# Who answers at 2 AM: designing coverage for the 128 hours

Published 2026-09-15 by Lumion. Guides & Playbooks. https://lumion.ai/blog/after-hours-admissions-coverage

> Most career schools cover 40 of the 168 hours a prospect might inquire. Adding shifts moves the wall. This playbook lays out the coverage grid: what an agent answers alone, what waits for a human, and how Monday morning picks up Friday night.

# Who answers at 2 AM: designing coverage for the 128 hours

*Playbook · Admissions & Enrollment · Speed to lead · All verticals · 17-minute read*

> **The one idea.** Most career schools staff 40 of the 168 hours a week a prospect might reach out. Adding an evening shift moves the wall. Turning on a chatbot flattens every question into the same shallow answer. Coverage is a routing decision: every message gets sorted by what it needs and what it risks, and every hour of the week gets a rule for who answers each kind. Nine cells. This playbook fills them in, step by step, and names what each step takes to run.

**What's in this playbook**

| Section | What you get |
|---|---|
| 1. Friday, 11:40 PM | The story every school is already in |
| 2. The two plans schools reach for | Why the shift plan and the tool plan are both half right |
| 3. The mechanism | Three tiers, three kinds of hours, the coverage grid |
| 4. Two rules for every play | End in an outcome; the agent does what people can't, shouldn't, or don't have time to |
| 5. The prerequisite | One system that owns every channel |
| 6. The five steps | Owner, time cost, done-when, what it requires, and the play that runs it |
| 7. The scoreboard | Three metrics, one self-test, and what a point of yield is worth |
| 8. The caveat | The rule that overrides the rest |
| Companion | The Night Shift Kit: printable grid, sort sheet, scripts, scoreboard |

---

## 1. Friday, 11:40 PM

*The story every school is already in.*

A 31-year-old forklift operator just got off a closing shift, sat down at the kitchen table, and filled out the inquiry form on a CDL school's website. He asked one question in the comments box: does the Class A program run on weekends, because he can't quit his job to train.

What happens next was decided months ago, whether or not anyone at the school remembers deciding it.

### The three fates

#### Same inquiry, three schools

That Friday-night question can enter one of three response systems. Here is what each one means for the prospect at 11:41 PM, and for the coordinator on Monday. Most schools are running the first or second without ever having chosen it.

| Version | What he gets | What the school has Monday morning |
|---|---|---|
| **The shared inbox** | A thank-you email. The inquiry sits until 8:15 AM Monday. | A cold lead who heard back from two other schools, one of them at 11:41 PM Friday. |
| **The chatbot** | A generic welcome and a link to the page he was already on. | Nothing. He closed the tab. |
| **The grid** | A text within a minute: yes, there's a weekend cohort, it starts October 17, here are two Saturday tour times, pick one. He picks one. He's told who will call Monday. | A program, a shift constraint, a Saturday tour, and a first name. The coordinator's call starts in the middle of the conversation. |

He is the typical prospect. Three-quarters of part-time students work, and four in ten of those work full-time hours. The people a career school wants to enroll research at night because night is when they aren't working.

We've written about why this matters: schools ghost students before students ghost schools, and the 128 hours a week most admissions teams don't cover are where most of the ghosting happens. This playbook is the other half. How a school decides who answers at 2 AM, what that system is allowed to say, and what it takes to run.

**Takeaway.** The same inquiry has three fates, and the school chose one months earlier, whether or not anyone remembers choosing.

---

## 2. The two plans schools reach for

*Both are half right, and both make the same mistake about the 128 hours.*

Ask an operator how they handle after-hours inquiries and you'll hear one of two plans: add a shift, or turn on a tool. Both come from a real place. The shift plan is what a school does when it takes the problem seriously and has budget; the tool plan is what it does when it takes the problem seriously and doesn't. This section lays them side by side, then puts the numbers behind the gap they're both trying to close.

