01
AI that knows your company knowledge
Connect Slack, Notion, Gmail, and documents, and agents answer on top of company knowledge with sources shown on every answer.
Multiplayer AI your whole team uses
Single-player AI
only gets you so far.
People and AI agents work as one team on top of your company knowledge. Permissions differ per person, and your data stays inside your company.
Dasis is a Korean multiplayer AI workspace that teams use together.
지식· 지식 그래프
열람 범위: 영업팀실시간
이
한빛전자협력사
진행 딜 3건박
박지현영업 담당
딜 소유자3분기 재계약딜
₩48,000,000초안검증실행
커버리지
92%
근거 있는 답변 출처 47건 · 연동 시스템 6개
빠르게 성장하는 조직들이 데이시스와 함께합니다
People and agents, one team
A workspace where people and agents work as one team
At most companies, AI is still single-player. In Dasis, your teammates and a set of role-separated agents share the same data, tools, and conversations, and work together. And the same agent answers differently depending on the permissions of the person asking.
Answers with sources attached
AI that answers from company knowledge
and leaves the evidence behind
Connect Slack, Notion, Gmail, and your documents, and agents answer on top of company knowledge. Every answer carries its sources, and results the team approves accumulate in the knowledge base. Company data is not used for model training.
Assets that accumulate for the team
A workspace that fits your team the more you use it
Repeated work is saved as a shared skill and runs the same way next time. An agent configuration one person builds can be reused as-is by the whole team, and answers the team approves accumulate in the knowledge base.
02
Different permissions per person
What an agent can read is determined by department, seniority, and project relationships (ReBAC).
03
On-premises option
For internal-network environments, deploy on-premises. Scope is confirmed after an environment assessment.
04
Freedom to choose your model
Switch between OpenAI, Anthropic, and Google models; local models are an option under on-premises deployment.
How each department uses it
Different work per department, one workspace
Agents take on the work each department repeats most. Irreversible steps such as sending or approval are confirmed by a person.
Enterprise security
When AI can reach your company knowledge, 'mostly secure' is not enough.
A dual-layer permission model that separates what an agent can access from who is allowed to use it. Tenants are isolated, every read and execution is logged, and irreversible actions do not run without human approval.
Security setup, 14 items
01Isolation and access control
- Per-tenant isolation
- Encryption in transit and at rest
- Per-tenant, least-privilege credentials
- Dual-layer agent permissions
- Customer data is not used for model training
- On-premises and private cloud deployment options
02Agent guardrails
- Irreversible actions run only after human approval
- A verifier agent cross-checks drafts
- Reversibility tiers decide what may run automatically
- Prompt-injection defense on externally ingested content
03Audit trail
- Every agent read and execution is logged
- Audit logs per workspace
- Ingested text is never executed as instructions
- Answers trace back to the sources they used
Questions security teams ask most
- Does our data leave our company?
- Company data is stored inside your Dasis workspace, and only the content needed to generate an answer is sent to the API of the model provider you selected. That content is not used for model training. If your environment cannot permit external transmission at all, the on-premises deployment option keeps everything inside your own infrastructure.
- Is a model trained on our data?
- No. Company data is not used for model training. Dasis does not train its own models on customer data, and requests to connected model providers are sent under no-training terms.
- Is it open-source based?
- We built our own security and permission layer on top of a proven open-source core. Permission management, audit logging, and on-premises deployment are implemented and supported by Dasis directly.
Multiplayer AI workspace
The best teams don't stop at using AI.
They run it.
The best teams don't put a chatbot on top of scattered knowledge and wait for it to find things. Dasis connects agents to the systems where work actually happens — email, CRM, and documents.
Permissions apply per person, and every execution is logged.
Irreversible actions run only after a person approves them.