Poldermarketing
What should be automated before increasing GTM headcount?
Automate the path from a signal to a logged next action before you add a GTM seat. Write one definition of a qualified lead, one owner per field, routing, a sequence that stops on a reply, and a report you can recompute. Keep positioning, discovery, and exceptions with a person. Hire when the remaining queue is conversations.
A new seat copies the leak you already have
Add a go-to-market person only after a signal already becomes one record, one owner, and one next action without someone retyping it. A new seat copies an ownerless lead, a duplicate, a deal with no next step, and a sequence that keeps writing after a reply.
RevPartners says if the motion is broken, do not hire yet. Find the leak in demand that never becomes a conversation, pipeline that does not convert, deals that stall, or customers who never expand. Their execution layer is routing, follow-up, enrichment, scheduling, and the lifecycle update. Admin past a fifth of the day, or selling time under half the day, is their test, and only when sellers already exist.
Madrona says to name who you sell to, what you say, and when you reach them. Tools amplify that strategy. The essay is co-written with two founders of Clarify, a CRM the firm backed, then argues for software that owns scoring and follow-up.
Match software to the way you sell now
Airtop says to pick how you sell now, not the model planned for a year and a half. One of its checks is where the last deals came from. Their sales-led order starts with list building and assumes a demo call to action plus a sales team, so the only seller should not copy it. Product-led starts by separating a sales-worthy signup from the rest.
| How the last deals happened | Automate first | Leave manual | A person is next when |
|---|---|---|---|
| Founder closes | Log, remind, one follow-up that stops | Volume research and the pitch | Same buyer, calendar full |
| Form or demo request | One record, an owner, a same-day task | A score with no labeled examples | Replies wait, tasks done |
| A list you build | Dedupe, blank on a failed lookup, suppression, exits | Send volume | Checked copy gets replies |
| A product signup | Split sales-worthy signups, flag a missed setup | Upgrade prompts | People reach value, a match still needs a person |
| Several buyers, long cycle | Pre-call brief, stale-deal alert, handoff note | The negotiation | Founder cannot both prepare and attend |
Index Ventures keeps founders selling while you learn who buys: spokesperson, a senior buyer, the learning, and precious leads that should not go to an unproven seller. HubSpot’s CMO, quoted there, says to introduce sales once the buyer and the user differ, past a contract-size bar in that chapter. RevPartners lists a partner program and a community as further ways to create pipeline. Neither one assigns the lead already sitting with you.
Ten records by hand are the specification
Write the path on ten real records, including one that fails, before any workflow runs. HeyReach says to walk a lead from one owner to the next without automation. If that walk is messy, the workflow amplifies it.
HubSpot’s sales automation guide requires, before the first workflow, duplicates removed, stages that match how you sell, consistent source, territory, and deal type, written routing, and the compliance posture for that market written down, before any outbound sequence. A US list and a Dutch or other EU list do not share one rule. Write why you may email the file, and do not start if you cannot. The knowledge-base page, updated 3 August 2026, only says to segment with a list and to test a workflow before you turn it on. The cleanup list is not on that page.
Name the trigger, the one system allowed to write each field, HeyReach’s owner, deadline, and fallback, and the exception (enrichment down, a duplicate, no owner). A no-show creates a task, not a new pitch. See when a small company should automate. If ten records are the whole month, keep the sheet.
Finish these jobs before you post a role
A row is done on a live record, not when an app is connected.
| Order | Job | Done when | Still a person |
|---|---|---|---|
| 1 | Stages and fields | One sentence per stage, one writer per field, duplicates blocked | Who the offer is for |
| 2 | Capture | Form, inbox, or signup makes one record | The offer on the page |
| 3 | Enrichment | A trusted source may fill a field. A miss stays blank | Research on one account |
| 4 | Route | High intent has an owner and a same-day task | The conversation |
| 5 | Follow-up | Reply, booking, unsubscribe, or bounce stops every channel | Buyer-facing lines, until a clean week |
| 6 | Meeting | Note, stage, and next task land on the record | Discovery and any commitment |
| 7 | Hygiene | No missing next step, stale stage, or overdue task | Keep or kill the deal |
| 8 | Handoff | Owner, context, and a short checklist move with the record | The relationship |
| 9 | Reporting | One weekly file, each word defined, totals you can recompute | The decision |
| 10 | Onboarding | A missed setup step or a usage drop creates a task | The save and the upgrade |
On row 3, a site visit, job change, funding event, hiring post, product use, or closed-lost record may create a task. None may enroll a sequence until that same event has already produced a reply you kept. On row 9 a model may draft the sentence and may not invent the total. See automate marketing reporting. If the customer cannot reach value alone, the next seat is support. Index calls heavy early support hiring, ahead of automation, debt that is sometimes right. Validate the ICP before a workflow enrolls anyone.
A score waits until a person has used it
Turn scoring on only after a person has labeled real records against a definition you can read aloud. HubSpot’s lead scoring tool, updated 2 September 2026, separates fit (stored properties such as title or company size) from engagement (visits, subscriptions, clicks). A combined score needs both. Treat an email open as weak. HubSpot lists it as an engagement example, and an open is not a person asking to talk. Some inboxes hide a real open, and Apple Mail can record an open nobody made.
