification may be subjective depending on the individual. The model within our Hybrid Processing infrastructure completes most of the work and passes only a few cases to employees, substantially reducing their workload and improving the quality of their checks.
Another avenue where Hybrid Processing comes in handy is categorising documents. Typically, insurance companies employ hundreds of workers to classify incoming documents and extract relevant information. With our infrastructure, most of the work is done automatically and only a small number of cases are flagged to be examined by employees. Similar to other cases, our model learns from the employee's decision.
Text processing can be another tedious workflow, however our Hybrid Processing infrastructure is able to identify and extract relevant texts in different forms such as documents or emails. Our trained models are able to convey any given dataset, present it in a comprehensible format with the appropriate texts and categorise them accordingly. Hence majority of the work is done and only a few outlier cases will be highlighted to be reviewed.
Save time and let your team focus on complex tasks.
Reduce labor costs and scale your operation.
Eliminate human errors and improve client confidence.
Our infrastructure seamlessly blends with your team and infrastructure.
Our model's performance improves continuously, no manual revision required.
We can provide light-weight interface for you or integrate into your current workflows.
The BMBF Grant has provided us the opportunity to boost the research and development of our Hybrid Processing infrastructure. It allows us to accelerate repetitive human workflows with an end-to-end human-AI workflow, which augments human intelligence.
IDnow verifies ID documents. In the past, they did this by live video authentication with human experts. Our solution checks the validity of holograms from video to prevent fraud. It contributes to IDnow’s new product line “AutoIdent” and decreased the costs per ID by 96% (from €2.5 to €0.1).
Previously, lawyers at DataGuard manually classify email texts into tickets to delegate them to the right person. In a bid to increase the efficiency of this manual workflow, we have supported this process with our Hybrid Processing infrastructure. Our model converts the internal structure of the dataset into a convenient format (e.g. class with typed attributes). As a result, only the 3 most relevant ticket types for them to pick from are presented - decreasing time needed to go through the list of more than 15 ticket types before picking out a suitable one.
ProSiebenSat 1, the largest German media company, needs to scan all media content for nudity. With humans, they do it by sampling and risk incurring huge fines. We built a model that automatically scans nudity in all media content and gets feedback from humans to keep learning. Our ready-to-deploy model achieved an automation rate of 30% already in the first model. Based on the success of the first project, we continue a co-development of our Hybrid Processing infrastructure.