Press Release: April, 2023
Agile Automations has upgraded its product set thanks to Artificial Intelligence (AI).
The ambitious automations specialist now has access to OpenAI through Microsoft Azure.
The Manchester-based business received Microsoft approval to use the API last week, after submitting a use case for their order capture product.
This utilises AI to read order forms, either email attachments or pdfs, and return the information as structured data, in turn reducing the need for manual input.
After making big strides in the public sector in 2022, Agile Automations is already upgrading the process for key clients including KCS and ESPO.
Both organisations work with Agile as a trusted automations partner to speed up the procurement process and improve risk management.
Chief Technology Officer Daniel Holgate said: “It is great news that we have Azure access to OpenAI and it will make a big difference to our capability, starting with the order capture product.
“We have seen some hesitancy around using Large Language Models (LLM) like ChatGPT because it is a new and unknown entity. However, most businesses will be familiar with Microsoft Azure as part of their licensing agreement and are comfortable with the idea of submitting data via Microsoft.
“Now that we have access to OpenAI via Microsoft Azure not only does it improve our service, but it gives clients that extra reassurance because Microsoft is a secure provider.
“It is a way of utilising AI without venturing too far into the unknown.”
As one of the UK’s largest suppliers to schools and nurseries KCS deal with thousands of order forms.
Each order, or invoice, contains the same basic data such as a delivery address, reference number, email address and phone number.
Prior to partnering with Agile Automations, staff would manually open each form and re-key the data into an in-house ordering system.
Agile Automations improved this process using OCR to scan the documents for key data fields and decipher which part of the form had the corresponding information.
This worked best with set templates, but it meant any minor change to wording or layout would go into an exceptions queue and require manual intervention.
Now, using OpenAI through Microsoft, the data back is more structured and accurate. The way that the LLM has been trained means it can recognise similar terminology and will find the right information regardless of the layout.
“The initial product was a success as we saw up to 80% straight-through processing thanks to some clever coding” explains Dan.
“However, we want to give our clients access to the latest technology and be on the front foot with advances in AI.”
The advanced process has seen straight-through processing increase to 95% with some data returned 100% of the time.
It saves time in that clients no longer need to provide order templates and there are fewer classed as ‘exceptions’.
Once the data is returned via OpenAI the Agile Automations team validate the information using custom-coded automations.