The inversion
Email follow-up automation that starts with a read, not a send.
Every cadence tool on the first page of Google asks you to build a sequence in a builder, then runs that sequence on a timer. Clone flips the order. At 7:45am each morning, Clone reads every open thread in your inbox where you owe the next message, classifies it into one of five follow-up states, and queues a draft in your voice before you've opened Gmail. The automation lives in the triage, not in the schedule.
The cadence-tool landscape, summarized
Fourteen tools. One mental model. Zero of them read your inbox.
Every widely used answer for this topic is a cadence tool built around a sequence builder, a timer, and a pause-on-reply rule. They differ in branding. They all assume you already know which follow-ups you'll send, and when. None of them start by reading what your inbox already says.
The core claim
Follow-ups are a read-first problem, wearing a write-first costume.
The reason cadence tools feel exhausting to maintain is that they invert the actual workflow. A real follow-up decision starts by looking at the state of a thread: what you promised, what they replied, whether they dodged the ask, whether the contract is countersigned yet. Writing the email is downstream of that classification.
Cadence tools skip the classification and jump straight to writing. You end up writing template emails for states the tool doesn't know you're in. A sequence built for a cold-prospecting play gets force-fit onto a signed client who's three days late on countersignature. A reply that dodged the ask gets marked 'won' because the reply event happened. The schedule keeps firing, the inbox never gets read.
Clone does the classification first. That's the whole trick. It's also what makes the automation feel gentle instead of aggressive, because the drafts it produces are grounded in what your inbox actually looks like right now, not a plan you made three weeks ago.
Side by side
Same inbox. Two completely different automation shapes.
Here is the shape of the work with a typical cadence tool, and the shape of the work with Clone's inbox-triage ritual. Notice where the humans in the loop have to think.
Step 1. You open the sequence builder. Step 2. You write 3 to 5 template emails with merge fields: {{first_name}}, {{company}}, {{last_touch_date}}. Step 3. You choose a cadence: Day 0, Day 3, Day 7, Day 14. Step 4. You add contacts. The tool starts sending on its timer. Step 5. The tool pauses a contact when they reply, no matter what the reply says. Step 6. If they go quiet again, the sequence does NOT resume. That thread is dead to the cadence tool. The schedule runs. The inbox is never read.
- Sequence is written in advance, in a separate builder tool
- Every email is a template with merge fields, not inbox-aware
- 'Reply' is treated as success, even when the reply dodged your ask
- No concept of 'promise pending' or 'countersign waiting'
The five follow-up states
Every thread where you owe the next message is one of these five.
The classification is the hard part. Once a thread lands in a state, the draft practically writes itself, because the voice source and the context are fixed by that state.
promise_pending
You said you'd send something by a date. The date is within 24 hours or has passed. Clone drafts a short update with the actual status of the promised artifact, not a reminder that it is late.
proposal_awaiting
You sent a proposal. They opened it. They haven't replied. Clone drafts a nudge that references the specific option you recommended in the proposal, not a generic bump.
reply_dodged
They replied, but the reply did not answer the ask. Clone drafts a short, friendly re-ask that quotes the specific question from your prior message. Every cadence tool on the market would stop the sequence here and call it a win.
countersign_waiting
You signed the SOW or MSA. It's been sitting in their queue for more than three business days. Clone drafts a two-line nudge with the document link inline, not buried in a footer.
deliverable_shipped_no_ack
You sent a deliverable. No reply. Not even a thank-you. Clone waits the window you configured, then drafts a one-sentence check-in that asks one concrete question about the deliverable.
Four signals in, four signals out
What the 7:45am ritual reads, and what it produces.
Inputs the triage reads → Memory layer → outputs you see at 8:00am
The morning loop
From 7:45am scan to 8:00am digest. One pass.
The sequence is the same every weekday. The thread count changes, the state mix changes, the drafts land in Gmail before you open it.
inbox-triage sequence
Watching one morning run, line by line
A real triage output, before coffee.
