How AI-assisted text interpretation simplifies your contract management

Text interpretation software (also called OCR) gets better and smarter, when artificial intelligence (AI) is used to interpret the text in digital contracts.
AI-assisted text interpretation for your digital contracts can be used for all types of contracts

Contract managers ensure that all the company's contracts and associated documents are kept in good order and that contract content is available so that the business can benefit from it.

An uncomplicated statement that can be taken for granted. At the same time, the words "good order" and "available" hide a lot of complexity. An example when using contract binders shows some of the problems. With contract binders, good order follows partly from the fact that contracts are actually inserted in the folder, and that they are classified according to a register. The HR department may sort employment contracts based on last name, while the CFO sorts the supplier contracts by organization number. Notes and reminders are attached to the respective agreement. This can be seen as "marking" or keeping track of information you deem relevant, ie extra important, for some reason. This is a good order based on the type of agreement. Unfortunately, it looks different in different parts of a company.

Accessibility means that the business can easily and without delay obtain relevant contract information, at the same time as unauthorized persons are denied access. From, for example, the supplier agreement at the CFO's office.

An investment in good order for the whole company pays off in better accessibility. This applies to both the contract binder and the digital contract solutions. Here we describe how Artificial Intelligence (AI) can help you maintain good order and improve accessibility.

The benefits of good order and accessibility

First, an example from reality that puts order and accessibility in a context.

A nationwide chain of franchise stores, as well as a group mother who has a shareholding in the stores. Each store purchases security services for cash handling and store monitoring. This type of contract often runs annually with automatic renewal unless the contract is terminated, for example 3 months before renewal.

Good order in this context requires that 3 dates be identified and recorded.

• Date of contract renewal, with possible cost increase according to the clause for index-linked prices.
• Date of contract renewal minus 3 months, which is the last day to terminate the contract.
• And a third date that may be 5 months before the contract is renewed and which gives the store manager time to review the current contract.
In this case, there were no such routines and most of the stores had the agreements renewed without review. At Group level, a significant cost saving was identified with centrally negotiated framework agreements, which was also implemented. But ... the effect was absent. Availability broke. The store managers do not have easy access to the framework agreement in the paper folder at the Group Management's office.

Find relevant information in your contracts

How do you find relevant information in contracts?

Let's take it from the beginning with the goal of simultaneously clarifying the benefits of AI-assisted text interpretation.

A typical contract contains many pages of text and there are no rules for where, in the document, the information you are looking for sits. Below are some examples of how information can be found:

• Manual retrieval. In addition to of course talking to the person who signed the contract, it is not uncommon for you as the contract manager to simply read through the document on the computer screen and from there try to find the information that is being sought. With this way of solving the task, there is a risk that information will be missed and that you simply do not have the time to go through the contract properly.
• Text search. If the file is in PDF format - or in some cases in image format - it may be possible in a PDF program to search for specific text. For example, Adobe has this capability in some of its programs (which, however, may have a license fee). With the help of text search, you can search for words or phrases, such as "Index increase" or "Valid until". With text search, all agreement text can be searched through, which is an advantage and increases the probability of a search hit. But you will only find text that matches your search phrase!
Fuzzy / smart search. Some IT systems allow you to use "search patterns" to find information. Examples are automatically identifying email addresses ( or web addresses (https: // It can also be about ignoring the difference between uppercase / lowercase letters or being able to handle simple misspellings of words. This provides a further improved search hit list, but since there are no rules for how text is written in contract text, there is still a risk of missing relevant information.
• AI-assisted text interpretation. The software generates suggestions for key information. The key information found is presented as lists where you as the contract manager can jump directly to the sections in the document where the key information is located. The interpretation made by the AI program works on text that is written in "natural language", ie a person has written the text without any special rules to relate to. An example is that the date does not need to be specified as "2021-12-24" to be identified as a date by a computer. It may just as well say "December 24, 2021". Both variants represent a specific date and the AI-assisted text interpretation "understands" this.

AI-assisted text interpretation


The video shows how key information is extracted from contracts through the use of AI-assisted text interpretation in Avtalsbanken.

The rapid development of AI-based text interpretation in recent years has led to a breakthrough in computer-assisted help to automatically identify and classify information in documents.

Today, technology can help you quickly identify and extract keywords in large masses of text. The text is analyzed with AI, which identifies dates, personal names, organization numbers, places, numbers, telephone numbers, etc.

