Absolutely brilliant! I've been discussing this with colleagues about how the rate is much better and how you can use COE to find the big jobs running and put doc intelligence in place there. Love the precision of delivery.
Great Video. I really learnt a lot. I have a challenge - 2 challenges actually. 1. How do I use the Azure document intelligence for a pdf with up to 30 pages with data on a page which goes beyond that page. 2. This is related to power automate - Can You explain or make a video to use gemini ai API key to return information to Copilot Studio rather than using ChatGPT OpenAI for this. I tried to create a custom connector but it didn't work
1. Are you trying to identify a particularly formatted page from 1 of 30 to the. Put through a model? 2. What are you trying to do with the Gemini api out of curiosity? Are you looking to return its result direct to Copilot Studio? Are you looking to use a specific topic to trigger it?
Hello Damien, Very nice video thanks for sharing. I have a question how's your experience with hand written documents or invoices. Earlier this year I was working on project using AI Builder and Flow to process hand written documents, I found AI Builder unable to process hand written documents 100% accurately. Ex: if in document text is Bagel, AI Builder is predicting text as 'Bogel'. Ex: if in document text is deport, AI Builder is predicting text as 'depont'. And AI Builder is not providing any option to modify this prediction text. Let me know if came across such scenario and how you overcame such issue. Thanks.
It was a sunny weekend and I had lots of gardening 😎 but managed to get this recorded and edited late last night. Very much enjoyed exploring and sharing this one.
Great video, thanks for sharing. I just have one question : Is there a reason why you use Power Automate instead of Logic Apps ? Is it because of cost issues ?
Hi, thanks for the great video! We are looking to improve our invoice processing using Azure AI Document Intelligence. Do you know if it is possible to build a Power Automate flow that lets u manually check the output data of the AI model and either approve or change some wrong values before the data gets exported into our internal processes? I am looking for such a soolution, but I am unable to find it as of yet. Thanks! :)
To elaborate a little on that, ideally we would want to set a threshold. For example, if confidence is 0.9 or higher we want to automatically accept the output. But if its lower, we want to manually check and maybe change the output.
Yes, this would be achievable, the output contains the confidence score and so you could factor that into either an app or an automation. For example, if a score is below 0.9 you could send an email or adaptive card, request a human in the loop verifies / updates and then save the changes. I don't have a demo yet though. Interesting use case.
Thanks for the quick reply. Do you have any recommendations on which tooling to use for such an automation? I can imagine power automate might be a suitable tool. And if you ever make a demo about this, I will be sure to check it out :)!
You could run both methods, in parallel, writing to two Dataverse tables recording the associated Confidence scores in those Dataverse tables. Then, later, run a Flow that compares those two Dataverse tables generating a 3rd one with the data that exceeds a value of 90% for the Confidence Score.
Great demo. Love that pricing comparison for AI builder vs. Azure PAYG services. As always with bigger power comes more responsibility/JSON.
I love a bit of JSON though 🤣 500 free credits a month though - worth it?
Thanks for sharing this method. Useful and cost effective.
Absolutely brilliant! I've been discussing this with colleagues about how the rate is much better and how you can use COE to find the big jobs running and put doc intelligence in place there. Love the precision of delivery.
Very much appreciate your comment - makes this worthwhile, thanks.
Great Video. I really learnt a lot.
I have a challenge - 2 challenges actually.
1. How do I use the Azure document intelligence for a pdf with up to 30 pages with data on a page which goes beyond that page.
2. This is related to power automate - Can You explain or make a video to use gemini ai API key to return information to Copilot Studio rather than using ChatGPT OpenAI for this. I tried to create a custom connector but it didn't work
1. Are you trying to identify a particularly formatted page from 1 of 30 to the. Put through a model? 2. What are you trying to do with the Gemini api out of curiosity? Are you looking to return its result direct to Copilot Studio? Are you looking to use a specific topic to trigger it?
great video as always, nicely technically explained. cheers!
Thank you! Cheers!
Hello Damien,
Very nice video thanks for sharing. I have a question how's your experience with hand written documents or invoices. Earlier this year I was working on project using AI Builder and Flow to process hand written documents, I found AI Builder unable to process hand written documents 100% accurately.
Ex: if in document text is Bagel, AI Builder is predicting text as 'Bogel'.
Ex: if in document text is deport, AI Builder is predicting text as 'depont'.
And AI Builder is not providing any option to modify this prediction text.
Let me know if came across such scenario and how you overcame such issue.
Thanks.
Interesting. What was the confidence score? I’m not sure what your options are but keen to hear if you find a solution.
Great 😃
Thanks for Sharing! Do you have tutorial on Azure Vault?
Not yet… 🤔
Nice Video
It was a sunny weekend and I had lots of gardening 😎 but managed to get this recorded and edited late last night. Very much enjoyed exploring and sharing this one.
Great video, thanks for sharing. I just have one question : Is there a reason why you use Power Automate instead of Logic Apps ? Is it because of cost issues ?
I primarily focus on Power Automate but 💯 a lot of my videos, including efficiency can be replicated on Logic Apps.
Hi, thanks for the great video! We are looking to improve our invoice processing using Azure AI Document Intelligence. Do you know if it is possible to build a Power Automate flow that lets u manually check the output data of the AI model and either approve or change some wrong values before the data gets exported into our internal processes? I am looking for such a soolution, but I am unable to find it as of yet.
Thanks! :)
To elaborate a little on that, ideally we would want to set a threshold. For example, if confidence is 0.9 or higher we want to automatically accept the output. But if its lower, we want to manually check and maybe change the output.
Yes, this would be achievable, the output contains the confidence score and so you could factor that into either an app or an automation. For example, if a score is below 0.9 you could send an email or adaptive card, request a human in the loop verifies / updates and then save the changes. I don't have a demo yet though. Interesting use case.
Thanks for the quick reply. Do you have any recommendations on which tooling to use for such an automation? I can imagine power automate might be a suitable tool. And if you ever make a demo about this, I will be sure to check it out :)!
Power Automate could certainly support this 👍 I’ll add it to me list but can’t make any promises
You could run both methods, in parallel, writing to two Dataverse tables recording the associated Confidence scores in those Dataverse tables. Then, later, run a Flow that compares those two Dataverse tables generating a 3rd one with the data that exceeds a value of 90% for the Confidence Score.