A procedure in several versions
The official document says one thing, the team's spreadsheet says another, and whoever joined yesterday follows the one they found first.
Memory for Operations
Kluco keeps operating procedures in a reviewed knowledge base, with an owner and a review deadline. From them, AI proposes processes with agent steps, rules and approvals by people, and every case leaves a record.
The problem
And requests travel by email until someone notices they have stopped.
The official document says one thing, the team's spreadsheet says another, and whoever joined yesterday follows the one they found first.
The request waits in someone's inbox, the deadline passes and nobody notices until the requesting area chases it.
Even with the procedure written down, automating it means describing every step again, and nobody knows where the process gets stuck.
Before and after
From the written procedure to the completed case.
| Without the company's memory | With Kluco |
|---|---|
| Without the company's memoryThe procedure exists in three versions | With KlucoThe current procedure is in the knowledge base, with an owner and the date of its last verification |
| Without the company's memoryDesigning the process starts from scratch | With KlucoYou paste the procedure and AI proposes the flow, saying what an agent can do and what stays with a person |
| Without the company's memoryThe approval is forgotten in someone's inbox | With KlucoThe approval has a deadline, and approved, rejected and deadline passed each have their own path |
| Without the company's memoryNobody knows where the request stopped | With KlucoEvery case has a timeline, and the indicators mark the step that is the bottleneck |
| Without the company's memoryRequests arrive with no owner or priority | With KlucoAn agent organises new requests with an owner, a deadline and a priority, and a person approves before saving |
What goes into memory
Documents come in from Google Drive, OneDrive, SharePoint, file upload, API or webhook, and every article keeps its version history. A procedure stored in the knowledge base can be pasted straight into process mapping.
Use cases
The most common use cases in Operations, with the part of the platform that does each one.
Paste the procedure the company already follows and AI proposes the flow, with agent steps and human approvals with a deadline.
AgentsInternal rules and procedures live in one knowledge base, with an owner, a review deadline and a version history.
BrainAn agent sorts new requests with an owner, a deadline and a priority, and a person approves before anything is saved.
AgentsBranches, units and teams query the same version of each rule. When it changes, everyone answers from the new one.
BrainWhen AI writes or updates an article, the change can wait for the approval of whoever owns the topic.
BrainPeople who just joined ask the knowledge base instead of interrupting the team, and learn the company's way of working through skills.
BrainAgents · AI agents
Paste the procedure and AI proposes the flow, saying what an agent can do and what should stay with a person. The canvas takes agent, rule, AI text, notify and wait steps, and every case keeps its own timeline.
Brain · the memory
The Brain turns operating procedures into articles with an owner, a review deadline and a version history. New people ask the knowledge base instead of interrupting the team.
LLM Management and Chat
The team uses the models the company allows in the Chat, which draws on the procedures in memory. In LLM Management, you track limits and cost.
Control
The company sets up each control below in the console.
Related
FAQ
No. Paste the procedure the company already follows and AI proposes the steps, saying what an agent can do and what stays with a person, with the source passage of each one. You review, adjust on the canvas and only then switch the process on.
Deadline passed has its own path on the canvas, just like approved and rejected. If it is not connected, rejecting and missing the deadline close the case. The decision shows up in the Approvals inbox and as a task for the approver.
Yes. You generate a key, and the ERP, a form or n8n opens a case by calling the process address. Send the order number from the other system along with it: the same call repeated does not open two cases.
The indicators show the last 90 days of each process: cases, cycle time, how many went through a person and how approvals ended. The per-step table shows the average time of each step and marks the bottleneck.
Next step
In the demo, we show AI mapping, approval with a deadline and the timeline of a case.