Triage workflow — the four-hour window and the per-category SLA.

The premise — speed is not a virtue, it is a math problem

Most operators treat reply speed as a posture. The disciplined ones treat it as a per-touch decay curve. A positive reply at hour zero is not the same asset twenty-four hours later — it is a fundamentally different opportunity with a lower terminal value. Reply triage is a perishable-inventory problem, and most outbound teams have no operational concept of perishability.

The empirical pattern: a positive reply touched within the first hour converts to a held meeting at 60% or higher. The same reply, from the same prospect, touched at hour 24, converts at approximately 25%. At hour 72, conversion falls to roughly 10%, and below that the curve flattens into background noise. The prospect has not changed. The intent has not vanished. What has decayed is the recency of the trigger — the original message no longer occupies any working memory, and the responder has to re-establish context cold.

The four-hour window is not a target. It is the empirical inflection point at which the curve begins to fall meaningfully. An operator who holds first-touch-on-positive-replies inside four hours during business hours, every business day, has resolved the single most expensive operational gap in their outbound stack. An operator who cannot, has not.

The per-touch decay curve

The numbers are worth stating concretely because the operational decisions downstream — staffing, on-call rotation, automation investment — all calibrate against them.

The fall from 60% to 25% across the first day is the single largest unit-economic lever in the entire reply-handling stack. A team that moves its median first-touch from hour 18 to hour 3 has, in a single operational change, roughly doubled the meeting throughput of the same campaign volume — with no incremental cost upstream, no new domains, no new sequences, no new copy.

The four-hour window — what it actually means

The four-hour SLA is business-hours-aware, time-zone-aware, and on-call-rotation-backed. Stating it as a flat "reply within four hours of receipt" is the most common misreading — a reply landing at 2am local time should not trigger a four-hour clock that expires at 6am.

The operationally correct definition: a reply received during the prospect's business hours has a four-hour-of-prospect-business-hours clock attached. A reply received outside business hours has a clock that begins at the next business-hours-window open. The relevant time zone is the prospect's, not the responder's — a Pacific-time responder touching an East Coast reply has a clock that closes three hours earlier than they instinctively expect.

The team that ignores this produces a dashboard reading "average response time 3.8 hours" while their East Coast prospects experience 7 hours. The dashboard is not wrong. The dashboard is measuring the wrong thing.

Per-category triage urgency

Treating all replies as equally urgent is the second most common failure mode. The five reply categories (Chapter 1) carry materially different per-category SLAs because the per-category decay curves are materially different.

The auto-responder SLA is not zero because the work is small: parse the return-by date, update the contact, requeue the sequence. The negative SLA is not zero because the work — suppression-list update — has no per-prospect urgency, only an aggregate compliance requirement before the next sequence send.

The positive SLA, in contrast, is everything. Every operational design downstream — the routing channel, the on-call rotation, the AE handoff — is a mechanism for protecting the one-hour-on-positive-replies number.

The on-call rotation pattern

Human triage requires human availability. This is the operational reality that operators with otherwise sophisticated outbound stacks routinely fail to plan for. The team-size-to-rotation pattern, in the configurations we have observed:

• 1–2 person team. No rotation; one person is always on. The SLA holds during their working hours and silently breaks outside them. The acceptable posture is to disable sends in time zones the team cannot cover, not to pretend the coverage exists.
• 3–5 person team. Daily rotation; one person owns the inbox each day, with a backup. The relevant discipline is that the on-call has no other deep-work assignments — interrupt-driven work and creative work do not share a person without one degrading.
• 6+ person team. Weekly rotation with split-shift coverage for the relevant time zones. A published schedule with named backups and an explicit escalation path for time-zone hand-offs.

The error mode is the team that announces a four-hour SLA but has no named on-call. The SLA holds opportunistically and the variance across days is enormous — median of 2 hours and a 95th-percentile of 38 hours. The 95th percentile is the number that matters; the high-intent reply that produced the 38-hour wait was a $40k opportunity.

The AE-vs-SDR triage split

The routing architecture (Chapter 2) determines which seat receives each category. The triage discipline reinforces it.

