The inbox
Intro threads, founder updates and data room links, spread across 400 emails and three people's inboxes.
Every deal leaves a trail: intro emails, a data room, IC debates and the reasons you passed. Mobtasks keeps that trail on one record. AI coworkers screen decks and work the data room against your checklist, cite every source, and wait for the deal partner before anything goes to a founder.
Pattern recognition is what a fund sells to its LPs. Today it lives in inboxes, memo drafts and partners' heads. When a principal leaves, the reasons behind hundreds of decisions leave too.
Intro threads, founder updates and data room links, spread across 400 emails and three people's inboxes.
IC_memo_draft3.docx is on a desktop. The IC debate that changed the decision is in nobody's notes.
“Why did we pass on Ledgerly?” Only the principal who ran the deal knows, and they left in March.
One channel per deal. Agents do the reading and the matching. Your people decide at every step that matters, and every step lands on the record.
A Monday pipeline screens each new deck against the thesis and drafts a one-page note per company. Your review is a stage, so nothing passes it by itself.
Search by text and by meaning across every past deal, memo and IC thread. Answers name their sources and come only from deals you may read.
Sync the data room to the deal folder. The agent screens every file first, then matches the room to your checklist and lists the gaps, with questions for the founders.
The memo cites the cohort file, the customer call and the deal you passed on. The IC debate stays in the thread, next to the memo it changed.
Questions and documents for a founder wait for the deal partner's approval, in the workspace or in Slack. The record keeps who approved what.
Board decks and monthly updates stay with each company. A pipeline drafts a summary of each update for the partners' meeting.
When an LP asks why you led a round, the record holds the memo, the IC thread, the approvals and the files they cited.
Scribe turns IC debates into the fund's positions, such as how you judge churn at Series A. Every line points to its source, and a partner promotes it.
The deal channel holds the data room, the questions, the answers and the sign-offs. The fund's playbook grows from them.
Ask in the deal channel. The agent matches every checklist item to a file and lists what is missing: churn, the signed cap table, IP assignment.
Every answer names the file, the IC note and the earlier deal it used, so you can check it before it goes in a memo.
An agent reads as the person who asked. A deal you cannot open stays out of its context and out of its answers.
Anything sent to a founder waits for the deal partner. Approve in the workspace or in Slack. Routine reading runs without approval.
Memos, models and trackers open in the deal. A folder on your Mac syncs both ways, and a conflict keeps both copies.
See agent spend per accepted memo and per deal, across every agent and model.
Chief of Staff asks up to three questions, then proposes AI coworkers and channels. You edit or untick each role before anything is created, and every new agent starts restricted.
An example team for a Series A fund. Your proposal follows the work you describe.
Every message, run, file change and approval is one event on an append-only record. It shows who did it, person or agent, what was approved and by whom.
Maya asks Diligence Analyst to check Quillo's data room against the Series A checklist
1 file held. A PDF with hidden instructions is quarantined before the agent reads it
Diligence Analyst matches 64 files to the checklist. 41 of 44 found, 3 gaps
Maya answers: use the cohorts to June, and ask for July
Questions parked. Sending to Quillo's CFO waits for the deal partner
Ben approves. The record keeps who approved, and the questions as sent
Search the record for a company and find the memo, the IC thread and the reasons, even after the principal who ran the deal has left.
Every decision links to the files, the debate and the approvals behind it.
See agent spend per accepted memo and per deal, across every agent and model.
A fund holds founders' data under NDA. This is how Mobtasks handles it.
On your own machines. A local worker runs each agent, and model credentials stay on that machine. They never pass through our servers.
Only the people you give access, signed in through your SSO with two-factor codes. An agent reads as the person who asked, so a walled deal stays out of its answers.
Untrusted files are screened before an agent reads them. An injection attempt is quarantined, and the run waits until the screen finishes.
The Run Inspector shows every byte of context the platform added to a run, and lists each workspace read the run made.
What its permission preset allows: Supervised, Standard or Autonomous. Risky actions park until a person approves, and the record keeps the decision.
The record is append-only and sealed with your workspace's own key. Export the whole workspace at any time, or delete it.
Bring one deal in diligence and your checklist. We'll set up the team with you during the private beta.