Keep me posted

AI has transformed how work gets done.It's time for depositary oversight to follow.

An AI-native platform for depositaries of alternative funds. Cash flow monitoring is the first solution in the box.

Period overview in DepositAIry.com Cash Monitoring: 98.1 percent of movements cleared with evidence, 26 waiting for the analyst.

The duty

Every movement. Every account. Every day.

Cash flow monitoring is the one depositary duty that cannot be delegated. The rules are short and stable. The work is not.

  • Art. 86(b)

    Reconcile everything.

    Every movement on every in-scope cash account, daily. Nothing filtered out first.

  • Art. 86(c)

    Spot what matters.

    Significant cash flows, and those inconsistent with the fund's operations, identified by close of business.

  • Art. 86(e)

    Follow up.

    Breaks and open items chased with the AIFM, escalated without undue delay.

  • Art. 86(d)

    Prove it.

    What was checked, by whom, on which data, with which result. Every period, and in the annual review.

Today this runs on bank statements, spreadsheets and inboxes.

How it works

AI does the groundwork. Before your team arrives.

Every movement is reconciled overnight. Routine flows clear under the thresholds and rules you set. Everything else reaches a person, with a recommendation.

  1. Done by: AI

    Bank feeds

    MT940, camt.053, PDF and spreadsheet statements, normalised into one record.

  2. Done by: AI

    Categorise

    Every movement categorised within seconds: capital call, distribution, fee, loan, FX.

  3. Done by: AI

    Match

    Against drawdown and distribution notices, invoices, the administrator's ledger and last month's patterns.

  4. Done by: AI

    Test significance

    Against your thresholds per fund: amount, share of NAV, category, new counterparty.

  5. Done by: AI

    Investigate

    Exceptions only. Reads the documents, the LPA and the payment history, and prepares a recommendation with its evidence.

  6. Done by: You

    Decide

    Accept the recommendation or change it: confirm, request a document, hold or escalate. The evidence trail writes itself.

Illustrative figures from our prototype: one month, one depositary.

Why AI-native

Not AI on top. AI everywhere.

Most oversight tools were built to route work between people. Adding AI to them speeds up a step. It does not change who does the work.

Workflow tools

Monitoring as process plumbing.

Who does the work
The analyst builds every match by hand, screen by screen.
What the analyst sees
A task list. A bank charge of a few cents blocks the period just like a ten-million drawdown.
What is remembered
Nothing. Next month starts from zero.
Evidence
Assembled afterwards, for the auditor.

Agents bolted on

A faster manual process is still a manual process.

Who does the work
Still the analyst, now with suggestions in the margin.
What the analyst sees
The same queue, every item opened, every answer constructed.
What is remembered
Little. The agent helps with a step, the flow around it stays the same.
Evidence
Still in folders and inboxes.

AI-native

Built around the recommendation. People keep the decision.

Who does the work
Agents reconcile every movement and prepare a recommendation for every exception before anyone logs in.
What the analyst sees
Only the exceptions, each with a recommendation and its evidence. The decision stays theirs.
What is remembered
Every decision. The second month costs less than the first.
Evidence
A by-product of the work, ready for the auditor and the CSSF.

Workflow tools decide who does the next step. An AI-native platform decides whether a person needs to.

The review desk

Review a recommendation instead of building one from scratch.

Review desk showing a short capital call receipt, the AI's recommendation, four evidence items and the decision buttons.
  1. The recommendation comes first.

    One line that says what happened, what is still open and what the AI recommends.

  2. Every claim cites its evidence.

    The notice, the bank confirmation, the LPA clause, the payment history. One click to the source.

  3. You decide.

    Accept the recommendation with one keystroke, or change it: confirm, request, hold or escalate. Your note can become a rule for next month.

  4. You stay in control.

    Nothing is confirmed without a person. Every decision is recorded: who took it, when, on which evidence, and what the AI had recommended.

Who does what

Agents do the boring and repetitive work. People decide.

AI agents

  • Read bank statements in any format
  • Categorise and match every movement
  • Read notices, invoices and agreements
  • Investigate exceptions and prepare recommendations
  • Draft document requests and assemble the evidence
  • recommendation + evidence, from the AI agents to your team
  • decision + feedback, from your team to the AI agents
  • becomes a rule, from your team to the AI agents

Your team

  • Set thresholds and procedures per fund
  • Review recommendations and evidence
  • Take the decisions
  • Escalate to the AIFM and, where needed, the CSSF
  • Teach the system with feedback
  1. Agents do the boring and repetitive work: document management, data analysis, matching. Your people review and take the decisions.
  2. Cash flow monitoring is the first solution in the box. But it won't be the last.
  3. Built by people who know depositary oversight, automation and AI from the inside.

Pilot

Shape it with us.

A small group of depositaries is co-designing DepositAIry.com with their own procedures and data. Want to hear more, or join them?

Keep me posted