• Positive replies → AE. The meeting-booking handoff is direct and immediate. Routing through an SDR for qualification first adds, in practice, four to eighteen hours of latency and a measurable drop in meeting-booking rate. The exception is a high-volume inside-sales motion where the SDR has full meeting-booking authority.
• Objections → SDR. The per-objection response library (Chapter 4) lives with the SDR. An AE responding to objections produces both worse responses and a downstream pattern in which the AE's working hours are consumed by reply-triage rather than discovery and close.
• Soft-pass → SDR. Nurture assignment (Chapter 6) is SDR work. Tag the contact, set the next touch date, return to the queue.

Multi-channel triage

A reply does not always arrive on the channel that sent the original message. The mature motion produces replies across email, LinkedIn, and inbound phone callbacks, and the triage discipline has to span all three.

• Email replies. The default channel. The four-hour SLA applies directly.
• LinkedIn replies. The timeline is shorter, not longer — message activity is itself a high-recency signal, and a reply unanswered for 24 hours competes with active feed-scrolling attention. Treat LinkedIn positive replies with a two-hour SLA, not four.
• Phone callbacks. A returned voicemail or inbound call is the strongest intent signal in the reply stack. The SLA is the next available calendar opening of the recipient, not a clock-based number. The failure mode is the AE returning a callback four hours later when the prospect has already been pitched by a competitor.

Response-time SLA enforcement

A per-category SLA without a metric and a dashboard is a slogan, not a discipline. The minimum-viable dashboard captures: time-to-first-touch per reply, broken out by category, displayed as a distribution rather than an average. The 95th-percentile is the operationally relevant number; the average is misleading.

The review cadence is weekly. The team meeting that opens with the prior week's triage distribution — number of replies per category, percentage within SLA, identification of the specific replies that breached, the named cause of each breach — is the meeting that holds the discipline. The team meeting that opens with a green "average response time" number is the meeting that performs discipline without producing it.

Per-segment triage prioritization

When two positive replies land at the same minute, which one gets touched first matters. The prioritization rule we observe operating cleanly in production:

• Named accounts — the high-ACV, strategically tagged target list — go first, every time, with no exception.
• ICP-fit prospects — the prospects matching the documented ideal-customer profile but not on the named list — go second.
• Low-tier prospects — replies from outside the ICP, often from sequences run as broader experiments — go last and may legitimately fall outside the four-hour SLA without operational concern.

The operator who treats all positive replies as identical produces, by random ordering, a dashboard in which the named-account 95th-percentile and the low-tier 95th-percentile look similar. The named-account 95th-percentile is the only one that affects revenue at scale; the low-tier number is operational vanity.

The lunch-and-weekend coverage problem

The four-hour SLA breaks predictably at the lunch hour, on Fridays after 4pm, and across the entire weekend. These are the gaps where the dashboards lie and the revenue leaks.

The lunch problem resolves with rotation discipline: the on-call does not take lunch unattended. The Friday-afternoon problem resolves with an explicit hand-off through the prospect's business day, not the responder's. The weekend problem is harder. A positive reply landing at 4pm Friday Eastern, without weekend coverage, has its clock running for 64 hours before a human touches it — per the decay curve, roughly a tenth of the asset it was when it landed.

The mature options: (a) weekend on-call with a reduced SLA — eight hours rather than four — explicitly published; (b) automated acknowledgement that establishes the conversation and buys time; (c) pausing Friday-afternoon sends in time zones the team cannot cover over the weekend. The unmature option, which we see most frequently, is to do none of these and accept the weekend gap as inevitable while continuing to send into it.

The volume-overload pattern

Reply volume scales with campaign volume and occasionally spikes above the team's steady-state triage capacity. The operational requirement is a load-shedding policy decided in advance, not improvised under load.

The order of preservation, when capacity is exceeded: named-account positive replies first, ICP-fit positive replies second, named-account objections third, ICP-fit objections fourth. Soft-pass replies, auto-responders, and negative replies are shed under load — they accumulate in a backlog queue and are processed in batch when capacity returns. The acceptable backlog age for these categories is 72 hours; beyond that, the negative-reply backlog risks compliance exposure on the next sequence send.

A team operating at 80%-of-capacity steady-state has the headroom to absorb spikes. A team operating at 110% has a four-hour-SLA dashboard that reads green on the average and breaches catastrophically at the 95th-percentile every time campaign volume increases. The capacity calculation is upstream of the triage discipline.

