Support replies grounded in your policies
Train product and refund documentation so an employee can find relevant passages while drafting a response.
Make selected files, text, websites, and connected content searchable for your AI employees. Eligible requests retrieve relevant passages from authorized trained sources. If training access is unavailable or no usable trained content exists for the request, document retrieval is skipped. Conversation memory and reading files through tools are separate capabilities.
Give your AI employees reference material for real work: product documentation, policies, brand guidance, and project notes. Choose files, pasted text, a website, or content from a connected app. After training completes, employees can find passages related to a question and use them when composing an answer. You do not need to paste the whole document into each conversation.
Sistava extracts readable text and indexes smaller passages by meaning. A question about returning an order can find a passage about refunds even when the wording differs. Selected passages arrive with source details, helping the employee work from your material. Training creates a searchable reference library; it does not rewrite the underlying AI model.
Searchable Knowledge is optional and separate from copying files to Drive. The system checks feature access and usable trained content before document retrieval. If the feature is unavailable, or no eligible trained content is available to the request, it skips that search. Having a paid plan alone does not trigger an empty document search.
Keep references current by retraining changed sources or scheduling a supported refresh. A trained copy is not a live connection to every change. For current order status, availability, or other changing records, an employee can use an available connected tool with the required access. Review source material when a decision depends on an exact detail.
Select the files or connected content needed for the work. When a connector offers a list, choose the pages, folders, channels, or records to include. If it cannot offer a useful list, an import of all accessible content requires explicit confirmation.
The employee receives selected passages related to the request rather than every document in full. Finding useful evidence helps answer a question, but retrieval and generation can still miss details or make mistakes. Ask for the source when precision matters.
Training is reusable across employees when the current requester has access. A private training source does not become public merely because an employee can use it. Source selection decides what enters training; access controls decide who can retrieve it.
A support draft can use a trained policy, a remembered preference for tone, and a current order record from an available tool. Conversation memory and live tool access are separate from document training, with their own controls and costs.
Choose material, complete training, and retrieve relevant passages when access allows.
Choose a source and confirm training. Sistava extracts readable text, splits it into passages, and indexes them. Check the completed training result to see which material is ready to use.
Before document search, Sistava checks feature access and usable trained content. Search stays within the sources authorized for the request. Matching passages and source details become context for the employee, alongside the current conversation and other available information.
Train product and refund documentation so an employee can find relevant passages while drafting a response.
Make selected policies searchable, then ask the employee to identify the material behind an important answer.
Train approved product sheets and positioning documents for employees working with authorized sources.
Retrain changed sources or schedule a supported refresh so later requests can use the revised content.
| Before | After |
|---|---|
| Paste the same reference into each conversation. | Train selected material once for reuse in authorized requests. |
| Search documents using only the wording you remember. | Find related passages by meaning, including different wording. |
| Assume an old copy reflects a current policy. | Review source details and refresh changed material. |
No. Automatic document retrieval checks the training entitlement before calling the search service. Ordinary eligibility checks still run, but that document embedding and search path is skipped.
There is no document retrieval without usable trained content. A trained website, pasted text, or connected source also counts as content, so uploaded files are not the only way to build knowledge.
No. Copying a source and paid indexing are separate actions. Use Make searchable or the supported training flow when you want material included in automatic retrieval.
No. Training indexes material you choose. Advanced Memory can capture and recall context from eligible conversations even without uploaded files; it has a separate setting and can use additional credits.
Retrain that source or use a supported scheduled refresh. Until refreshed, trained content reflects the indexed copy. Use a connected tool when the task needs a current record.
Yes, when access allows it. The documents do not need a separate upload for each employee, but retrieval still checks the current requester and source permissions.
I uploaded our 80-page product spec and connected our Notion wiki. Now the agent answers questions about our product better than most people on the team.
Custom Knowledge Training is part of Things that make them smart.
Combine conversations, remembered context, trained references, and available tools. Choose what becomes searchable and control Advanced Memory separately.