Copilot or Autopilot or Full Robot for Your Marketing
/ 9 min read
Start in Copilot, where AI proposes and you approve every action, until you would have approved about nine in ten proposals unchanged. Then move to Autopilot with tight guardrails such as a daily spend cap and a 10 to 15 percent limit on budget shifts per day. Hand over the whole plan in Full Robot only when a hard monthly cap, written stop rules and revenue reporting are in place.
The question most owners ask about AI marketing is not whether it can write an ad. It can. The real question is how much you let it do before you look. Too little autonomy and the tool becomes another inbox you have to clear every morning. Too much, too early, and a mistake in targeting or budget can run for days. This guide gives you a way to choose the right level for your business today and a safe path to widen it over time.
The three autonomy levels explained
MarketingRobot uses three named levels, and the same ideas apply to any AI tool that touches your marketing.
Copilot means the robot proposes and you decide
Every action, whether a new campaign, a budget shift, a social post or a newsletter, lands in an Approvals inbox. Nothing goes live until a person approves it. You get the analysis and the drafting without giving up any control. The cost is your time: you review everything.
Autopilot means the robot acts inside guardrails
You set limits once: maximum daily spend, maximum budget shift per day in percent, which channels the robot may use and which brand words it must never use. Actions inside those limits run without approval. Anything outside them, such as a shift larger than your limit or a new channel, goes to the Approvals inbox as in Copilot.
Full Robot means the robot runs the plan within a hard cap
The robot manages the whole marketing plan and budget across channels, inside a hard monthly cap and stop rules you define. You get a weekly report with revenue attribution instead of a stream of approval requests. You still keep a kill switch on every screen, and every action is logged and reversible.
The marketing copilot vs autopilot page shows how each level looks inside the product.
Which level fits your business today
The right level depends less on how good the AI is and more on your situation: how much money is at risk, how sensitive your brand is and how much time you have. Use this table as a starting point.
| Your situation | Suggested level | Why |
|---|---|---|
| First month with any AI marketing tool | Copilot | You need to see how it reasons before you trust it with money |
| Regulated field (health, legal, finance) or strict brand rules | Copilot for content, Autopilot for budgets later | Wording risk is higher than budget risk |
| Ad spend under about $1,500 a month | Copilot or narrow Autopilot | Few decisions per week, review takes little time |
| Ad spend $1,500 to $15,000 a month, owner short on time | Autopilot | Daily budget work pays off and guardrails limit the downside |
| Several channels, clear goals, CRM or store revenue connected | Full Robot | The robot can judge actions by revenue, not clicks |
| Agency managing several client brands | Autopilot per brand, Full Robot for mature accounts | Different clients tolerate different risk |
Two signals tell you a level is wrong. If you approve almost everything without reading it, you are paying for Copilot with your attention and getting little in return; move up. If you keep rejecting or reverting actions, the guardrails or the goals are not right yet; move down a level and fix them first.
Guardrails with real numbers
Guardrails turn trust into arithmetic. Here is a worked set for a business with a $4,000 monthly ad budget split between Google Ads and Meta Ads and a target cost per acquisition of $80.
| Guardrail | Example value | How it was set |
|---|---|---|
| Maximum daily spend | $150 | $4,000 divided by 30.4 days is about $132, plus roughly 15 percent room for strong days |
| Maximum budget shift per day | 15 percent | Large enough to act on a clear winner, small enough that ad platforms keep learning |
| Channels allowed | Google Ads, Meta Ads, email | LinkedIn and new channels need approval |
| Blocked brand words | "cheapest", "guaranteed", competitor names | Legal and brand rules the robot must never break |
| New campaign budget | Up to $30 a day without approval | A new test can lose at most $210 in a week before review |
Look at what the 15 percent shift limit means in money. If a Google campaign spends $60 a day, the robot can move at most $9 a day from it in one step. Over a week, steady shifts could take that campaign from $60 to roughly $19 a day if it keeps losing, which is fast enough to stop a leak and slow enough that a single bad day does not gut a good campaign.
Stop rules that end bad spend
Guardrails limit how far the robot can go. Stop rules decide when it must stop. Write them before results arrive, so neither you nor the software argues with the data afterward.
- Zero-conversion stop. Pause any campaign that spends two times target CPA, $160 in this example, with no conversions.
- Slow bleed stop. Pause or cut by half any campaign whose CPA stays above 1.5 times target, $120 here, for 7 days with at least 20 conversions, so the verdict rests on enough data.
- ROAS floor for stores. With a 40 percent gross margin, break-even ROAS is 2.5. Cut spend on any campaign below 2.0 for 5 straight days.