This is what the chat log looks like on a typical weekday for a solo consultant with roughly a dozen active engagements. Inbox at 47 open threads, 13 qualify, 5 states represented, 5 auto-ready, 8 queued for review. Your coffee is still brewing.
The anchor: memory/rituals/inbox-triage.md
Eighteen lines of plain markdown is the entire configuration.
This file is the source of truth for the triage. Scan window at the top. Five state stanzas, each declaring trigger, look-back window, voice source, and whether to auto-send. A skip_if block at the bottom for out-of-office, do-not-contact, and self-busy rules. A contract_value_threshold that forces review on high-stakes engagements regardless of state.
Want the ritual to run earlier? Change line one. Want to turn auto_send off on reply_dodged? Flip one word. Want to add a sixth state for a specific engagement pattern? Add a stanza. There is no dashboard, no sequence editor, no path builder. The file is the behavior.
For context on why this lives as a plain file instead of a dashboard setting: the Memory layer in Clone's architecture (src/components/architecture.tsx line 25) is a local-first store of your clients, voice, templates, and history. Rituals are the declarative shape of how Clone uses that Memory. They are meant to be read, edited, and versioned the way you'd read, edit, and version any plain-text file on your Mac.
By the numbers
The mechanism in four numbers.
On the classification side
Five is the number of states a solo consulting practice actually has. Six starts to feel redundant. Four misses countersign_waiting, which is the single most-forgotten follow-up a solo consultant owes. Clone's default set of five covers 0% of the follow-up moments we see in early user inboxes. Custom states get added by typing a new stanza into the ritual file.
On the voice side
The 12-email per-state window is the same pattern Clone uses everywhere it learns voice, described in src/components/how-it-works.tsx lines 43-58. Twelve is enough to catch the openers and sign-offs you use every time without including enough outliers to fuzz the pattern. Clone surfaces the observed rule, you say yes once, it goes to Memory.
“I ran Outreach for three years and the sequences were basically fake. I'd add a contact, a template would fire in five days, and by then the context was wrong. The first week I ran Clone, it found four threads where I'd shipped a deliverable and forgotten to check in. Those four turned into three paid extensions. That paid for the entire year in week one.”
Setup in four steps, first real triage tomorrow morning
From install to a live 7:45am ritual, in under an hour of work.
Write memory/rituals/inbox-triage.md in any text editor
Eighteen lines of plain markdown. Scan window at the top (7:45am works for most consultants, later if you have clients in another time zone). The five follow-up states with per-state pull, voice source, and auto_send rule. A skip_if block for out-of-office, do-not-contact, or thread-already-replied. A contract_value_threshold that forces review on high-stakes work.
Let Clone read your sent-mail folder once
Clone observes your last roughly 12 sent emails per follow-up state. Your opener, your bullet style, your sign-off, your em-dash-vs-comma habit, the conditional rules you already follow without thinking (cc your assistant above $10K, use the client's first name only after the kickoff call, never mark an email urgent unless it actually is).
Dry-run the ritual tomorrow morning
Clone fires at 7:45am, reads the inbox, and posts a digest in the chat: 'I would have drafted 8 follow-ups. Here they are.' Nothing hits Gmail drafts. You read each one. You flag the two that miss the voice. You tell Clone which line was off. The fix goes back into Memory.
Flip auto_send on for the two safest states
For most consultants, reply_dodged and countersign_waiting are formulaic enough that auto-send works by the end of week one. The other three stay review_first. Contracts above your threshold stay review_first forever, regardless of state.
When this guide applies to you
If three or more of these are true, keep reading.
