Lease agreement archive with AI: your hotline finds the right clause in the general terms

On a leasing company's hotline, many calls start the same way. The customer gives their agreement number and asks whether they can buy the car out early, what happens with the insurance, or whether the lease can be transferred to a new company. The answer is rarely in the agreement itself. The agreement refers to the general lease terms, and over the years the company has issued several editions of them. A customer who signed in 2021 has different terms from someone who signed last month, and an amendment may have changed the payment schedule or the leased asset as well. All of this sits in the lease agreement archive, but to make use of it, you have to know where to look.

So the agent has a customer on the line and several files open at once. They look for the right edition of the terms, scroll to the chapter on early termination and check whether an amendment changed anything. The customer waits, and picking the wrong edition means giving the wrong information.

This is one of the posts in our series "The AI-assisted archive in practice". For how the assistant connects to the archive and the limits it works within, see our overview of AI document search in CONTMAN.

A lease agreement archive: agreements, amendments and editions of the terms

In CONTMAN Archive, each type of document has its own category with fields. For a leasing hotline, this layout works well:

• "Lease agreements" with these fields: agreement number, the customer's tax number, signing date, leased asset and "Terms edition", with a value list of editions. The agreement number together with the tax number can form a unique key, so the same agreement never enters the archive twice.
• "Amendments" with an agreement number field marked as linking documents. Every amendment then appears next to its agreement.
• "General terms": each edition of the general terms as a separate document, with the fields "Terms edition" and "Valid from". The "Terms edition" field is the same definition as in the agreements, so an agreement and the edition of the terms it was signed under share the same value in that field.

Why each edition of the terms is a separate document

The archive keeps every version of a file, each one can be viewed, and the assistant can fetch the file of a chosen version. It reads text, however, only from the current version. If the 2021 edition were an older version of a single "General terms" document, reading the text would give the assistant the newest terms, even for an agreement they do not apply to. That is why it is good practice to save each edition as a separate document, with the date from which it applies.

For internal policies and procedures we advise the opposite, because only the wording in force matters there: a new edition is best added as the next version of the same document. We explain this in our post on employee onboarding with an AI assistant. An edition of the general terms, however, does not lose its relevance when a new one comes out: it still applies to the agreements signed under it.

Separate documents have one more advantage: the assistant can read two editions and set the same clause from each of them side by side. That comparison is prepared by the AI model, and the agent checks it in the documents. The archive itself does not show differences between versions.

Agreements signed on paper are usually scans: besides content processing, they need OCR, which has to be switched on first. Editions of the terms saved as PDFs with a text layer are read by content processing alone.

Questions an agent can ask the assistant

Once the archive has these links in place, the agent asks the AI assistant your company uses instead of searching through files:

• "Which general terms apply to agreement no. LS/2022/0458, and what do they say about early buyout?"
• "Find the clause on insuring the leased asset in the 2024 edition of the general terms."
• "Are there any amendments to agreement no. LS/2022/0458? Summarise what they changed."
• "Where in the terms that apply to this agreement does it say anything about transferring the agreement to another company?"
• "Which of this customer's agreements, by tax number, were signed under the 2021 edition of the terms?"

The agreement numbers and the editions of the terms are made up for this example.

For the first question, the assistant reads the agreement's description, including the "Terms edition" field. It then searches the "General terms" category for the document with the same value in that field, that is the right edition of the terms, and reads its text until it finds the passage on buyout. It finds the amendments among the documents linked by the agreement number.

The assistant helps find the clause but does not give a binding interpretation of it. The agent reads the wording from the document to the customer and passes disputed cases on just as before. The assistant's connection to the archive does not include the phone system or a CRM: the agent asks the question themselves, in the assistant your company uses.

Security on the hotline

Agents work with customer data, so the limits have to be clear:

• Every agent on their own account. The assistant always acts on behalf of the person who is signed in, so every agent connects it with their own account.
• Only the categories in the agent's role. Roles grant permissions for each category separately, and the assistant sees what the agent sees. Documents from the assessment of applications can sit in a category the hotline does not read.
• A trace in the history. The history of every agreement, amendment and edition of the terms shows when the assistant read it, with a note that the read went through an AI assistant.

What the assistant reads goes to the AI model your company uses, so the administrator decides which agents get the assistant. We describe roles and permissions in the archive in more detail in our post on secure document storage.

Coming soon: answers with a source card

In the next version, CONTMAN will include the built-in CONTMAN Assistant. Two of its features matter for a hotline: the answer takes shape before the agent's eyes, and it points to the documents it draws on with numbered references. A source card takes the agent to the document, so they can open the right edition of the terms straight away.

Getting started

1. Collect every edition of the general terms and save each one as a separate document in the "General terms" category, with the fields "Terms edition" and "Valid from".
2. Add the "Terms edition" field to the agreements category as the same field definition as in the "General terms" category, and an agreement number field marked as linking to the amendments. If you mark them as required, the rule applies from the moment you switch it on, and older agreements get filled in at their next edits.
3. Switch on file content processing in the categories for agreements, amendments and terms, and for scans, check that OCR is running in your installation.
4. Give your agents a role with the document read and the file download and preview permissions in these categories (without the second, the assistant cannot read the text of agreements and terms), and the permission to use the assistant. The assistant itself must first be switched on for the whole organisation by the installation operator.
5. Connect the assistant. Agents copy the server address from their account menu, on the screen for connecting an AI assistant.

How CONTMAN takes a lease from application to buyout is shown on our page on solutions for leasing. Book a demo, and we will show you how an agent finds the clause in the general terms that applies to a customer's agreement.