### The two plans

#### What each one actually buys

Five things to compare: what the plan is, what it covers, what it costs, where it breaks, and the mistake the two share.

| | The shift plan | The tool plan |
|---|---|---|
| **What it is** | Add an evening rep. Rotate weekend coverage. Give the newest coordinator the 4-to-midnight slot. | Turn on the chatbot. Set up an autoresponder. |
| **What it covers** | A 5-to-midnight shift five days a week covers 35 of the 128 uncovered hours. | All 168, which is the appeal. |
| **What it costs** | Roughly $67,000 in loaded compensation per rep, plus the loneliest shift going to the person most likely to leave. | The prospect's trust. The same shallow answer goes to a question about parking and a question about whether Pell will cover a $14,000 program. |
| **Where it breaks** | Seven in ten admissions coordinators in higher ed have been in the role three years or less. The evening shift does not improve that number. | Prospects can tell the difference between an answer and a deflection. The deflection teaches them the school isn't really there. |
| **The shared mistake** | Treats the 128 hours as one block to cover. | Treats the 128 hours as one block to automate. |

> "We wanted that interaction of the questions. We didn't want it to just be somebody entering their information and then not talking."
> **Michelle LaMantia**, Academy Director, Pivot Point Academy

> **The numbers behind the gap**
>
> - **44 percent** of prospective-student inquiries got no response at all when UPCEA secret-shopped continuing and online education units in 2025. The ones that did waited a median of more than three hours. That is the staffed-hours number.
> - **75 percent** of the questions students asked schools' AI assistants arrived outside business hours in EAB's 2025 sample of 25,000 interactions: 43 percent between 6 and 11 PM, 32 percent overnight.
> - **74 percent** of part-time undergraduates work, and 40 percent of those work 35 hours or more (NCES).
> - **71 percent** of admissions coordinators and counselors have been in their role three years or less; median tenure is two years (CUPA-HR, 2023).

Both plans miss the same thing. The message is the unit. The hour is just where it lands.

**Takeaway.** A shift moves the wall. A chatbot flattens the questions. Both treat 128 hours as one block, and the fix starts with treating the message as the unit instead.

---

## 3. The mechanism: coverage is a routing grid

*Every message has two properties: what it needs and what it risks.*

After-hours admissions coverage is the set of rules a school uses to answer inquiries that arrive when nobody is working: which messages get an instant answer, which get an acknowledgment with a named person and a time, and how the morning team picks up where the night left off.

This is the mechanism the rest of the playbook runs on, so it's worth slowing down. Sort inbound messages by what they need (a published fact, a judgment call, or a person) and by what they risk if answered wrong, and you get three tiers. Sort the week into the hours when staff are at desks, the evening when they might be reachable, and the hours when nobody should be working, and you get three kinds of hours. Cross them and every message has an owner before it arrives.

### The three tiers

#### Sort by what the message needs

The tier a message lands in decides who may answer it and how fast. Read the examples; most schools recognize their own inbox in the first column.

| Tier | What it needs | Examples | The right move at any hour |
|---|---|---|---|
| **Tier one** | A fact the school has already published | Programs, start dates, weekend or evening tracks, tuition and payment options as published, tour times, admissions requirements, location | Answer in seconds. End in a next step the prospect can take right now. |
| **Tier two** | Judgment, but not urgency | Transferring hours, switching days to evenings, restarting after a stop-out, comparing two programs, "am I too old for this," "can I really do this" | Acknowledge in seconds with a name and a real time. One true fact if there is one. Route the thread to a person who knows the school. |
| **Tier three** | A human, and the school saying so | Money beyond published options; Title IV, Pell, or state grant eligibility; aid or loan estimates; whether VA or Chapter 33 benefits will cover it; a prior balance; complaints; anything about a specific student's record | Acknowledge in seconds with a name and a window. No answer, no link, no reassurance about money. |

The encouragement questions sit in tier two on purpose. A single mother asking whether she can really do this deserves a warm reply and one true fact (there is an evening track; there is a student support office), and then a person. A veteran asking whether his Chapter 33 benefits will cover it is asking a money question, and that is tier three no matter how casually it's phrased.

### The three kinds of hours

#### Then place it in time

The three kinds of hours are about who's available, not the clock. A school with Saturday-morning staff has a different staffed block than one that doesn't; the point is to name yours.

| Hours | When | Who's available | What the school owes the prospect |
|---|---|---|---|
| **Staffed** | Your 40, whatever they are | The admissions team at their desks | A substantive reply in minutes, from a person if it needs one |
| **Extended** | 6 to 11 PM | Nobody, unless a coordinator chooses to clear a queue from a phone | Tier one answered; tier two and three acknowledged with a time |
| **Dark** | Overnight, most of the weekend | Nobody, and nobody should be | Tier one answered; tier two and three acknowledged with a real time; nothing initiated |

### The grid

#### Cross the tiers with the hours

Nine cells. Three are the agent's at every hour. Six are staff-owned and open with an acknowledgment. Filling them in is the design work, and the Night Shift Kit has the blank grid.