The AI-augmented triage pattern

An LLM in the triage loop, configured to read inbound replies and draft response candidates for human approval, is the highest-leverage operational change we have observed in the last twelve months. Every inbound reply is auto-classified and a response is drafted; the on-call reviews, edits, and sends. Median time-to-first-touch falls from 90 minutes to 15. The four-hour SLA holds at the 99th-percentile, not just the median.

The cost is non-trivial — token spend per reply runs in the low cents, response-quality QA requires a weekly review of sampled drafts, and the team has to develop the discipline of editing rather than rubber-stamping. The naive deployment, in which the LLM both drafts and sends without review, produces a measurable conversion penalty — the drafts are competent but generic, and the prospect can tell. The augmentation pattern, where the human approves every send, captures most of the speed benefit without the quality penalty.

The time-to-first-touch metric

The metric definition matters. Time-to-first-touch is measured from the timestamp of the inbound reply receipt to the timestamp of the outbound human-authored response, broken out by reply category and by named-vs-ICP-vs-low-tier segment. The aggregate distribution displayed on the dashboard:

• Median (p50) — operational health
• p75 — the visible variance
• p95 — the breach indicator
• p99 — the named-account incident detector

A team holding p95 inside the per-category SLA is operating with discipline. A team holding only p50 inside the SLA is operating on average and breaching specifically. The dashboard that surfaces only p50 is the dashboard that produces complacent quarterly reviews.

Common operator failures observed in production

• No on-call coverage. The SLA is announced; no one is named; the four-hour window holds opportunistically. Most common in 3-to-5-person teams.
• No per-category SLA. All replies share a single clock; positives drag through the same queue as auto-responders, producing wasted urgency on low-value categories and missed urgency on high-value ones.
• No metric tracking. The team believes it operates inside the SLA; the dashboard does not exist; the belief is unfalsifiable.
• Treating all replies as equal urgency. The SDR responding to a negative reply within five minutes while a positive waits four hours. The decay curve says the trade-off is roughly twenty-to-one against the SDR.
• No escalation path for after-hours. The reply landing at 7pm Friday Pacific has no defined home — not the AE's problem until Monday, not the SDR's, not the on-call's because the rotation ended at 6pm. The reply rots.
• Time-zone confusion. The responder's clock, not the prospect's, drives the dashboard. The team believes it operates at 3-hour median; East Coast prospects experience 7.
• Capacity-blind growth. Campaign volume scales 2x, triage capacity scales 1x, the SLA degrades silently until the quarterly pipeline review surfaces the conversion drop.

Pre-deployment triage checklist

• Per-category SLAs documented (positive 1h, objection 4h, soft-pass 24h, auto-responder 24h, negative 72h)
• On-call rotation published with named owners and named backups
• Time-zone-aware SLA-clock definition documented and implemented in the dashboard
• AE-vs-SDR category split agreed and reflected in the routing rules (Chapter 2)
• Multi-channel coverage — email, LinkedIn, phone — assigned to named owners with channel-specific SLAs
• Time-to-first-touch dashboard live, surfacing p50/p75/p95/p99 by category and segment
• Weekly triage review on the team calendar with a standing agenda
• Weekend and after-hours policy documented — coverage rotation, reduced SLA, or explicit acceptance of the gap
• Load-shedding policy documented for capacity-overload conditions
• Named-account vs ICP-fit vs low-tier segmentation reflected in the dashboard breakdown
• If using AI-augmented triage: human-approval gate enforced, weekly draft-quality QA scheduled

Where this fits

Triage is the operational layer that converts classification (Chapter 1) and routing (Chapter 2) into a held discipline. A correctly classified reply, routed to the correct channel, that no human touches inside the SLA, is operationally equivalent to a misclassified reply routed to nowhere — the conversion outcome is the same. The next two chapters cover what the human does once they touch the reply: objection handling (Chapter 4), which converts active resistance into a conversation, and meeting booking (Chapter 5), which converts a positive reply into a held calendar slot.

The four-hour discipline is the single number the operator should defend with the most institutional resistance. Every other reply-handling investment compounds against it. Without it, every other investment is lower-leverage than fixing it would have been.