AI Client Communication Workflow: What It Costs to Automate Client Texts

Published August 21, 2026By ABD Legacy LLC
AI client communication workflow ChatGPT Messages plugin cost client communication automation ROI

On August 20, 2026, OpenAI shipped an Apple Messages plugin for the ChatGPT desktop app on macOS: with permission granted, ChatGPT can read, search, summarize, draft, and send iMessage, SMS, and RCS messages (OpenAI release notes, Codex plugin docs). The software is plan-inclusive. The real cost of building an AI client communication workflow on it is the hardware, the seats, the review time — and the compliance overhead you choose to carry.

The ChatGPT Messages plugin in two sentences (and what it costs to reach it)

The plugin reads and searches iMessage, SMS, and RCS conversations on the Mac, and can find, summarize, draft, or send messages through Messages (9to5Mac, OpenAI). It is available on all plans in the ChatGPT desktop app for macOS, with the functional surface in ChatGPT Work and Codex (MacRumors).

The first cost gate is hardware. The plugin is included only in the Apple Silicon (arm64) build of the desktop app and does not run on Intel Macs (OpenAI, TNW). An agency on an Intel fleet faces a per-seat upgrade cost before anyone can use it. The second gate is the plan surface: it works in Work/Codex, not regular ChatGPT chats, web, mobile, Codex CLI, or the IDE extension (Unite.AI) — so the seats you price are the agentic ones, not consumer plans.

What an AI client-communication workflow actually costs

Line-item breakdown for the pricing site's audience:

ROI framing: run it through the AI automation ROI calculator — hours saved × billable rate vs. hardware + review overhead. For a 10-person agency already on Apple silicon and Work seats, the marginal cost of the workflow is near zero; for an Intel fleet or a heavy compliance posture, the same workflow has real line items.

Where the automation stops (and why that's a feature)

Selling this as "AI sends client texts unattended" is wrong on both the product and the pricing. By default, ChatGPT sends a message only after you approve the message and its recipients (OpenAI). Per-send you choose "Allow once" or "Always allow sending to this chat" — and OpenAI discourages persistent approval: it "removes your final chance to review a message before ChatGPT sends it as you" (OpenAI, TechCrunch). There is also a documented failure mode: tasks set to Full access (or otherwise disabling approval prompts) can break the send confirmation — the remedy is "Ask for approval" or "Approve for me" (OpenAI).

For pricing, this means the workflow is "AI-assisted client comms," not full automation: you price the drafting and triage win, set client expectations accordingly, and keep the human-review step in the deliverable. Admins can disable the plugin per workspace via the existing Computer Use control (OpenAI), so you can scope it to the staff who actually run client communication.

Model the cost of an AI client-communication workflow

Open the AI Agency Pricing Calculator →

Price Work/Codex seats, hardware, and compliance overhead against hours saved × billable rate.

Frequently asked questions

Do I need a new Mac for the ChatGPT Messages plugin?

If your Mac is not Apple silicon, yes. The plugin is included only in the Apple Silicon (arm64) build of the ChatGPT desktop app and does not run on Intel Macs — so agencies on an Intel fleet face a per-seat hardware cost before the software can be used.

Is the Messages plugin extra cost?

No. The plugin is available on all plans in the ChatGPT desktop app for macOS, with the functional surface in ChatGPT Work and Codex. There is no per-feature add-on — the cost is the plan seats and the hardware, not the plugin itself.

Does ChatGPT send texts automatically?

Not by default. Sending is gated by default — ChatGPT sends a message only after you approve the message and its recipients. You can choose "Allow once" or persistent approval per chat, and OpenAI discourages persistent approval because it removes your final review before a message is sent as you.

What does this cost to run for a 10-person agency?

The plugin itself is plan-inclusive. The real costs are hardware (Apple silicon per seat if you are on Intel), ChatGPT Work/Codex seats priced into retainer math, and one-time compliance review of consent and data handling — because OpenAI has not yet documented exactly which message content leaves the machine or published managed-Mac admin guidance.

Can we turn it off for some employees?

Yes. In managed workspaces, administrators can disable Apple Messages through the existing Computer Use control, so you can scope the plugin to the staff who actually run client-communication workflows.

Sources

Accuracy note: announced Aug 20, 2026; plugin included only in the Apple Silicon (arm64) build (Intel Macs out); available on all plans with the functional surface in ChatGPT Work and Codex; sending gated by default with "Allow once" vs persistent approval (OpenAI discourages persistent); known issue where Full-access tasks can break the send confirmation (remedy: "Ask for approval"/"Approve for me"); admins can disable via the Computer Use control. OpenAI has not documented exactly which message content is transmitted for processing, has not published enterprise admin guidance, and Apple was not part of the announcement — open questions, not confirmed facts. This article estimates costs; run your own numbers in the calculator.

Related reading