Questions and glossary
Frequently asked questions
44 canonical questions and answers — a single source of truth for customers, the team and AI systems.
Below is the full, canonical list of questions and answers about Slimaco AI — the same text is marked up with FAQPage (schema.org) structured data so both people and AI systems can use it.
An AI platform for organizations that captures company know-how and automates teams' repetitive work. Knowledge mining collects agreements, decisions and customer data from email, messaging and department systems such as CRM, ERP and helpdesk, and combines them into shared company knowledge. Agreements and decisions are added after a person approves them, while hard data, such as a customer's open tickets, updates continuously. AI agents work on that knowledge: they run tasks on demand or on a schedule in the apps the company already uses (Microsoft 365, Google Workspace and 1,000+ others), within permissions and human approval. The team uses that knowledge in the Slimaco assistant and soon also in ChatGPT, Claude, Gemini and Microsoft Copilot via MCP, always with the permissions of the person asking.
No. Company knowledge is the foundation in Slimaco, not the product itself. The assistant and agents work on that knowledge: they answer questions, prepare documents and run repetitive tasks in the company's apps.
No. Slimaco uses models as an execution layer and adds company knowledge, agents, permissions, integrations and governance.
A general assistant starts from an individual user's conversation. Slimaco starts from the organization's shared knowledge, roles and rules, and the assistant and agents work on top of them — agents can run tasks without a conversation, e.g. on a schedule.
A specialized task executor that the Slimaco assistant creates from a plain-language description. The agent gets tools in connected apps, a scope of company knowledge and a way of running: on demand or on a schedule.
Yes. An agent can run e.g. every morning, weekly or hourly: collect data from several apps, prepare a report or drafts and pass the result to the chosen people. Every run is recorded.
For example replies to quote requests using the price list, complaint handling by the current procedure, overdue payment reminders, reports and summaries, pre-meeting briefs, a new hire's first-week plan or market monitoring. You describe the task in plain language and Slimaco creates an agent that runs it in connected apps. Examples by department are on the Agents and automations page.
Not without your decision. Approval can be set in two ways: on every run of the agent, or on the first run with the action approved for a set period. A draft, saving to a mailbox and sending are separate actions.
The Slimaco assistant shows a card with the action's name, service, account and parameters. The user can allow it once, for the conversation, on a time-boxed basis, or decline.
Yes, when the integration is connected and the tool is enabled. Agents read and write data in the company's apps, and write actions go through human approval.
With Microsoft 365 (Outlook, Outlook Calendar, OneDrive) and Google Workspace (Gmail, Google Calendar, Google Drive), and, through the Composio integration layer, with more than 1,000 other apps, including HubSpot, Pipedrive, Salesforce, Slack, Microsoft Teams, Jira, Asana, Notion, Trello, Shopify and QuickBooks. The full list is on the Integrations list page of the knowledge base.
Yes. Integrations support multiple accounts from one provider, but choosing an account inside a conversation isn't yet a universal selector.
Yes. Answers based on company knowledge show citations and where the information comes from.
No. Access depends on role and on rules assigned to a collection, source or document. The same rules apply to assistants and agents.
A mechanism that analyzes day-to-day communication — email and conversations with the assistant — and picks up agreements, decisions and changes in how work is done. Captured fragments go to review and join shared knowledge only after the area owner approves them.
It shouldn't. Uploading, approval and indexing are separate states.
Slimaco says there isn't enough knowledge in the company instead of presenting an unsupported answer as fact. When needed, it can turn to open sources on the web.
The organization names owners for knowledge areas. Slimaco helps them: knowledge mining suggests new agreements, review lets them approve them, and the system points out gaps and outdated content.
A configured AI role with its own instructions, knowledge scope, tools, model and visibility.
It can. Slimaco can pick zero, one or several assistants based on their description. The user can also name an assistant explicitly with @.
A repeatable instruction for performing a task, described in plain language. A skill is not a separate agent or program code.
No. It can only narrow the way of working and the tools available within an already-permitted context.
Yes. Slimaco can split a task into steps and route them to different assistants or agents. Each step is subject to its own permissions, knowledge scope and approval rules.
It remembers context within one conversation. Memory across different conversations is planned.
No. Slimaco uses models such as GPT, Claude, Mistral, DeepSeek and Kimi and matches the model to the task. The model vendor can be changed without rebuilding the platform, and in BYO AI mode you can use your own model access.
