Cash position, runway and the call you may need to make.

What it does

The agent runs at entity level, then works through every fund listed in its setup. For each fund it does the following.

1. Checks data freshness. Verifies the fund's cash account data and pending — not yet executed — investment data are recent enough. A fund whose cash data is older than its configured threshold, or which has no cash data at all, is marked “blocked” and excluded from the calculation for that run. This is a deliberate safeguard, so that a runway or shortfall recommendation is never calculated off stale numbers.
2. Calculates the cash position for every fund that passed the freshness check:
   • Cash available — the most recent account balances tagged as cash, added up across the fund's entity and any child entities such as SPVs.
   • Upcoming investments — the total of any pending investments not yet executed.
   • Estimated expenses — a 90-day forward estimate built from account entries tagged as expenses. The method depends on how much history exists: four or more months of data uses a trend-based projection, two to three months uses a recency-weighted average, exactly one month repeats the most recent figure, and no historical data at all returns an estimate of zero with a “no data” flag.
   • Buffer required — the latest balance held in the reserve accounts.
   • Net position — cash available minus upcoming investments, estimated expenses and the required buffer.
   • Runway in days — cash available divided by the average daily outflow, being investments plus expenses spread over the 90-day window. Shown as “180+ days” in the email when there is no outflow at all to divide by.
   • Status — a fund is marked “shortfall” if its net position is negative, “watch” if its runway is under 90 days, and “surplus” otherwise.
   • Recommended capital call — for a fund in shortfall, the size of the gap plus the required buffer: enough to cover the shortfall and top the reserve back up.
3. Generates reports. Saves a report for each fund, and for any fund that was not blocked, also saves a formatted report document into that fund's private Reports folder.
4. Sends an email. A single consolidated report covering all configured funds goes to everyone on the recipient list.

Nothing is ever written back to your account entries, investment records or any other source data. This agent only reads and reports.

What it needs

This agent does not have a standard settings form like most others. Configuration happens in two different ways.

Setup warnings

Incomplete setup is not invisible: the agent's General tab shows a yellow warning banner whenever any of the following is true, and each item explains what will, or will not, happen until it is fixed.

• No funds configured — the agent would analyse nothing and still report success.
• A configured fund no longer exists, for example because it was deleted or moved — that row silently drops out of every run.
• The cash account types are not set — every fund would be reported blocked with a misleading “no cash data” message.
• No report recipients — the consolidated email would never be sent.

A clean General tab means the run will produce real numbers and the email will go out.

Other data inputs

• Account entries in the cash, expense and buffer categories, from your existing fund and entity ledger data in CapHive.
• Pending investment records that have not been executed, from your existing deal data.
• Each fund's entity and any child entities. Cash, investments and expenses are added up across the fund's main entity and all of its child entities, including SPVs.

Turning it on

An administrator enables the agent with a feature toggle under Admin → Edit Entity → Support Agents. No AI provider key is required — the reporting pipeline is fully rules-based, adding up account entries and running the expense projection maths, and it never calls an AI model.

Running it

Open AI → Support Agents → Cash Flow Assistant. The standard Run Now button is available once the feature is enabled for your entity, and runs the full check against every fund currently listed in the setup.

Configuration happens on the agent's Edit page, in the Funds & Staleness Thresholds table: add a fund row, pick the fund, set its two freshness thresholds in days, and save. There is no “run for just one fund” option — a run always covers every fund listed there.

Scheduling

Supported through the standard Schedule tab on the agent's page, the same way it works for every agent: set a frequency and time, turn it on, and save. Scheduled runs produce the same per-fund reports and email as a manual run.

There is also a separate daily reminder. Each day the system checks whether the next scheduled run falls within the configured reminder lead time; if it does, it emails the reminder recipients a preview of each fund's data freshness, so fund managers can fix stale data before the real run happens and that fund ends up excluded.

Where the output goes

• Fund Reports tab on the agent's page — the report for each fund from every run.
• Each fund's private Reports folder — a “Cash Flow Analysis” report file is saved there for every fund that was not blocked. Running the agent again replaces this file rather than keeping multiple copies.
• Consolidated email to everyone on the report recipients list. It shows an overall status banner (Action Required, Monitor or All Clear), a portfolio summary table with one row per fund giving status, net position, runway and recommended action, and a detail section for each fund that was ready to calculate: cash available, upcoming investments, estimated expenses, buffer required, net position and a cash breakdown by entity. Blocked funds show as dashes in the summary table, with “Update stale data” as the suggested action.
• Staleness reminder email, sent separately to the reminder recipients ahead of the next scheduled run, listing each fund's freshness status and the specific reason it would be blocked — for example that an entity's cash data is 45 days old against a maximum of 30.
• Chat assistant, a secondary feature. When chat is enabled for this agent, a chat panel answers natural-language questions using the most recent completed report for each fund as its source of truth. It has a built-in “what if” calculator for stress-testing scenarios, and is designed to always use that calculator rather than work out hypothetical numbers on its own. Sample starter questions: which funds are running low, show stale balances, recommend capital calls.

Reading the result

Each fund's report includes:

• Status — ready or blocked, the outcome of the freshness check.
• Block reasons — a plain-language list explaining exactly why a fund was blocked, such as stale cash data or stale pending-investment data for a specific entity.
• Fund status — shortfall, watch or surplus, shown only for funds that were ready to calculate.
• The underlying figures — cash available, a per-entity cash breakdown with last-updated dates, upcoming investments and the list behind that total, estimated expenses and which estimation method was used, the monthly expense projections and how many data points fed them, buffer required, net position, runway in days, recommended capital call, and currency.

Where it can fail

• The settings that actually drive the numbers have no on-screen panel. Which account types count as cash, expenses and reserve buffer must be set up before any of those figures can be anything other than zero, and there is no field for this anywhere in the product today. A fund manager who only fills in the Funds & Staleness Thresholds table will see every fund show zero cash available, zero estimated expenses flagged “no data”, and zero buffer required.
• Stale cash data quietly excludes the whole fund. If even one of the fund's entities, including a child SPV, has no cash-tagged account data at all, or its most recent entry is older than the configured threshold, the entire fund is excluded from that run's calculation. There is no partial result. The same applies independently for stale pending-investment data.
• Stale-but-plausible balances can still feed a misleading recommendation. The freshness check only looks at how recent the latest data point is, not whether it is accurate. A fund can pass on a balance that is recent but simply wrong, and still produce a shortfall or surplus recommendation that does not reflect reality.
• A blank report recipients list skips the email. The email step is marked “skipped” in the run history, and the agent's page shows a warning banner ahead of time whenever the recipient list is empty — check there rather than the run history.
• An empty fund list. With no funds configured, the freshness check and the calculation both report zero funds processed, and the run completes successfully with nothing to show. The warning banner flags this state before you run.
• The quality of the expense estimate depends entirely on how much history you have. A fund with only one month of tagged expense data gets a flat repeat of that figure for the whole 90-day forecast, with no sense of trend. A brand-new fund with no historical expense data gets an estimate of exactly zero, which understates true expected spending until enough history builds up.
• Combining a fund with its child entities assumes consistent tagging. Cash, investment and expense data is added up across the fund's main entity and every child entity; if account tagging is applied inconsistently between the parent and a child SPV, the combined total can double-count or miss a child's contribution with no visible warning.
• The saved report file is replaced, not versioned. Each new run deletes and replaces the existing Cash Flow Analysis file in the fund's Reports folder rather than keeping past versions. Only the report history list preserves the run-by-run record.

Questions we get asked