Apollo’s three questions gate any step: the work repeats, the data is clean enough to act on, and a person can override it. Their table marks discovery and negotiation as a poor fit, and outreach personalization as a draft a human still sends.
The sequence stops, or it does not start
A follow-up that cannot stop is not ready. HubSpot’s guide sets the exit before activation: a reply, a meeting, a stage change, or a manual removal. Test one internal contact through every step. HeyReach’s failure case is a Monday meeting that still produces a Tuesday invite.
AI lead research and outreach keeps a person on the send until a sample week is clean. Octave warns that identical outreach at volume burns the market you still need. Their ladder ends in an agent that sends alone. Stop at the human-reviewed draft. Version every workflow, keep a way back, and name an owner.
Leave judgment in the conversation
Keep a person on positioning, discovery, the relationship, negotiation, the close, and any case with no written rule. Octave’s not-yet list is negotiations, strategic relationships, novel situations, and high-stakes decisions.
A person confirms any stage that records a commitment. Operationalizing founder-led GTM logs, reminds, and prepares. The founder still hears the objection. On the last twenty high-intent records, measure the delay from the action to an owner plus a task. If you cannot explain that delay, it is the leak.
Add a person when the queue is conversations
Before anyone starts, write sequences from outbound that already got replies, give each account one owner (territory or round-robin, not both), and build the first-month report so meetings from the list, the copy, and the activity stay separate. Hand over the buyer, the lines you approve, what makes a meeting worth booking, and the handoff. A model may classify a reply, draft a note, or extract a task. It may not set the lifecycle stage, disqualify a record, or write the forecast.
Hire when the failure queue has been empty for a week and the conversation queue, on your written definition, still grows past one person. The failure queue is enrichment down, a duplicate, no owner, or a sequence that did not stop. Conversion on the definition has to hold, or the path is throwing deals away.
Index says not to hire a senior seller before the founders have run the motion. Early Zendesk sellers on that page needed a segment, or they took self-serve customers. Joe Cross, quoted there, says a generalist makes the first channel test unreadable, so use a specialist and do not hire them before the channel is proven. Index caps that test at three months. When to hire the first growth marketer is that seat, after this path exists.
Do not turn it on in these cases
Leave the build off when you cannot name the buyer, when each conversation still changes the offer, when nobody can approve a buyer-facing line this week, or when the workflow is only a notification. Leave it off when an unsubscribe fails to stop every channel, when the path to value is the constraint, or when the request exists to clean the CRM. Self-serve, a partner motion, or a community assumes trust you do not have yet.
What I would turn on before anyone new starts
I would turn on one path that stops on a reply, and keep prompts, keys, and notes in your accounts.
I am one person, an AI-native marketer, under the name Poldermarketing, and I work fully remote in Dutch and English. You can hire me for a bounded project or for a few days a week. For freshly funded startups anywhere, this is an all-in GTM solution: marketing, AI, and automation, from the first message to the first customers. I build and run that work. I do not only advise. I am strong in AI, content, automation, and workflows. Google Ads and Meta Ads are newer for me. I can set them up and review them.
The free growth scan shows how the current site reads. How I work is the shape of an engagement. Positioning and messaging is where I start when the sentence is the constraint.
Questions people ask
Should I automate outbound before I hire an SDR?
Not the send. Automate dedupe, a lookup that stays blank when it fails, suppression, and a stop on reply, booking, or unsubscribe. A person still checks the message until a sample week is clean. Founders stay in the conversations while the offer is still moving. Hire when those replies are waiting and list work is no longer the delay.
Does lead scoring replace a qualification call?
No. A score only repeats a definition a person has already used on real records. HubSpot's tool splits fit, which reads stored properties such as title or company size, from engagement, which reads actions, and a combined score needs both to be meaningful. An email open is one of their engagement examples, and it is a weak signal. If the properties are inconsistent, the score routes noise faster.
What if there is no CRM yet?
A single sheet with the same columns is enough to walk ten records and write the exceptions. Naming the stage, the owner, and the next action matters more than the logo on the database. Automate after that sheet has survived a week of real leads. A second tool that also writes the same field is how duplicates begin. Put the system of record in your company's name before anyone builds on it.
Can an AI agent replace the next GTM hire?
It can classify, draft, and extract inside a rule you wrote, with a person able to override the result. That is the test Apollo publishes: the step repeats, the data is clean enough to act on, and someone can review or stop it. Negotiation, a strategic relationship, a case you have not seen, and any high-stakes decision stay with a person. An agent that sends on day one fails that test.
When is the hire the right next step?
When one path has run for a week, the failure queue is empty, and one person cannot hold the conversations that already match your definition. It is also the right step when customers cannot reach value without a human, so the seat is support, or when founder-led sales is landing the same kind of customer and the calendar is the limit. A messy CRM is not that moment.