A marketing automation stack for a five-person team
How a five-person team runs marketing like a large one: four robot roles, what stays human, who owns the data, SaaS or owned automation, and a 90-day plan.
TL;DR
A five-person team can run marketing at the cadence of a much larger department if four robots take the repetitive work: a content robot, an ads robot, a communication robot and an operations robot. People keep strategy, creative direction and customer conversations. Everything writes to one database the company owns and can export at any time.
A marketing automation stack is the set of connected robots, one shared database and a small number of written rules that carry out a company's repetitive marketing work without a person pressing the button each time. It is not a pile of subscriptions or a chatbot that writes your posts. It is a division of labour: the machine takes the parts that repeat, the people take the parts that need judgement.
This post covers the four robot roles Guardino runs for clients, what stays human, who owns the data, and a 90-day rollout that does not break what already works. The Türkçe version is written natively, not translated.
Why a five-person team needs a stack, not more hands
A mid-size e-commerce company with five people in marketing usually has one person on ads, one on social, one on email and content, one on the shop, and a lead doing half of the other jobs. Most of each week goes to work that repeats: resizing photos, scheduling posts, shifting ad budget, sending the same order emails, copying numbers into a spreadsheet for Monday.
The volume is real. Instagram brands post a median of 3.69 times per week (Rival IQ, 2026), on one platform. Across seven, the weekly quota is beyond what one person can do by hand. On the email side, automated flows generate about 41% of email revenue (Klaviyo, 2026), so the flows nobody has time to build are the ones meant to pay for the team.
Hiring for repetitive work is the expensive answer. Giving it to robots and keeping the five people on judgement is the one that scales.
The four robot roles
We name robots by role, not by tool; tools change, roles stay. Every robot runs on the agency's own servers, writes to the client's database and has a written scope.
Content robot
The content robot takes raw input, usually product photos and short notes, and turns it into finished posts: cropped per platform, captioned in the brand voice, tagged and scheduled. It posts to 7 platforms, 2 to 3 times a day, which works out to 60 to 90 posts a month. It never invents a product claim; every fact in a caption comes from the product data in the database.
The team's job shrinks to sending photos and approving the weekly batch, about twenty minutes on a Friday.
Ads robot
The ads robot runs nightly. It reads the day's results from the server-side measurement pipeline, not the ad platform's dashboard, and applies the team's rules: move budget toward ad sets under the target cost per result, pause creatives past the frequency threshold, rotate in the next creative from the queue. It logs every change so a person can review it the next morning.
It never launches campaigns or raises the total budget; those two decisions stay human, on purpose.
Communication robot
The communication robot owns the messages a customer expects after an action: order confirmation, shipping update, delivery, a review request after a delay, a win-back message after a long silence. It is consent-aware. A customer who has not opted in to marketing gets transactional messages only. When someone unsubscribes, the robot updates the single customer record and every other robot sees it in the same run.
This is where the Klaviyo figure above lives: lifecycle flows are where a small team's email revenue is either happening or missing.
Operations robot
The operations robot talks to the team, not to customers. At 07:00 every day it sends a short summary: yesterday's revenue, ad spend, cost per result, posts published, emails sent, anything that failed. On Monday at 08:00 it sends the weekly report with the same numbers over seven days, plus what the ads robot changed and why.
The weekly report is the only meeting that needs to happen; everyone reads the same numbers before they talk.
What stays human
Automation fails when it is asked to do the wrong job. Three things stay with people in every stack we build:
- Strategy. Which products to push, which markets to enter, what the price should be, when to run a campaign. Robots execute a plan; they do not set one.
- Creative direction. The look, the tone, the one line that makes a campaign different. The content robot produces variations inside a direction a person gave it.
- Customer conversations. Complaints, real questions, anything with money on the line. A robot can route and prefill; a person replies.
Which of those human tasks stay inside the company and which a specialist can carry is covered in the hybrid marketing model.
Who owns the data
Every robot reads from and writes to one database: orders, customers, consent status, posts, ad results, email events. It belongs to the client, sits on infrastructure the client can log into, and exports in full as plain files at any time.
