deepopinion

An enterprise-grade platform offering AI-powered automation solutions to streamline complex knowledge work and unstructured data processing.

DeepOpinion

For the end-to-end optimization of text and document heavy processes the entire solution is developed as a no code platform empowering ensures to build human level AI skills that they can integrate into existing Bots or use to create smart Bots from scratch. With they are no code solution customers can create cognitive skills in minutes and just with a few clicks. As there is a clear trend for no code Solutions we are excited to have them here in Munich today. They will give more insights and what they do and how it adds to optimization in our industry. 

Next on stage deep opinion [Music] Millions of times people in your organization sit in front of the same type of email or document to extract the same data over and over again in input management claims underwriting customer service hundreds of people. And let's admit nobody like to do the same mundane task over and over again. It's a waste of human intelligence time and money. Well do you have the idea of process.

Automation and AI are not new. Most companies still fail to automate text and document process. This is where deep opinion comes into play and AI, an intelligent automation platform with AI at its core. My name is Steve. I'm co-founder deep opinion. With me, Stefan, likewise co-founder. 

Talking with experienced process automation and AI teams, we hear about the same three challenges over and over again. RPA bots lack intelligence to process unstructured data. Creating, deploying, and maintaining highly accurate AI skills requires huge resources, skills, and time. Integrating intelligent automation into actual processes and systems requires huge IT investments. 

With deep opinion, we solve all of these problems with our automl engine. We guarantee best-in-class AI performance for text documents. Our no code AI Builder makes it easy to train AI for everyone super fast and deploy it with one click. Our drag and drop automation Builder with 100 plus integration makes it super easy to deploy and integrate intelligent automation. 

So let's check. Share the AI Agent Store with friends.

Out how we can free up your people from the mundane task and to enable them doing the value adding task. The tasks that people really are good at. Here we are uh in the core feature of our product. This is the automation Builder. 

What you see here is a claims process that starts in put management recognizing if it is a claim. Processing the attachment, processing the email itself and then in the next step if it is really a claim going further. Processing the attached documents, processing the information that is in the email, checking the background for uh if that claim already exists. If it exists we will update the claim. If it doesn't exist we will file a new claim and then we Mark the email as read and move it to a folder. 

Now the magic really happens in those notes that you see here with a small deep opinion icon. Those are the AI skills. Those mimic human thinking, human text understanding, human document understanding. And what the best thing about it is you can actually customize them to your needs, to your processes and this is what we. 

Share the AI Agent Store with friends.

Will be looking at now. So let's create a new skill. Let's call it email intent detection and let's go create it now. As your teams have been doing those type of tasks for the last months and years, you have a treasure trove of training data available in your organization that you can use to train AI models to automate those tasks. So let's do exactly that. 

So I'm uploading here an Excel sheet that was just exported from a system and it just shows some tickets. Now I've uploaded it and here I just have to tell the system that this is the text and this is what the system, oh well, other way around, just clicking too fast. This side is the text and the other side is the actual thing that the AI should learn. 

So out of reading that text, please predict these labels. Now I'm confirming and uploading the training data. This now takes a few seconds and this is where already some of the magic happens. Some of the pre-processing happens and the AI already tuned and fine tunes itself to process this data set. So after a couple more seconds you will see. 

Please share the AI Agent Store with friends.

That the upload is finished and the AI will have all the training data ready to be able to train and really automate that kind of step from the human from the human annotations that have been done over the last months. So while we are waiting, yeah now it's finished. Okay great, so the next step is something goes wrong. That's amazing, let's just start it over again. 

So the next step is after we have uploaded the training data, we will train a model. Let's call it email intent detection. This is what we want to do here. Let's call it version five. Now you select one of the base models that we have and they are proprietary and they were free trained for example for the insurance industry. So you can already base your AI model on best-in-class base model for the insurance industry. 

Now I will hit training and here is where the real magic happens. Normally this is writing a few hundred lines of code, you know burning through a lot of GPU hours, a lot of costs, um multiple months of optimizing and this is where our automl framework takes over. So I

Just presented the training data set. I click train and in the end it will give me back the best model it can train for this exact data set and we can basically guarantee um Best in Class Best in Class performance. 

So you already see the model training. It's at 14, it's at fifty percent and it will finish soon. You always hear the score going up. Now the score is at 93, so 93 accuracy percent of accuracy is actually human level. If you compare that to benchmarks you find in research, that's even a bit above human level that you can reach. 

So now training finished. Um soon it will be available to us and to look at the exact numbers and details of that model. Now it's ready, so we can see the scores. We can look into the details of that model of how well it's doing. 

Now let's look at what those types of models are capable of. Here I have such a model open and it's able to classify into various categories that you see on the right. I will just run it and see what the response of the model is. So this is a usual email, it's obviously anonymized. 

Share the AI Agent Store with friends.

But you see here that the model recognizes that it's a change request. Now let's look at the next step. So you cannot only um classify but you can also extract information. Let's try to run that here. I just have the same email over here and I will run this now. You see it extracted address data. It extracted different types of insurance that this person mentioned. Bank details, birth dates, it's probably not a birthday, it's probably some other date, email addresses and all the other things that might be interesting to you to be able to fill it to the systems and mimic what the humans are doing. 

Now let's look at the next step. We can not only process text and emails but we can also process documents. So now I will just upload a document and this is a subrogation form that um you know a car expert hand over to the insurance for the repair or for the check he did. I already pre-ran this just to save time here. So you see here the document was processed and really all the information were extracted on the right and they are ready to fill into. 

Share the AI Agent Store with friends.

The systems that you need to fill it into the core system where you manage your claims. So just think about what the number of mundane and repetitive tasks and that you could automate in your organization with this type of technology. Because it's exactly these tasks that cost you millions of money, um thousands of hours of time. 

Um basically if you want to learn more about intelligent automation, if you want to discuss use case with us, we would be very happy to welcome you at our booth next to Munigre. Thank you. 

Please share the AI Agent Store with your colleagues.

Note, that there may be slight differences in text and video. Watch video for the original content.

DeepOpinion Video