AI.FO SIGNAL LIBRARY
AI.FO reads your accounting data and validates the numbers behind 48 signals before any narrative is written. This is the public view of that library: qualitative and redacted. The exact formulas, the exact thresholds, and your own numbers are the in-app view, behind sign-in.
Not every signal evaluates for every company. Which ones apply depends on that business and on what its data can support; some are labeled proxies and some report as not applicable rather than guess. The library below says so on each signal, plainly, rather than implying all of them fire for everyone.
8 recently added signals, each independently verified before it can publish. Open any one for what it detects, when it fires, and where it does not apply.
A recurring-revenue business is buying neither growth nor profit: the sum of its year-over-year revenue growth rate and its EBITDA margin sits materially below the published [redacted] line.
For the customer cohort a business already had at the start of a period, its recurring revenue is shrinking: contraction and churn among those existing customers outweigh expansion from the ones who stayed, so net revenue retention falls below the neutral line. Customers won during the period are excluded on purpose, because new logos cannot repair an installed base that leaks.
For the customer cohort a business already had at the start of a period, its recurring revenue is leaking through pure loss: contraction and outright churn among those existing customers, measured before any credit for expansion from the ones who stayed. A book can post healthy net revenue retention while its gross retention erodes, because upsell to a handful of accounts masks churn across the rest of the base. Gross retention is the read that exposes that leak, because it counts only what the business kept, never what it grew.
The business does not generate enough net operating income to cover its annual debt service (scheduled principal plus interest) with the cushion a commercial lender requires. Debt service coverage is the ratio lenders write into loan covenants, and a reading below the covenant line is a technical breach: the lender can call the loan, freeze the facility, or reprice it, regardless of whether payments are current.
Core operating profitability, EBITDA over revenue, is either sitting below a thin-margin floor or compressing materially year-over-year. EBITDA margin isolates how much of each revenue dollar the business keeps as operating profit before financing costs, taxes, and non-cash charges. A margin that is thin, or that is falling fast, means the business is losing the cushion it needs to absorb a cost shock, service debt, or fund reinvestment.
whether the cash a business is sitting on includes money that was never its own to spend, and whether the balance of that money is growing. Two legs. Leg A reads tax set-aside COVERAGE: an implied income-tax reserve (the disclosed [redacted]% assumption applied to year-to-date pretax profit, net of income tax already paid or accrued, floored at zero) plus the payroll-liabilities and sales-tax-payable balances, measured against unrestricted cash. Leg B reads trust-fund liability BUILD-UP: payroll and sales-tax liabilities rising across consecutive closes by a material amount.
a subscription or prepaid business is drawing down its deferred-revenue liability faster than new bookings and collections replenish it. Deferred revenue is cash a business has already collected for goods or services it has not yet delivered; recognizing that revenue draws the balance back down. When the balance falls across consecutive closes, recognized revenue is being funded by previously-collected cash rather than by new sales. This is a forward liquidity warning: the recognized revenue looks healthy today, but a materially declining deferred-revenue balance means part of that revenue is a drawdown of a finite reserve, and when the reserve runs out, recognized revenue steps down to whatever the current bookings pace supports.
whether what the owner takes out of the business is consistent with what the business generates, read in both directions. Leg A (over-extraction) fires when owner cash out (salary, guaranteed payments, draws, and distributions over a trailing window) runs ahead of the operating cash flow that funds it AND owner equity declined across the same window, so the business is being decapitalized while the profit and loss statement still looks acceptable. Leg B (under-payment) fires when the working owner is paid materially less than the company pays for comparable labor it hires, so reported profit is absorbing an unpriced labor subsidy and the margin the owner manages against is overstated. This is a READ of a relationship between two numbers already in the books: it does NOT opine on reasonable compensation, perform the IRS reasonable-compensation analysis, price equity, value the business, or judge whether a distribution was lawful. The full specification, primitives, and the derivation of every constant live in Part [redacted] under "Owner-Pay Sustainability"; this section is the base-signal summary.
