Sponsored

Sponsored: Onboarding data to prepare for successful AI adoption

“Looking into the future, most, if not all, organisations will make use of AI products at some stage,” says Mandre Stander, Business Unit Executive and Solution Architect at CoCre8.

02 December 2024

Mandre Stander, Business Unit Executive and Solution Architect, CoCre8.

“Future planning starts now. Organisations need to focus on putting the correct systems in place to capture, extract and feed data to the AI engine.”

AI is changing business operations in many ways. It can handle the mundane tasks, freeing up human resources to do more productive work, or bring in additional revenue streams. Some organisations use AI to translate documents so that their users can better understand and implement their practices.

The way value is derived from AI to automate tasks or assist with accessing information faster will be unique to every organisation. “My first question to customers is: what do you want AI to do?” Stander says. “If the objectives are not well defined beforehand, it’s going to be a struggle to get it approved or to realise the value for the shareholders or stakeholders.” No matter how you visualise AI benefiting your organisation, there are certain important protocols that must be followed to set things up for success.

Organisations generate a lot of data and receive many documents from outside sources. For example, a supplier invoice is not something that’s generated by your system. How do you make that data part of your system so that when you deploy your AI, you can simply ask it to call up that invoice instead of going searching for it?

If it’s not accurate, your AI will give you inaccurate data in response to your questions. How accurate the data is is of fundamental importance now, even if you are only planning on using AI in the future.

So, for instance, even when scanning a document, what are you getting? Does the hardware help enhance the image? The higher the quality of the image that goes into the software, the better the extraction. If the lines are blurry, it’s going to be very difficult for the systems to extract.

A grudge purchase

“This is where solutions like Kodak Alaris come in, with high volume accurate scanners coupled with Intelligent Document Processing (IDP) software. Many organisations are scanning their inbound documents; however, very few are extracting information from the documents and using it in other systems,” Stander says.

Kodak Alaris offers hardware and software solutions that can assist organisations to accurately extract information from paper and other data formats through intelligent document processing. Kodak Capture Pro software works seamlessly with scanners from many different manufacturers, and is ideal for use in a multi-location scenario. Kodak InfoInput software is intuitive, flexible and scalable, and can work from any source, whether email, scanned paper, mobile upload, or other sources.

Some people perceive digitisation as a grudge purchase in the sense that they have to be able to create digital documents. “Look at it in a different way. You’re investing in this process because later on, it’s going to carry tremendous value for the business.”

AI is only as good as the data that goes into it. So companies with multiple vendors that extract information off a document and auto-populate their systems with that information need to be very sure that the information extracted is accurate. Stander emphasises that data includes all externally received information. “It’s not just the physical and digital documents that you receive. Sometimes, systems provide information or documents at the back end and you have to run it through a piece of software to capture it.”

Typically, data is referred to as structured or unstructured. Structured data is everything that can fit in an Excel format in a row and a column – for instance, information like name, age, ID number.

Unstructured data

However, data that can’t fit into a row and a column, like an image, would be categorised as unstructured data. “Think of a security company that uses CCTV footage. They need to be able to extract information from the images like street names, licence plates, pedestrians walking. World data is exploding and close to 80% of it is unstructured data,” Stander says.

For organisations, unstructured data poses the biggest challenge because it requires different infrastructure. The traditional infrastructure that runs your ERP system can’t run unstructured data. So from the front end, you need to receive this information, but it takes fundamental changes to the back end to get to that point.

“Often, when we at CoCre8 have conversations with our customers, we ask the question: what are your business plans for the next few years?” Stander says. “We try to look that far ahead to make sure that whatever we deploy and implement today will complement the processes they plan on using in the future. Structured data infrastructure is not the right solution if a customer is going to rely 80% on unstructured data in two or three years’ time.” The first step is ensuring that your platforms are all supportive of your AI endeavours. From a platform perspective, assess if the platform supports AI, and the software adds value, and in a few years’ time when AI is more developed, whether it can handle more use cases.

The platform and the systems play a critical role in how well it’s going to perform. AI can only give answers based on the information that it has and the context it’s receiving. Extracting accurate information is fundamental to having AI work the way it should. “Get your systems in place now, and set yourself up for success the first time you enter your data, so that when you decide to use AI, it’s as streamlined and accurate as it can be,” Stander advises.

mandre.stander@cocre8.com

steven.kramer@cocre8.com