01Specialist agents
AI specialists for the work you do.
Agents with relevant expertise research sources, analyze information and write content, using the models, materials and tools your task needs.
Share your brief and materials. Get reports, plans and drafts from your AI team, ready to review.
Describe your task. Type @ to choose an agent
Task context, progress and interim results pass directly between AI agents, reducing repeated explanations and copying and pasting between steps.
01Specialist agents
Agents with relevant expertise research sources, analyze information and write content, using the models, materials and tools your task needs.
02Intelligent coordination
TITU Work breaks your goal into tasks and assigns the right agents, so you spend less time coordinating the work.
03Ready for review
Reports, plans and drafts from your agents are brought together for you to review, edit and use.
From organizing sources and researching markets to business analysis and proposals, agents coordinate around your goal. Choose a scenario to see how it works.
Example conversation
✓ Deliverable
Explains changes in revenue, profit, and collections while keeping unresolved items visible.
| Metric | Q2 | Q3 |
|---|---|---|
| Recognized revenue | CNY 3.2m | CNY 4m |
| Costs + operating expenses | CNY 2.6m | CNY 3.48m |
| Operating profit | CNY 600k | CNY 520k |
| Collections | CNY 3m | CNY 2.8m |
Financial-close basis; the CNY 600k order awaiting acceptance is shown separately and excluded from revenue.
Q3 revenue rose, but profit fell from Q2. Cross-check these three files, explain the sources of growth and cost increases, and prepare a management presentation. Keep orders awaiting acceptance separate from receivables due for payment.
Example conversation
✓ Deliverable
Interview evidence, revised claims, budget, and schedule come together in materials ready for your review.
| Page content | Revised version |
|---|---|
| Headline | Automatically organize notes and actions after meetings |
| Supporting copy | See discussion points and actions together for clearer follow-up. |
| Workflow | Transcribe → Summarize → Confirm actions |
Qinglan meeting device · October 20 launch | design draft
Qinglan's meeting device launches in October. It transcribes recordings, summarizes discussion, and extracts action items. Use these product files and interviews to prepare a launch page, key visual, customer announcement, and two-week schedule.
Example conversation
✓ Deliverable
Quotes, inspection evidence, and delivery history inform a plan you can adjust and verify.
| Parameter / result | Current scenario |
|---|---|
| Total / first / remaining | 1,000 / 200 / 800 units |
| Unit / shipping / inspection | CNY 103 / 300 / 300 |
| Quoted two-batch total | CNY 104,200 |
| Extra vs. single order | + CNY 600 |
First-batch quantity, shipping, and inspection are adjustable. Expedite fees remain unquoted.
We need 1,000 parts. Compare three suppliers on price, quality, and lead time, including the inspection photo. Recommend a plan based on more than unit price.
Example conversation
✓ Deliverable
Interview materials follow the role framework, preserve evidence gaps, and leave invitations for your approval.
| Candidate | Follow-up question |
|---|---|
| Chen Nan | Which decisions and rollout steps did you own? How did you handle disagreement? |
| Zhou Yu | When did customer research change your priorities, and on what evidence? |
| Song Jia | How did you align engineering and business for launch? What did you do when delayed? |
Ask the shared core question before individual follow-ups.
We have three product manager candidates. Compare their materials with the role requirements and first-round notes. Use our competency framework to prepare next-round questions and a shared interview record.
Example conversation
✓ Deliverable
Evidence-based risks and cash scenarios give document requests and further review a clear basis.
| Scenario | Balance after repayment | Minimum monthly balance |
|---|---|---|
| CNY 800k received in January | CNY 900k | CNY 400k |
| Not received before maturity | CNY 100k | − CNY 400k |
CNY 300k interest and principal repayment at maturity are model assumptions. The stress scenario has a CNY 400k short-term gap.
Yuanchuan Technology is applying for a CNY 5m one-year loan. Review finance, business, legal, and tax materials, focusing on repayment sources, critical gaps, and risks. Make the report traceable to its sources.
Verify account access, encrypt credentials such as API keys, and run code and process files in a cloud sandbox. Tools follow your permissions, and you confirm the specific actions that require approval.
Sign-in verification checks identity and controls access to protected work features.
AES-GCM encryption protects API keys and other credentials, reducing the risk of sensitive settings being read directly.
Code and files are processed in a cloud sandbox.
File access is limited by the sandbox environment and tool permissions.
Set tools to run automatically, require approval, or stay disabled.
Review each requested action and its purpose before deciding.
Common questions about TITU Work
Agents can organize sources, research, analyze information and write content around your goal. Reports and proposals are brought together for you to review, edit and use. These are example workflows using sample data.
Start by describing your goal and providing relevant materials. Clear instructions and business context help agents coordinate. You should still verify important information and final results.
TITU Work coordinates specialist agents around a shared goal, breaks down tasks and combines their results. This reduces the effort of assigning each step and repeating instructions.
You define the goal, supply materials, confirm actions that need approval and review the results. You stay in control of key decisions and the final outcome.
Depending on the goal, you may receive organized source materials, a research report, a proposal or written content. Agents combine their outputs for you to review, edit and reuse.
Use models such as GPT, Claude, Gemini, DeepSeek, Qwen and Kimi, or connect external agents such as Claude Code, Codex, Kimi Code and OpenCode. Available models depend on workspace settings and permissions. External agents also need their software installed on the execution device and the required provider account access.
Tools can be configured to run automatically, require approval or remain disabled. When approval is required, you see the specific action and its purpose before deciding whether to allow it.
Choose a task your team does regularly. Define the materials, permissions and expected result, then evaluate the workflow in a demo. Demo booking is not available yet.
TITU Work brings agents together around a shared goal, from the first task to a deliverable you can review. Key decisions and the final outcome stay with you.