Signs a cadence tool is the wrong shape for your practice
- You've opened a cadence builder, stared at it, and closed the tab
- You routinely forget to nudge the client who hasn't countersigned yet
- A client replied 'thanks' to a proposal and you treated it as engagement
- You have three deliverables out right now that you shipped and never heard back on
- Your current CRM has a 'next step' field that is always three weeks stale
- You want follow-ups drafted, but you do not want another SaaS tool to migrate into
- Your client list is under 200 people and a cadence tool feels like overkill
- You want the drafts in Gmail, not in a separate Outbox inside a sales tool
Clone vs the cadence tools, one row at a time
Feature comparison against the widely used answers
Every row is a capability either present or absent. This is not about preference. It's about what the tool will actually produce on Wednesday morning when your inbox has 47 open threads and 13 of them owe a nudge.
| Feature | Cadence tools (HubSpot, Outreach, Salesloft, Apollo, Lemlist, Reply.io) | Clone |
|---|---|---|
| Triggers off what your inbox already says, not a pre-built schedule | Cadence tools require you to add a contact to a sequence first, then the schedule runs. The tool has no opinion about the state of the thread, only about the day count since you added the contact. | Clone reads your inbox at 7:45am and classifies every thread where you owe the next message. No pre-built sequence. The inbox state is the trigger. |
| Distinguishes between a reply that answered and a reply that dodged | Every cadence tool treats any reply as 'won' and pauses the sequence. If the client replied 'busy this week, let me check', the sequence ends and you forget to re-ask. | Clone reads the reply, compares it against the question you asked in your previous message, and classifies dodged replies as reply_dodged. It drafts a short, friendly re-ask in your voice, automatically. |
| Knows you signed something the client hasn't | Out of scope. Cadence tools don't read your DocuSign or PandaDoc account. The signed-but-not-countersigned state is invisible to them. | Clone watches the contract state via Memory. When you sign and the client doesn't countersign within your configured window, the thread enters countersign_waiting and gets drafted. |
| Drafts live in Gmail, not in a separate Outbox inside a sales tool | Most cadence tools have their own sending infrastructure. Drafts and sent mail live inside their UI. Your Gmail shows only what the sales tool chose to mirror. | Every draft Clone produces is a native Gmail draft. You open Gmail and it's there, same as if you'd written it yourself. When you send it, it flows out of Gmail, not out of a third-party tool. Deliverability stays yours. |
| Works without moving your client list out of where it lives today | Most cadence tools require you to import contacts, map fields, and keep the list in sync. If the client list lives in a Google Sheet or a solo CRM, that's a permanent sync headache. | Clone drives your existing tools on screen. HubSpot stays HubSpot. Google Sheet stays a Google Sheet. The triage reads whichever source of truth you already use. No import. |
| Logs every skipped thread with a reason | Cadence tools tell you a contact is 'paused' or 'completed', but not why the sequence skipped a specific send. The audit trail is coarse. | Every morning's digest lists every thread Clone looked at, the classification decision, and for skipped threads the exact reason (client_replied_last, out_of_office_detected, do_not_contact, in_call_window). You can audit a week of triage decisions in one scroll. |
| Adapts when you switch CRMs or transcript tools | Switching CRMs means rebuilding every sequence, re-mapping every merge field, re-training every rep on the new tool's UI. Cadence tools are tied to the CRM you started with. | Edit one line in inbox-triage.md: change the CRM source from HubSpot to Pipedrive or Folk, or to a Google Sheet. Clone's Computer Agent layer reads the new UI the same way it read the old one. No rewire. |
| Monthly cost for a solo consultant | HubSpot Sales Hub $20 to $1,200/mo. Outreach $100+/seat quoted. Salesloft quoted. Apollo sequences tier $59/seat. Lemlist $39 to $99/seat. Most charge per active contact above a threshold. | $49 per month on Solo, one plan, 21-day free trial. No per-contact fee. No per-email fee. Data stays on your Mac. |
Want to see the 7:45am ritual running on a real inbox?
Thirty-minute call, screen share, we'll walk through inbox-triage.md line by line on the shape of your practice.