• AI software can help identify and interpret information in different types of documents. The document can be scanned as an image, saved in PDF format or other text-based format.
• The translation is handled by the AI program - not by you. The technology handles over 100 different languages. You thus get the same functionality with text in Russian, English, Swedish or Portuguese. Automagic.
• Scanned images have varying quality. Sometimes it can be very difficult to even interpret what is in the document. The technology here helps to improve the quality so that the information can be identified without you as the contract manager having to re-scan the document.
• Technology picks out and groups the words based on category. You can quickly identify and select important information. In the example with the stores, AI identifies "March 24, 2021", "3 months" and "annually" as words in the category time. You navigate to the place of the words in the contract to verify the meaning and then mark the start date, renewal date, reminder date. Technology makes it easier and faster to do the right thing from the start. And thus to get a good order.
The technology is good but not perfect. There is still a need for a person to go through the results. Regardless of this, many working hours are saved by letting the computers do the hard work and we humans control the result so that it is reasonable and useful.

Mark relevant contract information

En mängd "kom-ihåg-lappar" uppsatta och samlade på en plats kan användas för att hålla reda på avtalsinformation man vill hålla under bevakning
A collection of yellow "remember-notes" is a way to keep track of contract information

"Marking out" relevant contract information is part of creating good order. Once you have identified information that you for some reason want to keep a special eye on, how is the "marking" itself done?

• Yellow notes in binders! This may seem simple but is more common than you think and can work for small businesses that can go through all their agreements regularly. But of course this takes a lot of working time and it is easy to miss something.
• Spreadsheets. By creating a framework with rows and columns, you keep track of the information. The method is based on manual copying of information from the contract to the spreadsheet and works - but there are problems with information being missed or simply copying errors.
• If you use a business system or a contract management system, it is important to have support so that the information can be added as “extra data” linked to the document in question. However, it is still a critical part of the process to copy the correct information from contract documents.
• If you are already using a system with AI-assisted text interpretation, this must also help mark important data that you have identified. AI-assisted text interpretation has no value if it can not be used in a way that is beneficial to the business.
• Regardless of whether you have the contracts in binders or in a cloud service, it is worth considering how you want to be able to search for contracts and what monitoring you need. Is there e.g. a need to filter out contracts based on internal cost accounts? Then it may be worth the extra work to mark up the contract with a cost account.
Make sure that the contract solution you use supports both marking the contracts with any data and that you can search / filter on the same data. As in the case of the stores above, it is often the data in the contract itself that needs to be pointed out. With AI-assisted text interpretation, markup becomes more complete while going faster.

Set up contract monitoring

It should also be possible to monitor contract information that has been marked. Information in contracts and documents does no good before it has been communicated to a recipient who can act. This is part of improving accessibility.

It is common - at least in slightly larger companies - for different people or departments to be responsible for different parts of the business. The larger a company is, the higher the specialization. An extension of a supplier contract may have to be carried out by a purchasing manager, while an extension of an internship contract is handled by a person in the human resources department or the immediate manager.

Information from contract monitoring may also need to be received and processed in IT systems.

• Ask questions and map which people and systems are to be reached and when they need the information
• New contracts are added on an ongoing basis. How often must marked contract information be checked to determine whether an "alarm" should be sent to the recipient? Once a month, once a week or maybe daily?
• Whether the tagged information is stored in a spreadsheet, contract management system or business system; how can you, as the contract manager, set up automated monitoring and rules for who is to be notified?
• When the recipient has been notified by a contract monitor rule that an action must be taken for a specific document, how does he get access to the document itself and related information? How does the receiver show that "the alarm has been handled", ie the alarm is acknowledged? If no acknowledgement is made, the alarm will go off again until someone acknowledges it.
If you have already started looking at systems that support AI-assisted text interpretation, you should also check if alarms and alarms can be handled in the same system.

Curious to try?

AI-assisted text interpretation is a technology that is here to stay.

Already today, results are delivered with impressive accuracy. As it improves contract management so much, it is a well-invested time to start looking for solutions that work right now. There is also much to indicate that AI interpretation will be even better in the coming years.

Contract managers get a completely new tool in their toolbox that makes it easier to maintain good order and increase the availability of contract information. It will be easier to identify relevant contract information and set up contract monitors, which helps companies to meet the conditions contained in contracts and documents.

Would you like to contact us for more information? Email questions to or fill in your details in the contact form and we will get in touch.Show me more
Avtalsbanken has solutions for AI-assisted text interpretation in place and we have integrated the technology as a support for the user. There are no automatic decisions based on AI. At the same time, the development is very fast and the future will offer new solutions. We do not know which ones, but we make sure you have access to them.
In Avtalsbanken's web service, AI-assisted text interpretation is performed for each uploaded document. Take a look at this video for an introduction.Show me more