AI voice assistant for project management

Managing a project is largely a matter of writing things down: hours, tasks, issues, the agreements of a meeting. The PjMan assistant lets you do it by speaking or typing, just as you would tell a colleague —"log 2 hours on T0003"— and it prepares the record ready for you to review. You decide whether it is saved.

What you can ask it

• Record things. Tasks, phases, expenses, issues, risks, change requests, contacts, meeting minutes with their agreements, and time entries. If a required detail is missing, it asks you instead of inventing it. (Phrases that work)
• Look things up. "Which tasks are assigned to me?", "how many hours have I logged this week?", "how much have we spent against plan on this project?".
• Take you to a screen. "Open the risks of this project".
• Analyse. "How is this project going?" summarises progress, hours against the estimate, issues, delays and upcoming milestones, and gives a forecast end date when there is enough history. "Which projects are doing worst?" returns a traffic-light view of your portfolio with the reason for each colour. (Status, forecast and portfolio)
• Draft texts. A draft of the client status report or of the follow-up meeting, for you to copy and send: the assistant sends nothing.
• Turn notes into data. Paste the notes of a meeting and get the minutes with their agreements, the tasks and the risks, ready to confirm. (Pasted text)
• Read and write comments. “Summarise the comments on this task” or “comment on this issue that it is fixed” (with mentions, if you ask). What it posts can't be edited or deleted afterwards, so it always shows you the exact text on a card first. (Comments and mentions)

Dictation and reading aloud

The assistant panel has a microphone: you press it, speak, and what you say is written in the box as you talk. When you stop, the message is sent on its own. (Voice: dictation and reading aloud)

• Languages. Spanish, English, German and French, the one selected in the application.
• Browsers. Dictation uses the browser's speech recognition: it is available in Chrome-based browsers, in Edge and in Safari, but not in Firefox; there the button appears disabled and you type the message.
• Reading aloud, if you turn it on. The assistant can read its answers out loud. It is a setting of your account, off by default.
• Your voice, your browser. It is the browser that turns speech into text; PjMan neither receives nor stores the audio, only the resulting text.

Nothing is saved without you seeing it

Speech recognition can hear "twelve hours" when you said "two". That is why dictated input is handled with care: anything created from a dictated message always asks for a confirmation card, even if you have turned confirmations off in your settings. The card has three buttons: Confirm, Edit and Cancel. (Confirmation by risk level)

• Changing and deleting always ask for confirmation, whether you dictate or type.
• Undo for 10 minutes whatever the assistant created or changed, with a button. Some deletions cannot be undone —a task, an issue, a phase or a set of minutes—; in those cases the card warns you first and asks for a second click.
• One batch, one card. If you ask for several things at once ("today I spent 3 hours on the migration and 1 in the meeting"), they come in a single card where you tick or untick each item.

It knows which screen you are on

The assistant knows what you have open —the project, the phase, the task— and understands "this task" or "this project" without you naming it. It also understands the short codes PjMan uses: T0003, expense G0002. (The assistant panel)

Your permissions and your key

• The same permissions as you. The assistant goes through the same services as the application: if you cannot do something, it cannot either. A PM can propose almost anything in their projects; a consultant only what the application lets them do, such as logging hours on their tasks.
• Your own AI key. There is no shared key: each person adds their own in Settings, it is stored encrypted, and the usage is billed to their account with the model provider. You control the usage and the cost. (AI key)
• Or a free model, if you accept it. Without a key you can turn on a free model from an external provider, with its privacy notice, your consent and a daily quota. (Free model)
• Optional. Without either, the panel invites you to set it up and the rest of the application works as normal.

Try it

Create your account, add your key and start with something simple: "log one hour on this task". Or explore the demo first to see the application with sample data.