![The coverage grid](charts/coverage-grid.png)

Three cells are agent-owned: a written answer goes out at any hour and ends in a next step. Six are staff-owned: the agent's only job is the acknowledgment and the task, and a person does the rest. That split is the whole design. It puts the fast answers where the safe answers are.

It also tells you what the grid needs in order to exist. Something has to read each message and decide which tier it's in, at 2 AM, without a person. Something has to know the difference between a reply and an outreach. Something has to hand a thread to a named person with a clock on it. A shared inbox and three personal cell phones can't do any of that, which is why most schools are running one of the two half-right plans without ever having chosen it.

**Takeaway.** Nine cells. Three belong to the agent at every hour. Six belong to staff and open with an acknowledgment that names a person and a time.

---

## 4. Two rules for every play

*The grid says who answers. Two rules say how.*

The grid is a routing decision. It doesn't say what a good reply looks like or where the agent stops and a person starts, and those two questions are where most AI deployments in admissions go wrong. Rule one is about the reply: it has to end in an outcome. Rule two is about the line between the agent and the team: the agent takes the work people can't, shouldn't, or don't have time to do, and nothing else.

### Rule one: every play ends in an outcome

A reply that ends with "let us know if you have any other questions" is a deflection with better manners. The prospect asked one thing and now has to decide what to do next, alone, at midnight. Most won't.

Every play ends in a next step the prospect takes inside the conversation, or a commitment the school makes with a name and a time on it. That is what separates a conversation from an autoresponder.

#### Open-ended, and the same reply with an outcome

Five replies most schools send today, and the version of each that ends in something the prospect can do.

| Open-ended | Ends in an outcome |
|---|---|
| "Here's a link to our programs page." | "The weekend Class A cohort starts October 17. Two Saturday tour times this month, pick one." |
| "Someone will reach out soon." | "Marcus will text you Monday before noon with what we'd need from you." |
| "Great question! Our financial aid office can help with that." | "Dana runs financial aid. She'll call you Monday between 9 and 10. Want a different time?" |
| "Thanks for calling. Please leave a message." | "Sorry we missed you. What can I help with tonight, and what's the best way to reach you tomorrow?" |
| "Feel free to apply anytime." | "Here's the application. It takes about ten minutes. Want me to save your spot on the October 17 start while you finish it?" |

This is also why the agent has to be built on the school. A generic bot can only end in a link, because it doesn't know the cohort dates, the tracks, the tuition options, or who Marcus is. An agent that has read the school's site and documents can ask the qualifying question ("which shift do you work?"), answer in the school's own terms, and close on a real slot. The outcome rule is the test: if a tool can't end a tier-one conversation in a booked tour, a chosen callback window, or a started application, it isn't ready for the dark hours.

### Rule two: the agent does what people can't, shouldn't, or don't have time to

Career schools run on the human relationship. The coordinator who remembers a prospect's shift, the financial aid director who walks a veteran through his benefits, the instructor who calls a student who stopped showing up. Nothing in this playbook replaces any of that, and a school that uses an agent to replace it will lose the students it's built to serve.

The grid works because it draws the line in the right place. The agent takes the work people can't do (answer at 2 AM), shouldn't do (quote aid amounts from memory, text from a personal cell after quiet hours), or want to do but don't have the capacity for (respond to every inquiry in under ten minutes, log every send, keep every promise on a clock). People take everything that needs judgment, trust, or a voice.

#### What the agent takes, and what people keep

Each row is one piece of after-hours work. The left column is why the agent takes it; the right column is why people keep the work beside it.

| The agent takes | Because people | People keep | Because the agent |
|---|---|---|---|
| The 2 AM answer to a published fact | Can't be there | The tier-two answer that needs judgment | Doesn't know this school's exceptions |
| The acknowledgment with a name and a time | Can't promise a time they don't know they can keep | The call to the veteran about his benefits | Must never touch eligibility or money |
| The missed-call text-back, in a minute | Shouldn't be answering a work line at midnight | The conversation with the single mother who isn't sure she can do this | Can offer one true fact; only a person can offer belief |
| The morning digest, sorted by tier | Don't have time to reconstruct the weekend | The Monday call that starts in the middle | Set it up to be that easy |
| The log of every send | Shouldn't have to remember what they promised | The decision about what the agent may say at all | Runs whatever line the school draws, and no further |

The relationship is symbiotic when it's designed and corrosive when it isn't. Designed means the school chose, cell by cell, what the agent does alone and where a person steps in, and the agent's job in every staff-owned cell is to make the person's first touch easier than it would have been. Undesigned means the agent answers whatever it can and the school hopes for the best, which is the chatbot plan with a better model behind it.

