Give your AI agents current, cited knowledge
Your agents read one knowledgebase that flags replaced policies. They apply the rule in force that day, and stop when they can't back a step.
- 1. Supplier matchedHarlow Civil Pty Ltd is the party to contract 2026-114
- 2. Over the contract rate by $500Contract 2026-114 · schedule B · page 22
- 3. Rate as at 12 May 2026, the invoice dateA later schedule exists and was not used
- 4. Needs the finance managerDelegations policy · section 3.2 · page 5
- 5. No purchase order on fileStopped, and sent to a person
Stop agents reading stale policy
When each agent indexes its own folder, a policy change reaches one agent and not the others. Old versions sit beside new ones, and search treats them as equally valid.
An agent then acts on the old rule with a citation that looks fine. With Certant every agent reads the same knowledgebase, so a correction or a replaced document reaches all of them at once.
- Travel policy v3 · section 4.2Claims must be lodged within 14 days of travel
- Expense procedure · section 2Claims must be lodged within 30 days of travel
How agents use the knowledgebase
Agents read the knowledgebase at each step, so each decision cites the document it relied on.
- Connect your documents once Policies, contracts and records go into one knowledgebase.
- Agents read it at each step Each decision cites the document and the page it relied on.
- Corrections reach every agent A fix made once is kept in a ledger and applied everywhere.
- People approve what matters An agent stops at an approval step and waits for a person.
Checks your agents inherit
These come from the knowledgebase, so every agent gets them without extra work.
| Feature | What it does |
|---|---|
| Replaced versions flagged | Replaced documents are flagged, so an agent can tell an old version from the current one. |
| Rules as at a date | The agent uses the rate or rule in force on the transaction date, and records which one. |
| Conflicts go to an owner | When two documents disagree, a named person decides, and the decision is kept. |
| Stops when unsure | If the knowledgebase can't support a step, the agent stops and asks a person. |
| Full execution logs | Every run, step and prompt is recorded, and runs can be compared and replayed. |
Build agents in the Agent Builder or ask the AI FDE. Your own agents can also read the same knowledgebase through the MCP server under a scoped key.
What teams deploying agents ask
How do multiple AI agents share one knowledge repository?
Each agent reads the same Certant knowledgebase, with its own instructions and permissions. A correction or a replaced document reaches every agent at once.
How does it know which version of a policy is current?
Certant flags documents that have been replaced or that contradict each other, and tells the reader when an older version says something different.
What stops an agent acting on a wrong answer?
Each step cites its source, the agent stops when the knowledgebase can't support a step, and actions that matter wait for a person to approve them.
Can our own agents use it?
Yes. Agents outside Certant can use the knowledgebase through the MCP server under a scoped key.
Bring a job you'd like automated
Book a call and point an agent at it, or start free and build one on your own documents.
Contract 2026-114 scored 4 of 5 for risk. It is with the risk team, and nothing has been sent to the vendor.