These are the AI project management questions teams ask us most, answered in a few lines each. Every answer says how it works in AlianHub, so you can check it yourself in the demo. If your question is about cost, safety or control, start with sections two and three.
Key takeaways
- Most AI questions are really trust questions.
- An AI in AlianHub can never do more than the person it acts for.
- Deleting, billing and permissions do not exist as agent actions.
- Costs are capped per agent, and you choose the model.
- You can start with zero autonomy and earn your way up.
The basics
1. What does AI do in AlianHub?
It answers questions in the Ask box, plans projects from a brief, and runs agents that comment, create subtasks, post digests and propose bigger changes.
2. What is the difference between Ask and an agent?
Ask answers when you ask. An agent is a workspace member that is assigned work and acts within its allowed actions and autonomy level.
3. Do I have to use AI?
No. Views, time tracking, chat, automations and Email to Task all work without it.
4. Which AI models work?
AI Assist runs on your own LLM API key, and multiple providers are supported. You can choose the model per question or pin one per agent.
5. Can I use the Claude or ChatGPT account I already have?
Yes. AlianHub ships an MCP server, so Claude, ChatGPT or Claude Code can connect and work as you. See Connect your AI.
Control and safety
6. Can an agent delete my work?
No. Deleting a task, doc, comment or project is not an available action for agents.
7. What are autonomy levels?
Four levels, L0 to L3. L0 answers and suggests, L1 proposes every change, L2 runs low-risk changes, and L3 can also run on a schedule.
8. What counts as low risk?
A comment, a nudge, a status change within a list or a report.
9. Where do proposals go?
To the AI Inbox, where a person approves or declines. A decline can include a reason the agent sees.
10. Can an agent mark a task Done?
No. Agents may set In progress or In review. A person closes the task.
11. Can I undo what an agent did?
Changes an agent makes on its own come with Undo, and every action is written to the audit log.
12. How do I stop everything at once?
One switch pauses every connected agent in the workspace. A change a person already approved is still made.
13. Can one project be stricter than the rest?
Yes. A project can be stricter than the agent, so a sensitive project can hold every agent to proposals only.
The safest AI action is the one that does not exist.
Cost and data
14. How do I control AI cost?
Each agent has a daily action limit and a monthly spend cap. A workspace can also set a daily AI budget.
15. Is AlianHub itself free?
Yes. It is free and open source under AGPL-3.0, self-hosted, with no per-seat pricing.
16. Where does my data live?
On your servers. Only the model call is external.
17. Can a connected AI see everything?
No. It sees only what you can open and changes only what your role allows. It cannot read your direct messages.
18. What is AI memory?
Private notes about how you like to work, with up to 50 short facts. Only you can see it, admins included, and you can choose "Forget everything".
Getting work done
19. Can AI plan a project for me?
Yes. The AI project generator takes a brief, asks questions and shows a plan of tasks, subtasks and checklists. Nothing is built until you approve.
20. Are there ready-made agents for my team?
Yes. Team packs turn on role agents for 15 teams in one click. They start paused, and a pack can be undone.
21. Can I write my own skills?
Yes. You can create, version and retire your own skills, and every skill has a Dry run.
22. How do I send many tasks to agents?
Select the tasks and the router proposes an agent per task, refusing those that need a person.
23. Can agents run on a schedule?
At L3, yes. A Reporter can post a digest every Monday morning, for example.
24. How many agents can one person supervise?
AlianHub's guidance is to plan one person to supervise 3 to 5 seats, where a seat is one connected AI working a role.
25. Can I see how well an agent is doing?
Yes. Health views show error rate, approval rate and cost per agent and per model, plus a quality view over 7, 30 or 90 days.
Quick reference
| Question | Short answer |
|---|---|
| Can agents delete things? | No |
| Who approves wider changes? | A person, in the AI Inbox |
| Who closes tasks? | A person |
| Who pays for the model? | You, with your own key |
| Where is the data? | Your servers |
How to do this in AlianHub
To answer most of these questions yourself in ten minutes:
- Open the Ask box and ask "What should we get done?" with "Everything you can open".
- Open the Agents tab and check each agent's level, skill and what it writes.
- Open an agent's runs and limits to see actions today, spend this month and the pinned model.
- Approve a proposal in the AI Inbox, then press Undo.
- Open AI, then Connect your AI, and copy the address for your own app.
Tip: do this in the live demo before any internal AI meeting. Answers you clicked through yourself settle debates faster than slides.
A business scenario
A finance lead blocks an AI rollout because nobody can say what agents may touch. The operations manager walks them through questions 6, 10, 12 and 14 in the demo.
The lead sees that billing and permissions are not agent actions, that the Invoice Preparer only drafts, and that spend is capped. The team turns on the Finance pack paused, starts at L1, and month-end timesheet checks begin moving faster with the finance lead still approving.
Show the limits first. Show the features second.
For deeper reading, see AI agents for project management and the AI controls page.
FAQ
What are the most common AI project management questions?
Teams mostly ask what AI can change, who approves it, what it costs and where data lives. This post answers 25 of them for AlianHub.
Is AlianHub AI safe for client or regulated work?
Destructive and permission actions do not exist for agents, projects can be held to proposals only, and every action is logged. Review it against your own policies.
Do I need technical skills to use AlianHub AI?
No. You ask in plain words, approve in the AI Inbox, and connect your AI by pasting an address.
Where can I ask a question that is not here?
Use GitHub Discussions, linked from the community page, or contact the team.
Got a question we missed? Try it in the live demo, self-host AlianHub for free, and share this list with the person on your team who has the most questions about AI.