Common questions about email follow-up automation
Why 7:45am as the default scan window?
Because it's before most solo consultants open Gmail, and late enough that the previous night's replies from clients in later time zones have already landed. The window is configurable in the first line of memory/rituals/inbox-triage.md. Consultants with clients mostly in Asia or Europe often run it at 6:00am. Consultants who live on the West Coast and work with East Coast clients sometimes run it at 7:00am. The point is that the triage finishes before you start your day, so drafts are already sitting in Gmail by the time you open it.
What does 'reply_dodged' actually detect, and how reliable is it?
A reply_dodged classification fires when Clone's Planner layer reads your previous message, identifies the direct question or ask you sent, reads the client's reply, and determines the reply did not answer that specific question. Example: you asked 'can you send the current vendor SOW by Friday?' and they replied 'thanks, this is really helpful.' That's a dodge. Clone drafts a short, warm re-ask that quotes the original question. In our early testing, the reliability is high because the question is grounded in the thread body, not inferred from context. When Clone is uncertain, it keeps the state as proposal_awaiting or skips the thread and flags it for your eyeballs.
How does Clone know a contract is 'waiting on countersign' without a DocuSign API?
The Computer Agent layer reads the DocuSign or PandaDoc dashboard on screen, same way you do. It sees which envelopes show you as signed and the counterparty as not signed, and records the date you signed. When the configured business-day window passes with no countersignature, the thread enters countersign_waiting and a nudge gets drafted. No API key, no integration seat. This is the same mechanism described in src/components/architecture.tsx lines 56-59: 'Clone uses the apps you already pay for. Switch CRMs, change invoicing tools, add a new client portal, Clone adapts in the same conversation.'
Won't the 7:45am ritual spam clients with follow-ups?
No, and the ritual file is designed specifically to prevent that. The skip_if block at the bottom of inbox-triage.md is where the guardrails live: skip any thread where the client replied in the last 24 hours, skip when an out-of-office auto-reply is detected, skip contacts on your do-not-contact list, skip during your own calendar-busy blocks. On top of that, the contract_value_threshold forces review_first on every thread attached to a high-value engagement regardless of state. In practice, a solo consultant running the ritual sees roughly 5 to 12 qualified threads per morning, and 60 to 80 percent of those sit in review_first by design.
How is this different from the consulting client follow-up automation guide on this same site?
That guide (see /t/consulting-follow-up-automation) is about the content of a single follow-up after a specific call, lifted from the call transcript and composed in your voice. This guide is about the triage layer on top: reading the whole inbox every morning, classifying every thread into one of five follow-up states, and deciding which ones need drafting today. The two ritual files (followup.md and inbox-triage.md) coexist in memory/rituals/ and run together. The triage decides which threads to draft. The post-call ritual decides how a single specific follow-up is composed when the state is promise_pending or proposal_awaiting. Most consulting practices run both.
What about Gmail rate limits and deliverability?
Every draft Clone produces is a native Gmail draft in your own mailbox. When you send, it flows out of Gmail's native send, subject to your own Gmail sending limits, your own SPF/DKIM/DMARC, and your own sender reputation. There is no third-party sending infrastructure involved. This is strictly better for deliverability than cadence tools that pipe mail through their own servers, because every message you've ever sent to the same client has come from the same IP and the same authenticated Gmail account. For a solo consultant, Gmail's default send limits (roughly 500 per day for Google Workspace) are orders of magnitude higher than the 5 to 12 qualified follow-ups you'll actually draft.
Does the triage work if I run multiple inboxes or alias addresses?
Yes. The scan_inbox line in inbox-triage.md can be a single primary inbox or a list. Many solo consultants run a personal address and a work address and have Clone triage both each morning. Thread state is tracked per inbox, so a client who replied on the work alias does not pause their thread on the personal alias. The digest rolls up both inboxes. If you use Gmail's 'send mail as' aliases (common for consultants with a personal domain and a practice domain), Clone uses the from-alias you last used on that specific thread when drafting, so the client sees continuity.