This matters more than any single robot. Spread across five vendor dashboards, the data shows what five vendors chose to show. In one place the company owns, the reports are the company's own numbers, and switching a tool means changing a connector, not losing a year of history.
No intermediary sits between the client and the platforms either: ad accounts, social accounts and the domain stay in the client's name; Guardino gets access, never ownership. That is the principle behind the marketing automation service: the robots are ours to run, the data and accounts are yours to keep.
Rented SaaS bundles or owned automation
Most small teams start by renting: one subscription for scheduling, one for email, one for ad rules, one for reporting. It works until the fifth tool.
| Question | Rented SaaS bundle | Owned automation |
|---|---|---|
| Where does the data live | Split across each vendor's account | One database in the client's name |
| Can you export everything | Partly, in each vendor's format | Yes, full export as plain files |
| Who writes the rules | Limited to what each tool exposes | The team, in plain sentences, applied by the robot |
| Cost as volume grows | Per-seat and per-contact tiers climb | Server cost stays flat; robots do not charge per post |
| Cross-tool logic | Manual, or through a fifth tool | Native: one consent change reaches every robot |
| Setup time | Days per tool, weeks to connect them | 8 weeks in the Launch program |
Rented bundles are fine for a one-person shop with one channel. They get expensive exactly where a five-person team arrives: several channels, real volume, one truthful number.
A 90-day rollout plan
Switching on four robots at once is how you break a working business. Our order:
Days 1 to 30: measurement and data.
- Stand up the single database and connect the shop, ad accounts and email sender to it.
- Put server-side measurement in place so ad results come from your own events, not a pixel half the browsers block; the event dictionary and weekly reconciliation are in measurement, conversion and retention.
- Record consent status for every existing contact. Nothing gets automated until this is correct.
- Switch on the operations robot first. For a month it only reports, so the team learns to trust the numbers.
Days 31 to 60: communication and content.
- Switch on the communication robot with transactional messages only, then add lifecycle flows one at a time, starting with the review request.
- Switch on the content robot at 1 post a day on 2 platforms; raise it toward 2 to 3 a day once weekly approval is above 90%.
- Write the two-page brand voice document, with examples of what not to say.
Days 61 to 90: ads and handover.
- Write the ad rules as plain sentences: target cost per result, frequency ceiling, minimum spend before a creative is judged.
- Run the ads robot in log-only mode for a week, read its proposed changes every morning, then let it act.
- Hold the first Monday review that runs entirely from the weekly report. If it needs a spreadsheet, fix the report, not the meeting.
Ninety days is roughly one Launch program plus the first month of the monthly program; the process page shows the same sequence from the client's side.
Do not automate a broken step
If order emails are wrong today, the communication robot sends wrong emails faster. Fix the template, sender domain and consent records first. Automation multiplies whatever it is given.
Common failure modes
- Automating before measuring. The ads robot optimises toward a number that was wrong from the start. Fix: the operations robot runs alone for the first month.
- No written scope. Someone asks the content robot to answer comments, and it does, badly. Fix: one page of scope per robot; everything else is a human task.
- Consent as an afterthought. A win-back email reaches someone who unsubscribed two years ago. Fix: consent is a field in the customer record, checked on every send.
- Approval fatigue. The team stops reading the weekly batch and approves everything. Fix: keep the batch readable in twenty minutes, or split it between two people.
Checklist before you switch anything on
- One database in the company's name, with a scheduled export.
- Server-side measurement live and reconciled against the shop's own order count.
- Consent status recorded for every contact.
- A one-page scope per robot and a two-page brand voice document with examples.
- Ad rules written as sentences, with a name and a date.
- A 07:00 daily summary someone reads, a Monday 08:00 report that replaces the status meeting, and a named owner for strategy, creative direction and customer conversations.
What this costs
The Launch program builds this stack in eight weeks for $9,900: platform, single database, measurement, first flows, social setup. The monthly program then keeps the robots and reports running; Model B is $990 a month plus 4% of online net revenue, so the fee moves with results. Both are on the programs page; the social media engine page shows what the content robot produces in a month.
A five-person team does not need to become fifteen to run seven platforms, nightly ad optimisation and lifecycle email. It needs four robots with clear scopes, one database it owns, and the discipline to keep strategy, creative and customers in human hands.
