Together we can structure your unstructured data using natural language processing and a clever online platform. Using our knowledge from past projects, we've distilled seven simple steps for a successful project.
Before integrating our platform with your business we start with an intake period where we gather all the necessary information. This is done through a series of workshops with all the relevant stakeholders and users. We first visualise the business processes that will be aided with Klassif.ai. Furthermore, we take a look at all the data that is currently available. Finally, we try to answer the following questions:
The goal of these session is to visualise the business processes and list all the required steps that are required to integrate Klassif.ai. If all the necessary information is clear for us, we start with the next step: Integration.
In order to start the next phase, we will need to integrate our platform with your systems. Luckily, Klassif.ai comes with many built-in integrations such as:
New integrations can easily be added to the platform!
The platform is fully integrated with your systems and data gathering and labeling can begin. Moreover, Klassif.ai can replace your business processes through the platform.
The Klassif.ai platform is built around an intuitive, highly customisable user interface. In this phase we will work with your users to customise the platform to their needs. We will analyse their way of working in order to keep the impact on their work as low as possible. This means structuring the elements on the platform, changing the branding to your company and adding the necessary custom fields.
In this phase, we make use of the standard SCRUM practices, including two-week sprints, daily standups and two-weekly sprint reviews. After every sprint review, you can review the changes and provide feedback. Furthermore, we can organise a user testing session to evaluate the user interface.
After this phase, the platform will be ready for use by your business experts. If any changes are needed afterwards this is of course still possible!
Using Klassif.ai your domain experts will keep processing the data manually and simultaneously they will generate valuable labeled data. This is done using our innovative labeling platform which allows users to manually structure data and send this data to your systems. In the meantime this data is also stored for the next step.
The labeling process can be aided by using some of the built-in models for dates, addresses and more. Once, enough data is gathered we start the model building phase. The first model will already greatly reduce the labeling time and the customer can expect this time to decrease with each new iteration of the model.
The outcome of the labeling phase is a labeled dataset which will be used for the model training phase. The customer fully owns this dataset and is free to use this data for other projects as well.
In the training phase, we use our extensive knowledge in natural language processing to build the perfect model for your documents. Typically the model building phase consists of the following steps:
In this phase, your business experts will be working together with our AI engineers to make the best possible model.
After the training phase, the platform is ready to automate your processes and structure your data. This means we are ready to deploy the platform for primetime use.
After the model training phase, the platform can be deployed for automating your workflows. Klassif.ai can run on our cloud platform as a software-as-a-service (SaaS). Besides our own cloud, Klassif.ai can also be deployed on all mayor cloud providers. We can even deploy it on your own on-premise architecture!
The platform is ready to be used by your business experts.
Now that your business experts are using the platform daily, they are still gathering valuable data. Periodically the customer can choose to retrain and evaluate the new model. If the new model outperforms the old one, the customer can choose when to deploy is. It is also possible to manually trigger a retrain whenever the customers wants.
Using such an approach, the platform is always performing at the highest accuracy.