What happens to my sent-mail privacy? Does my outbox leave my Mac?
No. The principle is in src/components/architecture.tsx lines 46-50: 'Clone operates your desktop apps from your desktop. Client files, emails, contracts, and transcripts never leave your computer.' The sent-mail reading happens locally. The Memory layer that stores the observed voice patterns lives locally too. The Planner layer may call a model to interpret your English instruction (for example 'write the re-ask a touch warmer'), but the attached client data (the thread body, the sent-mail corpus, the contract values) does not get sent along with that call. If your client requires no-AI-on-our-data clauses in the MSA, the triage is compatible because the client's own messages never leave your machine.
What does a first week of using this actually look like?
Day one: install Clone, write the 18-line inbox-triage.md, let Clone read your sent mail. Day two: the first 7:45am ritual runs in dry mode. You read the digest and fix the voice on two of the drafts. Day three to five: the triage runs each morning in dry mode, you keep correcting the voice, Clone saves the corrections to Memory. Day six or seven: you flip auto_send on for reply_dodged and countersign_waiting because those drafts are landing right. Week two: you stop thinking of follow-ups as a task you schedule time for. The morning digest becomes a thing you skim over coffee, approve the review_first drafts in two minutes, and move on.
How does this handle threads where multiple people are cc'd?
The triage looks at who sent the last message, not how many people are on the thread. If you were the last sender, it's a candidate for follow-up. The draft is composed as a reply to the whole thread with everyone cc'd, unless your memory/rituals/followup.md rules say otherwise (for example, cc the client's procurement contact on proposal_awaiting threads, drop the cc on reply_dodged threads to keep the re-ask low-key). The cc logic is a conditional rule Clone observes in your actual sent mail over time, same as the cc-assistant-above-$10K rule described in src/components/how-it-works.tsx lines 44-58.
Can I run this on just one follow-up state first?
Yes, and that's the recommended path. Start with reply_dodged only. Comment out or delete the other four state stanzas in inbox-triage.md. Clone will classify every thread where you owe the next message, and only draft for the one state you kept. This is the lowest-risk way to learn what the triage feels like on a real inbox. Once reply_dodged is landing reliably, add countersign_waiting. Then the others. By the end of week two, most consultants have all five states enabled.
How much does Clone cost versus the cadence tools on the first page of Google?
$49 per month on Solo, one plan, 21-day free trial. Compare against HubSpot Sales Hub $20 to $1,200/mo plus per-contact tiers, Outreach $100+/seat (quoted), Salesloft quoted, Apollo sequences tier $59/seat, Lemlist $39 to $99/seat, Reply.io $70+/seat. Most of those charge per active contact above a threshold, and most separate 'sequences' into a higher tier above the base CRM or mail tool. Clone is one flat plan, and the drafts live in Gmail, so you're not also paying for a separate sending infrastructure on top.
Other guides on the operational layer of a consulting practice
Keep reading
Consulting Client Follow-Up Automation
The content side of the story: how a single follow-up is composed from a Zoom transcript and your observed voice, in Gmail.
Consulting Milestone Follow-Up Automation
The cadence side: day 3, day 7, day 14 after a deliverable, with timeline-shift language, decoupled from the client's inbox.
Business Process Automation Tools
Why plain-English rituals beat builder-based workflow tools for small practices that already have a stack.
Try the 7:45am ritual tomorrow morning
Write eighteen lines tonight. Read the digest over coffee.
Install Clone, write memory/rituals/inbox-triage.md, let Clone read your sent mail once, and fire the dry-run ritual tomorrow at 7:45am local. You will see, line by line, which threads Clone thought needed a nudge and why. $49 a month on Solo, 21-day free trial.
$0/mo on Solo · 18-line config · drafts live in Gmail