Skip to main content

Documentation Index

Fetch the complete documentation index at: https://help.daylit.com/llms.txt

Use this file to discover all available pages before exploring further.

For every open invoice, Daylit’s AI generates a predicted payment date — an evidence-backed estimate of when you should expect to receive payment. Predictions are anchored to real data about the customer’s payment behavior, the age of the invoice, and what has been communicated, giving your AR team and finance leadership a more reliable basis for cash flow planning than averages or assumptions alone.

What payment predictions are

A payment prediction is a structured forecast that answers the question: “Based on everything Daylit knows about this invoice and this customer, when will this invoice most likely be paid?” The prediction includes a primary predicted date, a confidence score, the reasoning behind the estimate, the data sources used, and alternative scenarios with their own probabilities. It is generated by the AI each time the invoice’s signals are refreshed — it is not a static field.

What factors the AI considers

Daylit’s prediction model draws on multiple data sources for each invoice:
  • Customer payment history — How long this customer typically takes to pay relative to their due date. Whether they are a reliable payer, a chronic late payer, or someone whose behavior has been shifting recently.
  • Invoice age and aging bucket — How many days past due the invoice currently is, and how that compares to the customer’s typical pattern at the same age.
  • Customer behavior signals — AI-detected patterns such as payment slowdown, ghosting, promise-to-pay, or payment imminent that adjust the prediction up or down.
  • Communication history — Whether the customer has made a commitment to pay by a specific date, whether they are responsive to outreach, and the overall tone of recent conversations.
  • Invoice characteristics — Invoice amount, payment terms, and whether this invoice has custom terms that differ from the customer’s default.
The AI combines these factors and chooses a primary basis for the prediction:
BasisWhat it means
PromiseThe customer made an explicit commitment to pay by a stated date in the communication thread.
HistoricalThe prediction is derived from the customer’s past payment timing patterns.
PatternStatistical or behavioral patterns — for example, a consistent 15-day delay beyond terms — are used as the primary anchor.
JudgementThe AI made an inference when data is limited or conflicting, such as for a new customer with few closed invoices.

Where to see predictions

Payment predictions appear on the invoice detail page under the AI Insights tab.
1

Open the invoice detail page

Click an invoice number in the Invoices table.
2

Select the AI Insights tab

Click the AI Insights tab in the detail page navigation.
3

Find the Payment Prediction section

Scroll to the Payment Prediction section. The primary predicted date is displayed prominently alongside the confidence score.

The prediction fields

Each payment prediction contains the following fields:
The single most likely payment date in YYYY-MM-DD format. This is null when there is insufficient data to make a reliable estimate — for example, a brand-new customer with no payment history and no communication thread.
A score from 0 to 1 (displayed as a percentage) expressing how certain the AI is about the predicted date. A score of 0.85 or higher indicates high confidence; below 0.5 indicates the prediction is speculative and should be treated as a rough guide only.
A clause-style explanation of why the AI arrived at this prediction. For example: “Customer averages 12 days beyond Net 30; no response to last two reminders; no promise on file.” This is the AI’s plain-language rationale, not marketing copy.
The specific data inputs the AI used: for example, payment_history, communication_thread, or customer_signal. Knowing which sources were used helps you assess how much weight to give the prediction.
Up to two or three other possible payment dates with their own probability scores and a brief basis clause. For example, an optimistic scenario where the customer pays in response to the next reminder, and a pessimistic scenario where the account goes silent for 30 more days. Use these to understand the range of outcomes rather than treating the primary date as a certainty.
The primary driver of the prediction: promise, historical, pattern, or judgement. A promise-based prediction is the most reliable because it is anchored to a specific commitment made by the customer. A judgement-based prediction carries more uncertainty.

Using predictions for cash flow planning

Payment predictions are most useful when you treat the full set of predicted dates — not just the primary date — as a probability distribution. For individual invoices: Use the predicted date and confidence score together. A high-confidence prediction of a date three days from now can inform a near-term cash position. A low-confidence prediction on a 90+-day-overdue invoice tells you the outcome is genuinely uncertain and the account may need escalation. For the portfolio: When predictions are aggregated across many open invoices, you get a forward-looking view of expected cash collections by week or month. This is surfaced in the Cash flow dashboard. For prioritization: If an invoice has a low predicted confidence and a basis of “judgement,” that is a signal to invest in outreach — more communication data will improve both the prediction accuracy and the underlying collection outcome.
When a customer makes a commitment to pay by a specific date, log it as a note or ensure it appears in a synced email. The AI will detect the promise and switch the prediction basis to “promise,” which significantly increases confidence.

How predictions improve over time

Payment predictions are not static. They are recalculated each time the AI refreshes a customer or invoice’s insights — which happens on an automatic schedule as new payment data and communications sync into Daylit. Several things make predictions more accurate over time:
  • More closed invoices — Each paid invoice adds to the customer’s payment history, giving the model a stronger behavioral baseline.
  • More communication data — Emails, call notes, and promise-to-pay records give the AI concrete anchor points instead of having to infer purely from historical patterns.
  • Feedback signals — When the AI predicts a date and the invoice is paid earlier or later, that outcome feeds back into the model’s calibration.
For customers new to your portfolio or new to Daylit, expect lower confidence scores and “judgement” basis predictions until enough data accumulates. The predictions become materially more reliable after three to five closed invoice cycles with a customer.

Managing invoices

View, filter, and act on your open invoices, including sending reminders and recording payments.

Cash flow dashboard

See predicted cash collections aggregated across your full AR portfolio.

Customer AI signals and insights

Understand the customer-level signals and conversation summaries that feed into invoice predictions.

AI insights

Learn how Daylit’s AI generates predictions, signals, and strategy plans across your portfolio.