An AI contract classification agent reads each document and assigns a contract type, such as NDA, MSA, SOW, order form, amendment or renewal.
An unlabelled document enters the intake.
The agent reads title, parties and operative clauses.
It proposes a type such as NDA, MSA or order form.
The document is tied to a parent agreement if one exists.
A person confirms uncertain classifications.
The type decides which review and fields apply next.
It adds business labels, such as spend category, department and risk tier, based on content and the linked vendor record.
It states why each label was chosen and sends doubtful cases to a person.
Repositories fill with documents of every kind, often saved under vague file names. Without consistent labels, nobody can filter, route or report on them.
'Final_v3' says nothing about type.
One PDF can hold an MSA, an order form and a schedule.
Procurement and legal use different terms for the same thing.
High-risk agreements get no extra care because nobody labelled them.
Spend by category is wrong if categories are wrong.
Read the structure and key phrases to decide whether it is an NDA, MSA, SOW, order form or other type.
Separate combined files into their component documents and classify each.
Add category, department and risk tier using the content and the linked vendor.
Record the text or feature that led to the label.
Send low-confidence cases to a reviewer and apply confirmed labels to the record.
Bring a small sample and watch Flo work through it.
| Aspect | By hand | With the agent |
|---|---|---|
| Sorting | Staff open each file to decide what it is. | The agent proposes a type from the content, not the file name. |
| Naming | Files carry names like scan0042. | Types and parties are assigned from text. |
| Routing | Wrong-type documents go to the wrong reviewer. | The type decides which review path is used. |
| Mixed bundles | A bundle holds several documents. | Bundles are split and each part classified. |
| Taxonomy | Types are used inconsistently across teams. | One type list is applied to every file. |
| If | Then | Who decides |
|---|---|---|
| the document matches a known type clearly | assign it and move to review | Contract administrator |
| the document mixes types | propose a split and ask for confirmation | Contract administrator |
| the type is not on the list | mark as other and suggest a new type | Contract administrator |
| the document looks like an amendment | link to its base agreement | Contract administrator |
| the file is not a contract, such as a quote | classify as supporting document | Procurement lead |
| classification confidence is low | hold for human labelling | Contract administrator |
A change document has a full-agreement title. The agent reads the operative text and flags the mismatch for an administrator.
A signed copy and a draft look alike. The agent notes the signature status and the administrator confirms which governs.
A single file contains both. The agent proposes a split and an administrator approves.
Terms do not fit the type list. The agent assigns the nearest and flags it for a new category.
Talk it through with the team. We will map it to your steps and show where a person stays in control.
Keep it short and distinct.
Compare with an administrator's labels.
Let the type pick the review path.
Add types that keep appearing as other.
This shows how often people disagree with the agent's labels. A rising rate suggests unclear label definitions or new document types.
Quick answers to what people ask most about contract classification.
Book a demoIt is the process of tagging each contract with its type, business category and risk tier so that it can be found, routed and reported on.
Manual tagging is slow and inconsistent. Automatic labels with stated reasons give a uniform basis for routing and reporting.
Yes. It can split a bundle and classify each part separately, with a person confirming where the boundary is unclear.
Your team does. The agent applies the definitions you set and reports where documents do not fit them.
Contract metadata extraction is the capture of structured facts from an agreement, such as parties, effective date, term, renewal type, notice period, value and governing law.
→Contract hierarchy mapping is the linking of related documents into a parent and child structure, such as an MSA with its order forms, SOWs, schedules and amendments.
→Contract summarization is the condensing of an agreement into a short statement of what it covers, what it costs, how long it lasts and what the main obligations and risks are.
→Contract signature routing is the process of deciding who must sign an agreement, in what order, and sending it to them with the right pre-checks complete.
→Bring a small sample of real items. We will show what Flo handles on its own and where it hands over to your team.