![One inquiry, one weekend, two kinds of touch](charts/handoff-cadence.png)

*Friday 11:40 PM to Monday 9:05 AM for the forklift operator's inquiry. Every agent touch is a fact or a commitment. Every human touch starts where the agent left off.*

**Takeaway.** Every play ends in something the prospect can do or someone the school has committed to. The agent takes what people can't, shouldn't, or don't have time for, and people keep everything that needs judgment, trust, or a voice.

---

## 5. The prerequisite: one system owns every channel

*Before the tiers, before the scripts, before anything else on this list.*

This is the prerequisite most schools skip because it isn't interesting: every way a prospect can reach the school has to land in one place, on one record, with every send logged. Without it, nobody can sort last month's inquiries because nobody has them, nobody can hand off an overnight thread because it's on a rep's phone, and nobody can prove a text went out inside quiet hours. Use the check below to find out whether the grid can exist at your school yet.

One record per prospect is the part schools underestimate. A form on Friday night, an Instagram DM on Saturday, and a call to the campus line on Sunday are one person, and the system has to match them on phone and email before Monday. Otherwise the coordinator calls Devon three times from three lead records, and Devon concludes the school doesn't know who he is.

### The prerequisite check

#### Can the grid exist at your school yet?

| If this is true today | The grid can |
|---|---|
| Every channel lands in one inbox on one student record, deduplicated on phone and email | Route by tier, hand off overnight threads by name, and measure the slowest cell |
| Replies go out from numbers the school owns | Prove every send stayed inside quiet hours, and see every promise a rep made |
| The phone line and the text line are the same system | Turn a missed call at 11:40 PM into a text at 11:41 PM |

| If this is true today | The grid can't exist yet |
|---|---|
| Reps reply from personal cells | The school can't see the thread, can't measure it, can't hand it off, and can't prove what was promised |
| Chat, forms, texts, and DMs go to different inboxes | Nobody can sort last month's inquiries because nobody has them, and one prospect becomes three records |
| The phone rings to voicemail after 5 PM | The highest-intent after-hours channel is the slowest one |

Any reply from a personal phone is outside the grid. If that's true at your school, this comes first, before the tiers, before the scripts, before anything else on this list. On Lumion this is the Unified Inbox: chat, text, email, and the school's phone line on one student record, matched on phone and email, with every message, every call summary, and every action the agent took visible in the thread.

**Takeaway.** If a reply can leave from a personal phone, the grid can't exist yet. One inbox, one record per prospect, every channel. This comes first.

---

## 6. The five steps

*Each step has an owner, a time cost, a done-when, and a line for what it requires.*

Five steps, in order. The first three are design work: what the agent knows, what last month's inquiries were, and what the overnight rule is. The last two are operating rules: what happens in the evening and what happens the next morning. Each step opens with four lines. Read the last one first; it tells you whether the step is a decision your team can make this week or a capability you need to have before you start.

### Step 1. Approve what the agent knows. Set what it never says.

| | |
|---|---|
| **Owner** | Admissions director, with whoever owns financial aid language |
| **Time** | About an hour to review; ten minutes a month after |
| **Done when** | The agent's knowledge has been read and approved by a person, and the never-answer list is set |
| **Requires** | An agent that builds its knowledge from your website and documents, and a review step before it goes live |

This is where chatbots and agents part ways. A chatbot arrives empty, and someone has to type every answer into it, which is why so few ever get finished. A purpose-built agent arrives having read the school's site: programs, tracks, start dates, tuition and payment options as published, tour times, requirements. Lumion's web agent does this before launch, from the school's own pages and the documents the school uploads (the program sheets, the cohort calendar, the enrollment agreement). The school's job is to read what it learned, correct what's stale, and approve it.