In standard SaaS, company knowledge is stored in the EU, and searching it runs on Slimaco's own infrastructure. Only the question and the knowledge fragments needed for the answer reach the language model, with personal data masked. The default model, GPT-5.4 Mini, as well as DeepSeek V4 Flash and Mistral Large 3, run in Microsoft Azure in the EU data zone, with an EU-only processing guarantee. The other selectable models, including Claude, process data globally. The full list is on the AI models page.
Yes. Slimaco runs primarily as a SaaS service, but it can also be deployed on-premise or in dedicated client infrastructure — the deployment model is agreed individually.
Yes. Slimaco runs as SaaS: you set up a trial account (7 days, no card), connect mailboxes and documents and describe the first task. Full implementation by our team — process analysis, preparing knowledge, integrations and launching automations — is an option, not a requirement.
It starts with Discovery AI: diagnosing processes, knowledge sources, owners and metrics, and choosing a first, limited scenario — e.g. a single automation. More areas are added once the first one delivers results.
No. It takes over repetitive work and relieves experts of the same questions, but high-risk decisions remain with a human.
Especially organizations with several teams, scattered knowledge, repeatable processes and a need for control — most often companies of roughly 20–500 people.
Through the time it takes to complete repetitive tasks, the number of tasks done by agents, time to access knowledge, the number of questions to experts, answer acceptance, onboarding, errors and financial results.
Automation in the context of the company. Agents and the assistant work on the organization's approved, continually updated knowledge, with access control and human approval. Over time that knowledge becomes the company's memory, used by every new model and every new automation.
It depends on scale and purpose. For one person's individual, ad-hoc tasks, regular ChatGPT access is usually enough. If the goal is for the whole organization to work on shared, reliable knowledge, with repetitive tasks done by agents following company rules and under control, you need a platform like Slimaco AI. You don't have to give up ChatGPT: soon you'll connect it to Slimaco via MCP.
Microsoft 365 Copilot is a good tool for employees in the Microsoft ecosystem. Slimaco has a different starting point: it builds shared, approved organizational knowledge with versions and owners, picks up new agreements through knowledge mining, creates agents from a task description and isn't tied to one model vendor. It also works with Google Workspace and more than 1,000 other apps. Soon Copilot will be able to use Slimaco knowledge via MCP.
No. Soon you'll connect them to Slimaco via MCP and they'll answer based on the whole company's knowledge, with the permissions of the person asking. Teams that prefer one place for knowledge, conversation and agents use the Slimaco assistant. It runs on GPT, Claude and other models, so writing, analysis and research are on the same level.
Soon, yes. Slimaco exposes company knowledge via MCP (Model Context Protocol). At launch, external AI tools read knowledge; as a next step they'll also be able to run Slimaco agents. The organization's admin decides whether the connection is enabled and which tools may use it.
Only the knowledge fragments needed for the answer that the asking person is allowed to see, with personal data masked. Further processing is the tool vendor's responsibility under its agreement with the organization. We recommend connecting company accounts, not private ones.
Yes. Queries from ChatGPT, Claude, Gemini or Copilot via MCP use SU just like questions asked in the Slimaco assistant.
From documents (manual upload, OneDrive, Google Drive), email and calendars (Microsoft 365, Google Workspace), messaging (Slack, Microsoft Teams) and department systems such as CRM, ERP, helpdesk and invoicing. Agreements and decisions join shared knowledge after a person approves them, while hard data, such as the number of a customer's open tickets, updates continuously.
Both. Company knowledge — the organization's approved sources and documents — comes first. When the answer isn't inside, Slimaco turns to open sources on the web, and for new topics it can run its own online research.
No. Slimaco integrates with the existing environment — Microsoft 365, Google Workspace and more than 1,000 other apps — and isn't tied to one language-model vendor. The team can also keep its AI tools: soon ChatGPT, Claude, Gemini and Copilot will connect to Slimaco via MCP.
Slimaco delivers the most value to organizations with several teams, scattered knowledge and repeatable processes — most often companies of roughly 20–500 people. It also works for smaller teams with complex, repeatable procedures — e.g. law firms use Slimaco from 5 users, and the smallest package is recommended for teams of about 10 people, with unlimited users.
Knowledge typed into private chats isn't collected anywhere or available to the rest of the company — that's one of the problems Slimaco solves. From the start, Slimaco builds shared organizational knowledge: from approved sources and from agreements that knowledge mining picks up from day-to-day communication.