This page is access-controlled. Please enter the password you were sent to continue.
Daylit is the behavioral and conversational surface that sits above Sage Intacct's AR module and Sage AR Automation's cadence engine — closing the five gaps the rule-and-template architecture cannot naturally grow into, while Intacct stays the structured source of truth.
Daylit has been in market since 2023, serving AR teams across services, software, distribution, and professional-services verticals. It was built from the first release on the pattern Sage's own February 2026 messaging has now named — working in the background, keeping decision-making with the rep, every action visible, tuned for accuracy on finance-specific tasks.
Pre-drafted reminders and follow-ups generated per-customer, written in the rep's voice, queued for approval, and sent through the rep's actual mailbox. Templates become voice anchors and safety rails — the text that ships varies per customer context.
Every customer reply classified in real time into one of ten canonical AR actions, extracted into structured fields, and auto-logged as the corresponding action in the ERP. Full visibility, correction, and reversal on a dedicated activity feed.
Behavioral intelligence that watches payment patterns, invoice-age distributions, customer health, and engagement — and recommends cadence changes the rule engine cannot see. Every decision is a trace joined to the outcome that followed.
The modern AR stack separates into three layers. Daylit operates on the third — and writes back to the other two through documented APIs. The architecture is partnership, not replacement.
Five distinct lanes where the rule-and-template architecture stops and customers keep walking. Each is documented with verbatim G2 voice. Presented in order of market opportunity — largest whitespace first.
Every major AR automation product is architected around the outbound motion. Templates go out on a cadence. Replies come back, and a human reads them. The productivity multiplier that has defined AR automation marketing for fifteen years lives entirely in one direction.
The inbound side is where AR teams actually spend most of their time. "We never received the invoice." "We disputed this in March." "Check is on the way, number 4412." None of this happens in the automation. All of it happens in the AR rep's inbox, by hand — as the cost center that the outbound automation was supposed to eliminate.
There is no field in the DUNNINGDEFINITION schema for customer behavior. A customer who has paid every one of their last forty invoices within seven days is treated, at day thirty-one past due, identically to a customer who has slipped three times in six months.
A competent credit manager treats them differently. The product treats them the same. Intacct's R1 2026 release now exposes customer health insights on the customer record — the raw material is present. What is missing is the layer above the rule engine that modulates selection, timing, and intensity based on what the customer's behavior is actually telling the system.
The business cost of system-address send is not that customers do not receive the email — they receive it. The cost is that the customer recognises the email as not from a person, and adjusts accordingly. They treat it as a notification, not as correspondence that requires a reply.
The open-rate, reply-rate, and days-to-payment gap between human-inbox send and system-address send is the single most-replicated finding in AR automation research of the last five years. Sage's current architecture is structurally unable to address it — the system address is baked into the send path.
Intacct's dunning templates are, by architectural design, static. The tone is uniform; the specificity is none; the variation is zero. Customers often do not articulate that the copy sounds like a template — they articulate it indirectly, through complaints about volume, about clunkiness, about "8 e-mails in a row."
2026 is the first year where generated copy can reliably read as though a person composed it, grounded in this customer's specific context and written in the rep's own voice. Sage's February 2026 Sole Trader release named the standard: "human-first AI working quietly in the background."
Collections case management is capable for the use case it was designed for — an AR clerk tracking cases they themselves are working. Where it stops is at the shared workflow. Disputes in mid-market typically are not AR's to resolve; they are Sales Ops', Customer Success', or Operations'. AR's job is to make sure the dispute does not slip, does not cause the cadence to fire inappropriately, and does not disappear from anyone's queue.
A keyword filter for "dispute" returns a single review out of roughly ninety in the responsive Intacct corpus. That is not because disputes do not happen — it is because the product does not foreground the workflow. The absence of the word is the evidence that the workflow is absent.
An AR manager at a $180M mid-market software company, Tuesday morning. Thirty-three minutes of oversight replaces a full morning of inbox reading, invoice lookup, case creation, and manual send. The controller's rule engine is exactly as it was. What has changed is the layer above it.