Then the second list, which no scrape produces because it's a decision:

**The never-answer list:**
- Eligibility determinations
- Aid amounts, or anything that sounds like one
- Discounts, waivers, or anything that sounds like a promise about money
- Complaints
- Anything about a specific student's record

The agent's defaults already cover most of it: it never quotes a price it wasn't given, never negotiates terms, and always discloses that it's an AI. The school tightens from there. Ten minutes with the financial aid director on this list is what makes every fast answer at 2 AM a safe one.

> **What a chatbot makes you do instead.** Write out every answer to every question by hand, keep it current by hand, and hope the bot matches the question to the right one. The first two are why most school chatbots answer three questions and deflect the rest.

**Step takeaway.** An agent built from the school's site arrives knowing the programs. The school's hour of work is reading what it learned and drawing the line it never crosses.

### Step 2. Sort last month's inquiries into the three tiers

| | |
|---|---|
| **Owner** | One coordinator |
| **Time** | An afternoon the first time. Under an hour if the inbox already tags them. |
| **Done when** | You have three percentages and a list of your fifteen most common tier-one questions |
| **Requires** | Every inbound message from the last 30 days in one place, with the hour it arrived |

Pull every inbound message from the last 30 days, all channels, and sort each one. Most schools find the large majority is tier one, and that the same fifteen or twenty questions account for most of it. That is the number that makes the rest of the grid cheap. If most messages have a published answer, the dark hours are mostly a lookup problem. The remainder is where the people go.

Tally by channel and by hour while you're in there. The channel a prospect reaches for at 11 PM is the one that has to answer, and the hours tell you where your wall is.

Write the percentages down. They're the first row of a scoreboard the school has probably never kept. On a system that already classifies each conversation, this afternoon becomes a report. (The Night Shift Kit has the sort sheet with the twenty standard questions pre-tiered.)

**Step takeaway.** Three percentages and fifteen questions. If most messages have a published answer, the dark hours are a lookup problem and the people go where the rest is.

### Step 3. Set the dark-hours rule

| | |
|---|---|
| **Owner** | Admissions director |
| **Time** | An hour to decide, a week to watch |
| **Done when** | Each tier has a play with a real owner and a real time, and respond-only is enforced by the system |
| **Requires** | Routing by tier at any hour; an acknowledgment that raises a task on the student record; a system that treats a reply and an outreach differently; a phone line the agent can answer, with text-back for anything it doesn't |

**The first half of the rule is what the dark hours are for: replying.** A prospect who reaches out at 2 AM has opened the conversation, and answering is the school keeping up its end. Anything the school starts on its own (a check-in, a reminder, a re-engagement text) waits for the morning. Federal telemarketing rules (47 CFR 64.1200) restrict solicitations to 8 AM to 9 PM in the recipient's time zone, and the most restrictive states stop at 8 PM. A reply to a prospect's own inquiry is a different thing from an outreach, and the system has to know which one it's sending.

> **Dark hours are respond-only.** The school replies to what comes in and initiates nothing. Lumion enforces this by play type: replies to an inbound message go out at any hour, and anything the school initiates holds until quiet hours end. A rep's personal cell has no such setting.

**The second half is the three plays.** One per tier. Each one is configured once and runs every night.

| Play | Tier | What the agent does | Ends with | Who owns the rest |
|---|---|---|---|---|
| **First touch at any hour** | One | Answers the question from approved knowledge, on the channel the prospect used. | A tour slot, an application link, or a callback window. Never "someone will be in touch." | Nobody. The thread is done until the prospect replies or the tour needs confirming. |
| **Heard and scheduled** | Two | Acknowledges with a name and a real time. Raises a task on the student record with the whole thread attached, due inside the window the acknowledgment named. | A named person's reply, inside the window. | The named owner, on the task. |
| **Sensitive handoff** | Three | Acknowledges with a name and a window. Sends no facts, no links, no numbers. Routes the task to the named owner. | A phone call from that owner. | The named owner, on the task. |

What the three sound like at 11:42 PM:

> **Tier one.** Yes, the weekend Class A cohort runs Saturdays and Sundays, and the next one starts October 17. Two Saturday tour times this month: October 3 at 10 AM or October 10 at 10 AM. Which works?

> **Tier two.** Got it. Transfers depend on the program and where the hours came from. Marcus handles those and he'll text you Monday before noon with what we'd need from you.