Grouped by what each signal watches. Every card links to a plain-language page for that signal; the exact figures live in the authenticated app.
The company is running low on cash relative to how fast it's burning.
Gross burn is growing significantly faster than net burn, revenue is currently masking a structural cost problem.
Multiple hypothetical stress scenarios produce emergency-level runway.
The company's stated burn rate doesn't match what's actually happening in the bank.
The business fails multiple independent stress scenarios, indicating low resilience to adverse events.
Customers are taking longer to pay their invoices, cash is getting stuck in receivables.
The overall time from spending cash on delivery to collecting it from clients is lengthening.
Core operations are not generating enough cash to cover near-term obligations.
The company is paying vendors increasingly late, a behavioral indicator of cash stress.
The company's short-term financial cushion (current ratio) is declining over time.
The stated revenue growth plan requires more working capital than is available without crossing the liquidity floor.
A single client accounts for a disproportionate share of revenue, creating existential dependency.
Top client share is increasing over time, concentration is getting worse, not staying stable.
Project-based AR (lumpier, slower to collect, lower margin) is growing as a share of total AR.
For the customer cohort a business already had at the start of a period, its recurring revenue is shrinking: contraction and churn among those existing customers outweigh expansion from the ones who stayed, so net revenue retention falls below the neutral line. Customers won during the period are excluded on purpose, because new logos cannot repair an installed base that leaks.
For the customer cohort a business already had at the start of a period, its recurring revenue is leaking through pure loss: contraction and outright churn among those existing customers, measured before any credit for expansion from the ones who stayed. A book can post healthy net revenue retention while its gross retention erodes, because upsell to a handful of accounts masks churn across the rest of the base. Gross retention is the read that exposes that leak, because it counts only what the business kept, never what it grew.
Revenue is growing but gross margin is declining because contractor costs are growing faster.
Revenue mix is shifting toward lower-margin project work, reducing the blended profitability of the business.
Total labor (payroll + contractors) is consuming too large a share of revenue, and the trajectory is worsening. S14 uses payroll growth rate projections and scenario analysis (hiring freeze impact, payroll cliff timing, revenue-per-FTE trajectory) to assess where labor costs are heading, not just where they are today. For the point-in-time measurement, see S36.
Total operating costs are consuming a growing share of revenue, cost discipline is eroding.
The company's pre-tax profit margin is declining year-over-year, organic profitability is weakening.
Monthly spending is materially above the operating budget.
Revenue generated per employee is declining as headcount grows faster than revenue.
The same expense categories are over budget across consecutive months, indicating structural, not one-time, variance.
The current-month labor-to-revenue ratio is above the [redacted]% threshold, a point-in-time measurement of labor cost pressure.
At current payroll growth rates, payroll will hit the [redacted]% critical ceiling within [redacted] months.
A recurring-revenue business is buying neither growth nor profit: the sum of its year-over-year revenue growth rate and its EBITDA margin sits materially below the published [redacted] line.
Software and subscription spending is growing faster than revenue, or there are overlapping tools in the same category.
A single vendor accounts for [redacted]%+ of total spend, creating supply-chain concentration risk.
The business has available debt capacity it's not using while cash is getting tight.
The business needs more capacity (utilization is high) but doesn't have enough cash buffer to safely add a headcount.
The company is spending too little on capital assets relative to its revenue, risking deterioration of its delivery infrastructure.
Cash is tightening and DPO is rising, creating a window to renegotiate vendor terms from a position of relative strength before vendors notice the strain.
The business does not generate enough net operating income to cover its annual debt service (scheduled principal plus interest) with the cushion a commercial lender requires. Debt service coverage is the ratio lenders write into loan covenants, and a reading below the covenant line is a technical breach: the lender can call the loan, freeze the facility, or reprice it, regardless of whether payments are current.