Overnight and first-thing-Tuesday, 17 customer replies have landed on her own Outlook mailbox. All 17 have been classified. Eleven promise-to-pays are already logged in Intacct with the cadence paused to the stated pay dates and verification follow-ups scheduled. She scans the eleven, spots that two customers specified pay dates Daylit extracted as the wrong calendar day — a "next Friday" read as this Friday — and corrects both in-line. The PTP dates update in Intacct and the follow-ups reschedule. The other nine are clean.
Three replies were disputes. All three have Collections cases open in Intacct with the extracted reasons, and two are routed to the account manager with a 24-hour SLA — pattern-matched to billing-address changes the team has resolved cleanly before. She reads the routing confirmations and moves on. The third dispute is flagged in the feed as low-confidence-escalation because the extracted reason ("contract amendment not reflected") does not match a prior pattern for this customer, and the disputed amount is $42,000 on a renewal invoice. Daylit opened the case and paused the cadence but did not auto-route. She reviews, assigns it to the VP of Finance, and adds a note. The CFO's dashboard will show it by lunch.
Two replies were remit confirmations; Daylit has set up the match-on-arrival against expected payment amounts. One reply was from a customer saying their AP lead is new and asking who to contact going forward. Daylit updated the contact on the customer record and drafted a friendly reply in the rep's voice, queued in her outbound queue for approval. She approves it with the draft as written.
She opens the outbound queue. Daylit has drafted 23 messages for the day — a mix of first notices, second notices, and two behavior-flagged escalations for customers whose payment patterns have shifted in the last sixty days. Each draft reads like she wrote it: her opening ("Hi Sam — quick one on INV-4412"), her typical bridge, her closing ("happy to get on a quick call if easier"). She scans all 23, edits three, approves 20 in a batch. They send from her Outlook, thread with the customer's prior mail.
She has reviewed the night's inbound activity, corrected two extraction errors, escalated one anomalous dispute, and sent the day's outbound queue. On a rule-based system, this would have been a full morning's work. Here, the actions have mostly already happened. Her role was oversight, correction, and judgment on the cases that needed her.
Five non-goals define the edge of the relationship. Everything Daylit does writes back to Intacct through documented APIs. If Daylit disappeared tomorrow, the customer's AR history would be intact.
No invoice storage. No customer master. No payment posting. No aging bucket management. No GL replication. Those are Intacct's.
Card acceptance, ACH, embedded Fortis, and Sage AR Automation's customer-facing portal are all outside Daylit's surface. Payment links in outbound email point to the customer's chosen Sage-operated surface.
Reads them. Selects among them with behavioral logic. Does not create or edit. Creating a dunning definition remains a controller action inside Intacct, under the controller's audit trail.
Outbound waits for approval before it leaves the mailbox. Inbound actions fire automatically but are fully visible, correctable, and reversible. Low-confidence classifications are flagged for attention before routing.
Every structured action writes back to Intacct so that the ERP remains the authoritative record. If Daylit disappeared tomorrow, the customer's AR history would be intact in Intacct. That commitment is what makes the partnership model coherent — Daylit gets stronger as the ERP gets stronger, and the ERP does not have to cede ownership of anything it already owns.
Daylit as a named component of Sage's AR stack. Integrated natively with Intacct, coordinated alongside Sage AR Automation, sold through Sage's channel. Revenue shared under a negotiated economic structure. Roadmaps coordinated quarterly. Go-to-market investment on both sides.
Mid-market Intacct customers selected jointly from Sage's book. Each chosen for fit with the architecture above. Outcome metrics in CFO vocabulary — DSO, days-to-payment, productivity gain, payment-rate lift, dispute resolution time. Baselines from the customer; targets agreed in the design session.
Daylit-on-Sage or co-branded with Sage's existing AR products. Integrated natively with Intacct via the public API. Coordinated with Sage AR Automation. Sold through Sage's channel as the AR upgrade for customers who want the behavioral and conversational layer above their existing dunning configuration.
Roughly 90 days to commercial shape. 6 to 9 months to full integration and channel enablement. An explicit review at the end of the second quarter post-launch, where both parties assess the evidence and decide what the next phase of the partnership looks like.