> **Tier three.** That's a real question and I want to make sure you get a real answer. Dana runs financial aid and she'll call you Monday between 9 and 10 unless a different time works better. Want to pick one?

Two details that decide whether the plays hold. The time has to be a real one; "tomorrow" at 2 AM on a Saturday is a promise nobody will keep, so the acknowledgment has to count business days, not calendar days. And the tier-one next step has to be real too. "Pick a Saturday tour time" needs a tour calendar with inventory on it. If the school doesn't have one yet, the agent offers two named callback windows instead ("Monday between 9 and 11, or between 1 and 3?"), which is still a commitment the prospect makes and still beats a link to the page he was already on.

**The phone.** In the trades, the after-hours inquiry is as likely to be a call as a form: the number on the Google listing, dialed from a truck cab or a break room. A call that rings to voicemail at 11:40 PM is the most expensive silence in the grid, because the prospect had the phone in his hand and the school's number already dialed.

The phone gets the same three plays as every other channel. On Lumion the school's line rings to the agent after hours: it answers, handles tier one from the same approved knowledge, acknowledges tier two and three with a name and a window, and the call summary lands on the student record before the coordinator arrives. A call the agent doesn't take, or a caller who hangs up, gets a text back within a minute from the school's own number:

> Sorry we missed you. This is Apex Technical's admissions line. What can I help with tonight, and what's the best way to reach you tomorrow?

The call and the text are one thread on one record, so Monday's coordinator sees the whole thing.

**For a one- or two-person team.** Every acknowledgment names you, and the windows stack. Tour confirmations and tier three get the first hour of the morning, tier two gets the rest of it. Promise what you can keep and the scoreboard will tell you whether you're keeping it.

**Step takeaway.** Respond-only overnight. Three plays, one per tier, each with a real owner and a real time. The phone gets the same three.

### Step 4. Decide what happens between 6 and 11 PM

| | |
|---|---|
| **Owner** | Admissions director |
| **Time** | One decision |
| **Done when** | The school has chosen "tonight" or "tomorrow" for tier two, and if tonight, someone is paid for it |
| **Requires** | Nothing new. The dark-hours plays already cover it. A task queue a coordinator can clear from a phone makes "tonight" possible. |

This is the second peak. In EAB's sample, 43 percent of student questions arrived between 6 and 11 PM. With the dark-hours rule in place, the evening already works: tier one gets answered, tiers two and three get acknowledged with a time. The only question is whether the school wants to turn "before noon tomorrow" into "tonight."

| | Evening shift | Coordinator on call |
|---|---|---|
| **What it is** | A rep at a desk, 5 PM to midnight | One coordinator clearing the tier-two task list from a phone, about 15 minutes, three evenings a week |
| **Hours added** | 35 | 0. The agent covers the hours; the coordinator shortens the wait on tier two. |
| **Annual cost** | ~$67,000 loaded | ~45 minutes a week, compensated and rotated. Under $1,500 a year at a loaded coordinator rate. |
| **Who does it** | The newest hire, alone | Whoever's turn it is, from the couch |
| **What it changes** | Where the wall is | Whether a tier-two promise says "tonight" or "tomorrow" |

Forty-five minutes a week against a $67,000 shift is the cheapest acceleration a school will ever buy. It is also optional, which is the point. The grid holds without it.

The evening is also when warm leads answer. Lumion's evening nurture play sends check-ins to warm leads between 6 and 9 PM, inside quiet hours, on the night they're most likely to reply, and never overnight. That's the outreach half of the week, and the evening is where it lives.

**Step takeaway.** The evening already works once the dark-hours rule is in. The only decision is whether "tomorrow" becomes "tonight," and that costs 45 minutes a week.

### Step 5. Design the morning handoff

| | |
|---|---|
| **Owner** | Whoever opens admissions each morning |
| **Time** | 45 to 60 minutes, in two blocks, done by 11:30 AM |
| **Done when** | Every overnight tier-two and tier-three thread has a human touch, in the middle of the conversation, inside the window the acknowledgment promised |
| **Requires** | An overnight digest: every thread the agent finished, every call it took, every task it raised, sorted by tier, with the whole conversation attached |

This is the step that separates schools that cover the dark hours from schools that only look like they do. An overnight acknowledgment that promises "before noon" and delivers at 3 PM has taught the prospect the school's promises are approximate. The handoff is where the promise is kept or broken, and it is the metric almost no school tracks.