Core operating profitability, EBITDA over revenue, is either sitting below a thin-margin floor or compressing materially year-over-year. EBITDA margin isolates how much of each revenue dollar the business keeps as operating profit before financing costs, taxes, and non-cash charges. A margin that is thin, or that is falling fast, means the business is losing the cushion it needs to absorb a cost shock, service debt, or fund reinvestment.
whether the cash a business is sitting on includes money that was never its own to spend, and whether the balance of that money is growing. Two legs. Leg A reads tax set-aside COVERAGE: an implied income-tax reserve (the disclosed [redacted]% assumption applied to year-to-date pretax profit, net of income tax already paid or accrued, floored at zero) plus the payroll-liabilities and sales-tax-payable balances, measured against unrestricted cash. Leg B reads trust-fund liability BUILD-UP: payroll and sales-tax liabilities rising across consecutive closes by a material amount.
a subscription or prepaid business is drawing down its deferred-revenue liability faster than new bookings and collections replenish it. Deferred revenue is cash a business has already collected for goods or services it has not yet delivered; recognizing that revenue draws the balance back down. When the balance falls across consecutive closes, recognized revenue is being funded by previously-collected cash rather than by new sales. This is a forward liquidity warning: the recognized revenue looks healthy today, but a materially declining deferred-revenue balance means part of that revenue is a drawdown of a finite reserve, and when the reserve runs out, recognized revenue steps down to whatever the current bookings pace supports.
whether what the owner takes out of the business is consistent with what the business generates, read in both directions. Leg A (over-extraction) fires when owner cash out (salary, guaranteed payments, draws, and distributions over a trailing window) runs ahead of the operating cash flow that funds it AND owner equity declined across the same window, so the business is being decapitalized while the profit and loss statement still looks acceptable. Leg B (under-payment) fires when the working owner is paid materially less than the company pays for comparable labor it hires, so reported profit is absorbing an unpriced labor subsidy and the margin the owner manages against is overstated. This is a READ of a relationship between two numbers already in the books: it does NOT opine on reasonable compensation, perform the IRS reasonable-compensation analysis, price equity, value the business, or judge whether a distribution was lawful. The full specification, primitives, and the derivation of every constant live in Part [redacted] under "Owner-Pay Sustainability"; this section is the base-signal summary.
Revenue is growing but DSO is deteriorating, the company is effectively financing its own clients' cash cycles.
Margins are compressing while utilization is already maxed, there is no operational lever to escape the problem.
Multiple constraints are converging simultaneously, the business is losing the freedom to make strategic moves.
A combination of immediate liquidity risks creates a scenario where multiple failure modes can activate simultaneously.
Payroll is growing faster than revenue while hiring capacity is already constrained.
Revenue per unit is falling while costs rise, pricing hasn't kept pace with cost structure.
The most severe composite, cascading liquidity failure, payroll cliff, and multiple stress scenario failures all converging.
Revenue is growing but each new dollar of growth destroys value, the growth itself is making the business worse.
Fixed-asset stock relative to annual revenue falls outside the healthy PS benchmark band, signaling either under-investment in delivery infrastructure or capital tied up in assets relative to revenue. C9 is a balance-sheet stock variant of the same investment-adequacy concept S32 measures on a P&L flow basis.
Multiple distinct cost-efficiency problems are firing simultaneously, indicating that cost-per-output is deteriorating across several surfaces at once rather than as a single isolated issue. C10 is a multi-signal aggregation; it reads the firing state of four base signals and fires when at least [redacted] of them are active.
Every signal traces to a published methodology, and an independent, fail-closed verifier re-runs the engine every night before anything ships. See the nightly proof page for the standing coverage guarantee, or how verification works for the path a number takes before it reaches a founder.
Prose fields are free of digits (methodology tags excepted), word-form magnitudes, and constant names; identifiers, enums, and constant names are shape-validated; formula, severity tiers, exact threshold values, manual-calculation steps, precise applicability blocks, and full source citations are omitted, leaving only authored qualitative summaries.