When customers text back: running a two-way inbox on top of RealGreen automations

What happens when customers text back your automations? They will, constantly, and it is the side effect we see companies plan for least. Every texting automation you turn on (visit reminders, review asks, payment links, upsell offers) starts conversations. They reply to the pre-visit notice with a gate code, to the invoice link with a billing question, to the upsell with "how much for the back lot too?" Those replies are not noise; they are the highest-intent inbound channel you have, warmer than your web forms and cheaper than your phone lines. But only if somebody answers. An unanswered text reads worse than no text at all, because the automation proved you can reach them and the silence proved you were not listening. Before you scale sends, build the inbox.

How many replies are we talking about?

Plan with arithmetic, not vibes. A 2,000-customer company running a normal automation stack might send on the order of 14,000 automated texts a season across reminders, summaries, AR nudges, and campaigns. At a 4-7% reply rate, that is roughly 600 to 1,000 inbound messages a season, clustering hard around send mornings and route days. Most take under two minutes to resolve; maybe a fifth need a schedule lookup or a callback. Call it two to four staff hours a week in season. Illustrative numbers, but the shape is reliable: this is a part-of-someone's-job workload, not a hire, and far too much to leave to whoever happens to notice.

Shared inbox or personal phones?

Shared inbox, and this one is not close. Texts routed to a manager's cell phone die on vacation days, walk out the door with departing employees, and leave zero record for the office. The requirements for the tool are short: one business number for all automated sends and replies, a team inbox where any trained person can see and answer conversations, assignment so a thread has an owner, and every message tied to the customer record so "who is this?" never happens. CRM-layer tools built for home services handle this natively: it is a core reason companies bolt a system like HighLevel onto RealGreen rather than texting from a disconnected app. The specific tool matters less than people think: we can connect RealGreen to pretty much any inbox platform with an API, a Zapier connection, or a native integration. The disqualifying failure is context: an inbox that shows a bare phone number instead of "the Hendersons, full program, invoice open, visit Thursday" turns every reply into a research project.

Who answers, and how fast?

Set a service-level agreement even though nobody but you will ever see it: replies within 30 business minutes in season, and a courteous next-morning catch for evening texts. Texting carries chat-speed expectations; a customer who texts at 10am and hears back at 4pm experienced you failing, even if the answer was perfect. Ownership should be a named primary with a backup, not "the office." The same discipline governs your after-hours behavior: an auto-acknowledgment ("got it, we open at 8 and you are first in line") honestly buys you the night, and it is the same principle as the missed-call text-back play on the phone side: acknowledge instantly, resolve during business hours.

Can AI draft the replies?

Draft, yes. Send, mostly no. The current honest state of the art: AI is genuinely good at reading a thread plus the customer record and drafting the likely answer ("your next application is scheduled for Thursday the 14th"), and a human tapping approve on a correct draft turns a two-minute reply into a ten-second one. That human-approves pattern is where we would start, and for many offices stop. Full auto-send is defensible only for the narrowest confirmations, and never for pricing, complaints, cancellations, or anything with an edge of frustration in it: misreading tone in those threads is expensive exactly in proportion to how much the customer cares. We took the longer version of this position in our piece on AI answering services: the technology clears the bar for drafting and triage well before it clears the bar for unsupervised conversation.

What gets escalated, and to whom?

Most replies are logistics. The ones that are not need a pre-agreed path, written down where the inbox owner can see it:

• Cancellation language goes to whoever owns saves, same day, as a call, not a text thread.
• Complaints open a ticket and get a human call within one business day; the text reply just confirms it is being handled.
• Billing disputes move to phone or email where detail lives better than in bubbles.
• Buying signals ("how much to add the back lot?") route to sales the same hour: this is the warmest lead your stack produces.
• STOP and opt-out language is handled by software, instantly and without exception: that part is law, not judgment, and the details live in our texting-compliance guide.

Why log every thread against the customer record?

Every thread should end up attached to the customer, visible beside their program, invoices, and visit history, and readable by the next person who touches the account. This is what turns texting from a side channel into institutional memory: the tech sees the gate-code exchange, the CSR sees last month's complaint before answering this month's question, the renewal caller sees the upsell reply that never got followed. And conversation data compounds: a season of threads is an unfiltered record of what customers actually ask, which quietly rewrites your templates, your FAQ, and your campaign copy. If your texting tool cannot write conversations back somewhere durable and customer-linked, it is a dead end, however pretty the inbox.

When is two-way texting not worth it?

Fairness section. If nobody in your office can own a 30-minute SLA in season, do not open the channel: send one-way notifications from a number that says so, and route questions to phone and email you can actually staff. A visibly ignored inbox is worse than an honest no-reply. And if your automation volume is tiny: a few hundred sends a season: replies will be too rare to justify tooling; a forwarding number and a spreadsheet will hold until volume grows.

The takeaway: automations start conversations; the inbox is where you keep the promise they make. One shared number, a named owner, a 30-minute clock, AI drafting with human sends, and every thread on the record: that is the whole system.

Your rollout plan

1. Consolidate all automated sends onto one business number wired to a shared, record-linked inbox.
2. Name the inbox owner and backup, and set the in-season SLA at 30 business minutes.
3. Write the escalation card: cancels, complaints, billing, buying signals, opt-outs, each with an owner and a clock.
4. Turn on AI drafting with human approval, and review a sample of drafts weekly for the first month.
5. Audit monthly: reply volume, response times, threads logged, and what the questions are telling you to fix upstream.