AI records orders automatically

Tip to try out

If you want to process sensitive data, you should first selectively anonymize it. Personal data, confidential price information or customer details fall into this category. The cloud service pdfFiller blackens fields in the best CIA manner. This allows training data to be prepared to provide additional protection when processing in AI models. pdfFiller also offers numerous other functions such as PDF processing, eSign and workflows - and can be integrated into any company workflow via API.

The gold standard

Until now, powerful OCR platforms such as docparser in conjunction with an automation platform such as Workato were the gold standard in the automated processing of PDF documents or scans. docparser works according to the principle of marking each relevant area in the document and, if necessary, assigning further evaluation rules, the content of which is then imported into another system, e.g. an ERP system. When used correctly, the software is very powerful and efficient. The results are convincing across the board.

This process is ideal for similar documents. For example, if a company receives numerous orders that are similar in their layout and therefore have a predictable structure, we can achieve an evaluation rate of almost 100% in this way. Errors can be identified and ejected for further automated or manual correction or evaluation.

The disadvantage of these discrete OCR solutions is their dependence on a predictable document structure. For example, if a company has orders from numerous different customers who each use completely different PDF layouts, a separate evaluation logic would have to be configured for each PDF layout. This can be a challenge.

Nevertheless, discrete OCR is still our first choice when document diversity is low due to its predictability and transparent evaluation rules. In contrast to AI processes, each result can be traced step by step - and corrected precisely and permanently in the event of errors.

Discrete OCR shines through traceability and transparency.

The new gold standard

Since mid-2023, a new gold standard has begun to establish itself with the availability of GPT 3.5 Turbo and, more recently, GPT 4.0: Generative AI. In contrast to discrete OCR, we "trust" that the GenAI model (Foundation Model) has already "seen" countless content-like documents during its training and has been calibrated through reinforcement training, i.e. manual feedback from many testers, so that it can distinguish an article number from a supplier number and understand complex price table structures.

This is where we are today. It is now possible. Anyone who can use generative AI to solve problems in everyday business has a clear advantage. The industry is struggling to make AI usable. We know how to do this - and what needs to be done in detail.

GenAI conquers document processing...

Relieve customer service, helpdesk and back office

Where is the best place to start? Where work is not part of the core business and does not provide a competitive advantage over others - and where you might want to make better use of your employees. For the time being, we will focus on customer service, helpdesk and the usually overloaded back office.

PDF, Word, e-mail and scanned orders are entered manually in many places. This ties up resources and costs money. Thanks to generative AI, over 98% of this work can be automated. The process is as follows:

1. PDF etc. Orders are made machine-readable via an OCR service.
2. Prompts with relevant instructions run over this imported data with precise instructions on what is to be understood and how.
3. The generative AI model carries out its work and produces a result (completion).
4. Completions are received and converted into standardized order structures.
5. These orders are imported into the ERP or TMS via an interface; the order is now available there. Job done!

All this happens fully automatically. Orders can be received in any way. Whether via e-mail, file server or interface (API) - everything is possible.

Manual work in document capture is no longer necessary.

A further expansion stage is that the AI evaluates responses from the ERP or TMS system and then takes another action - as an "automated dialog" with itself, so to speak. If the ERP or TMS system has several APIs, these can be used, for example, for entire automated booking processes or contract creation. For example, after the incoming order has been transmitted, another API can be used to trigger an automated search for a free transport slot, a suitable proposal can then be automatically selected via an instruction given to the AI, the order can be automatically confirmed - and if there are any changes, either on the part of the customer or the transport service provider, further measures can be taken accordingly.

This is where artificial intelligence comes into its own, because it imitates human behavior. It can combine the decision factors perfectly, as it can also learn this. In contrast to discrete OCR, generative AI covers many more case constellations as it imitates the flexibility of the human brain. It seems to "think and act" according to the situation.

We expect this functionality in particular to be increasingly incorporated into all software. This will take time, as every piece of software has to be adapted in many places. However, the trick will be to use AI models or even platforms to automate end-to-end business transactions across multiple systems. Because no company can get by with software that can do everything. We are already working on this today.

Successful projects

How do you achieve a successful project? There is an effective "recipe" in automated document processing that we like to use as "best practice".

1. Record the quantity structure of the orders to be automated. How great is the economic potential? What is worthwhile? Which process should be used where?
2. Describe the functional target process as simply and completely as possible. What exactly should happen operationally and how? Where are the hurdles? How should work be done in the future when the process is automated? Details matter!
3. Acquire or call in technical expertise. Who takes responsibility and technical depth for success? Where do you go first? Which technologies and providers are suitable, which less so? Apart from all the promises, what do you need to look out for?
4. Try out points critical to success. How do we achieve the desired quality of results? What combination of procedures, tweaks and technologies should be used to achieve the desired success in a measurable way?
5. AI project budgeting, procurement, implementation. Do we understand what is being done? You can read how this works here.

There is one fly in the ointment: You will certainly encounter problems. This is completely normal. Remember that you are turning a big wheel. OpenAI, for example, has required tens of billions of dollars for it to perform as it does today. The same goes for other foundation model providers behind every AI platform. And you can now simply call them up via an API. But there is a lot going on behind that. That needs to be mastered!

Problems happen, problems have to be solved. This can range from an unfavourable recognition rate ("accuracy") to questions of result interpretation and standardization ("target system specifications"). There are solutions and methods for this. These must either be developed in-house or obtained externally.

Ultimately, AI offers an enormous opportunity to get rid of those tedious tasks that employees have previously resigned themselves to and stoically endured - at high opportunity costs: keyword "copy & paste". Since, thanks to AI, the computer no longer just does exactly what you tell it to do, but also creates something new in the process - that is generative AI at its core - the possibilities are not yet fully described and captured here. It all depends on the specific task and company situation. Take the initiative! You will be richly rewarded.

The universe is infinite - and the possibilities of AI

About Business Automatica GmbH:

Business Automatica reduces process costs by automating manual activities, increases the quality of data exchange in complex system architectures and connects on-premise systems with modern cloud and SaaS architectures. Applied artificial intelligence in the company is an integral part of this. Business Automatica also offers automation solutions from the cloud that are geared towards cyber security.