It's also the step that breaks if it's designed for a quiet morning. Between 8:30 and 10:30 the front desk is taking walk-ins, tours, and the phone. A coordinator facing fifteen overnight threads before 10 AM will reach for a template, and a template is the acknowledgment sent twice. So the morning runs in two blocks, and the acknowledgments the agent sent overnight promise times that fit them.

![What a designed handoff looks like](charts/morning-view.png)

*A concept drawing of the overnight digest. The point is the shape: finished tier-one threads, tier-two tasks with a clock on them, one tier-three task already routed.*

**Block one, 8 to 9 AM: the urgent third.**
- Confirm every tour picked overnight for today or tomorrow, by call and text.
- Check that every tier-three task landed with its owner and that the window the agent promised is on that person's calendar.
- Return every missed call, including the ones that got a text back. The text bought time; it didn't close the loop.

**Block two, 9 to 11:30 AM: the consultative rest.**
- Read every overnight tier-two thread, all the way back to the question.
- Reply in the middle of the conversation. "You asked about transferring hours from Dayton. Here's what we'd need." Never "thanks for reaching out."
- Keep the time the acknowledgment promised. If the promise was "before noon," noon is the deadline, and the scoreboard counts it.
- Move the stage and close the task, so the record shows the promise was kept.

On Lumion the morning starts at 8:05 AM, the first permitted minute after quiet hours, with the morning warm-up play: a one-line nudge to overnight contacts who didn't take the next step ("Morning. Still want that Saturday slot?") and the task digest for the coordinator, tier three and tour confirmations on top. By 9 AM the handoff is half done before anyone has opened the inbox.

Pivot Point Academy's version of this is instructive. Their web agent books the tour in the chat, so the coordinator's first human touch the next morning is confirming a tour with someone who already chose a program and a time. Her job is keeping a conversation the agent already started.

**Step takeaway.** The handoff is where the overnight promise is kept or broken. Two blocks, urgent first, and the acknowledgments promise times that fit them.

**Takeaway.** Steps one through three are the design and take a week. Four and five are the operating rules and take a decision and a morning.

---

## How the grid runs on Agent Teams

*Everything above is one customer journey.*

Every step in this playbook maps to something the platform runs, and this section shows the mapping plainly. The school marks each stage agent-owned or staff-owned once, during setup. The agent runs the agent-owned stages at every hour and raises a task at every handoff. The grid below is the same nine cells from section 3 with the platform's plays in them.

![The grid as Agent Teams run it](charts/platform-grid.png)

| Play | Cell | Requires |
|---|---|---|
| **First touch at any hour** | Tier one, all hours | Web agent on chat, text, and the school's phone line |
| **Schedule fit** | Tier one, evenings and weekends | Cohort schedule in the approved knowledge; a tour calendar or booking link |
| **Call catch** | Tier one, dark hours | The school's phone line on the platform; the agent answers, and texts back within a minute on any call it doesn't |
| **Heard and scheduled** | Tier two, all hours | Tasks on the student record with a deadline; a named owner per topic |
| **Sensitive handoff** | Tier three, all hours | Same, plus the never-answer list |
| **Morning warm-up** | Tier one and two, 8:05 AM | Outbound agent for the nudge; the task digest works without it |
| **Evening nurture** | Warm leads, 6 to 9 PM | Outbound agent |

Three things the platform enforces that a shift or a chatbot cannot: respond-only in the dark hours, by play type; the approved knowledge as a hard boundary, with escalation for everything outside it; and a log of every send with what sent it, what it was allowed to say, and why.

**Already on the site.** Western Welding Academy cut first response from 49 hours to under 10 minutes. Pivot Point Academy went from about two qualified leads a day to four, and enrolled its first student through the web agent on the day it went live.

**Takeaway.** The grid is one customer journey. Tier-one stages run on their own at every hour; everything else raises a task with a name and a clock on it.

---

## 7. The scoreboard

*Three metrics. The third keeps the first two honest.*

Most schools track an average response time, and an average is the number that hides the problem: a fast Tuesday afternoon covers for a dead Saturday. Three metrics replace it. The first finds the slowest cell of the grid. The second counts whether the overnight promises were kept. The third samples the answers for accuracy, because a fast wrong answer is the one failure the first two will score as a success.

| Metric | How to measure it | Target | Why it matters |
|---|---|---|---|
| **Time to first substantive reply** | By hour of day and day of week. Report the slowest cell. | Under 10 minutes in every cell, including 2 AM Sunday | An average lets a fast Tuesday afternoon hide a dead Saturday |
| **Promises kept** | Share of overnight tier-two and tier-three threads with a human touch inside the window the acknowledgment named | 100 percent. Below 90 means the day team isn't cashing the checks the agent wrote. | An acknowledgment kept late is worse than none; it teaches the prospect the school's word is approximate |
| **Tier-one accuracy** | Read 20 overnight answers against the approved knowledge. Twenty minutes on a Friday. | 20 of 20 | A fast wrong answer is a failure the response-time metric will count as a win |

Two supporting rows worth keeping: **missed calls without a text-back** and **off-grid replies** (anything sent from a personal cell or inbox), both with a target of zero.

> **Get the first number tonight.** Fill out your own form at 9 PM on a Friday. Ask a real question in the comments box. Then call the main line and hang up after four rings. Time the first reply of any kind, then the first reply that actually answered the question, then check whether it ended in a next step. Most operators who do this don't need the rest of the playbook to know what they're going to find.

### What a point of yield is worth

Coverage design doesn't create demand, so the honest way to size this is exposure. Take the inquiries the school already pays for, and ask what one point of inquiry-to-enrollment yield is worth at that volume. Then look at the sort from Step 2 and how many of those inquiries arrived when nobody was there.

| Inquiries a month | One point of yield, per month | Per year |
|---|---|---|
| 200 | 2 enrollments, about $30,000 | $360,000 |
| 500 | 5 enrollments, about $75,000 | $900,000 |
| 1,000 | 10 enrollments, about $150,000 | $1,800,000 |

*At $15,000 average tuition. Replace with your own numbers.*

Whether the dark hours are worth one point or three at a given school depends on the sort. A school where half the inquiries arrive after hours and none of them get a substantive reply until Monday is not looking at one point. This table is what the first one costs.

**Takeaway.** Report the slowest cell, count the promises kept, and check twenty overnight answers a week. A fast wrong answer is a failure the first number will count as a win.

---

## 8. The caveat

*One rule overrides the rest.*

Two things the grid does not do, before the rule that governs everything it does.

Coverage design doesn't create demand. It stops the school from losing demand it already paid for. A school with no inquiries at 2 AM has a marketing problem this playbook won't fix.

And the grid has one rule that overrides the rest.

> **THE RULE THAT OVERRIDES THE REST**
>
> Wrong at 2 AM is **worse than silent** at 2 AM. A prospect who waits until morning for a real answer is annoyed. A prospect who was told at midnight that his benefits would cover it and finds out Monday that they won't is gone, and he tells people. The three tiers exist so that the fast answers are the safe ones and the risky answers are the slow ones. Any tool or process that blurs that line is worse than the shared inbox it replaced.

Next: Back to Friday night →

---

## Back to Friday night

*CLOSE*

*The inquiry has an owner because the school decided.*

The forklift operator's inquiry has an owner at 11:40 PM because the school decided, months earlier, what a weekend-schedule question is (tier one), who answers it after hours (the agent, from what the school approved), what the answer ends with (a Saturday tour slot), and who picks it up Monday (the coordinator, in the middle of the conversation). None of that required an evening shift. All of it required deciding, and a system that could carry the decision through the night.

That's the whole difference between a school that covers 40 hours and a school that covers 168. One of them wrote it down, and then handed it to something that never sleeps.

**The Night Shift Kit** (four pages, printable): the blank grid, the sort sheet with twenty questions pre-tiered, the three acknowledgments as fill-in templates, the two-block morning checklist, and the weekly scoreboard. **Agent Teams** runs the grid: lumion.ai/engagement.

---

*Sources new to this piece: 47 CFR 64.1200 (FCC telemarketing rules). NCES, College Student Employment (2020). UPCEA, 2025 Enrollment Process Review: Secret Shopper Analysis. EAB, What Are Students Asking University Chatbots (Oct 2025). CUPA-HR, The Higher Ed Admissions Workforce (2023). Rep cost, 128-hour gap, Western Welding, and Pivot Point figures as published on lumion.ai. Western Welding re-engagement figures held pending reconciliation. Evening window standardized on EAB's 6 to 11 PM. The handoff view is a concept drawing.*