- Spend spike stop. If daily spend runs more than 30 percent above the trailing 7 day average, hold all increases and send a notice for review.
- Monthly hard cap. When month-to-date spend reaches the cap, all paid campaigns pause until the next month or until you raise the cap yourself.
- Tracking stop. If conversions drop to zero across all campaigns for 24 hours, assume tracking broke, freeze changes and alert you, instead of cutting every budget.
The last rule matters more than it looks. A broken tracking tag can make every campaign look like a failure. A careless automation would pause everything. A good one recognizes that a sudden drop to zero everywhere is a measurement problem, not a marketing one. More on how MarketingRobot pauses campaigns is in the post on turning off losing Google Ads campaigns automatically.
A staged path from Copilot to Full Robot
Moving up a level should be a decision based on evidence, not on how busy you feel this week. Here is a path most small businesses can follow over about three to four months.
Stage one is Copilot for four weeks
Approve or reject every proposal. Keep a simple tally. Say the robot sends 120 proposals in the first month and you approve 108 unchanged, edit 9 and reject 3. That is a 90 percent unchanged approval rate. Read the rejections closely: they show where the goals or the brand rules need to be clearer. Move on when the unchanged rate stays at or above about 90 percent for two weeks in a row.
Stage two is narrow Autopilot for four weeks
Turn on Autopilot for budget work only. Set the daily spend cap, a budget shift limit of 10 percent a day and the stop rules above. Keep content, new campaigns and email in Copilot. Review the activity feed twice a week. Count how many automatic actions you would have reversed. If that is fewer than about 1 in 20, the guardrails are doing their job.
Stage three is wide Autopilot for four weeks
Raise the shift limit to 15 or 20 percent a day, allow new test campaigns up to a set daily budget and let SEO articles, social posts and newsletters publish without approval as long as they stay inside your brand rules. Keep anything involving new channels or offers in the Approvals inbox. Connect your CRM or store if you have not yet, because the next stage depends on revenue data.
Stage four is Full Robot with a hard cap
Set a hard monthly cap a little above your normal spend, for example $4,500 on a $4,000 plan, so the robot can respond to a strong week without exceeding what you can afford. Keep every stop rule. Switch your attention from approvals to the weekly report: spend, revenue by channel, actions taken and what the robot plans next. If revenue tracking shows a channel losing money for two weeks in a row, the robot should already have cut it; if it has not, step back to Autopilot and check the rules.
| Stage | Duration | Autonomy | Move up when |
|---|---|---|---|
| 1 | 4 weeks | Copilot, everything approved | 90 percent of proposals approved unchanged for 2 weeks |
| 2 | 4 weeks | Autopilot for budgets, 10 percent shift limit | Fewer than 1 in 20 automatic actions reversed |
| 3 | 4 weeks | Autopilot for budgets and content, 15 to 20 percent shift limit | CRM or store revenue connected and reports match your books |
| 4 | Ongoing | Full Robot with hard monthly cap and stop rules | Weekly review only, step back if rules are breached |
What you keep control of at every level
Autonomy is not abdication. Even in Full Robot, several things stay with you.
- The goal and the budget. The robot works toward the target and inside the cap you set.
- The kill switch. One click pauses the robot on any screen, and you can revert the last action.
- The record. Every action is logged in the activity feed with what changed and why, and each one is reversible.
- The accounts. Campaigns live in your own Google Ads and Meta Ads accounts, content on your own site, email in your own Mailchimp, Klaviyo or Brevo account. Access is scoped per channel and you can disconnect any time.
- Who can approve. On Pro and above, seats have roles (owner, approver, viewer), so a junior team member can watch without changing anything.
Security details, including encrypted tokens and SSO on the Scale plan, are on the security page.
What each level costs
Autonomy levels map to plans. Copilot is available on Starter at $199 a month or $99 a month billed yearly. Autopilot with guardrails and all eight robot modules come with Growth at $499 a month or $249 a month billed yearly. Full Robot with a monthly budget cap and stop rules starts on Pro at $1,499 a month or $749 a month billed yearly. The subscription is never a percent of ad spend; you pay Google and Meta directly. Compare plans on the pricing page.
A common approach is to start on Growth even if you plan to stay in Copilot for the first month, because the move to Autopilot is then a setting rather than a plan change. If you are still deciding which channels to fund, read how to split a marketing budget by channel first, and if you are comparing this approach with hiring people, the AI marketing team page lays out the trade-offs.
Frequently asked questions
Is it safe to let AI spend my ad budget?
It is as safe as the limits around it. With a maximum daily spend, a cap on budget shifts per day, a hard monthly cap and written stop rules, the most an error can cost is known in advance. Start in Copilot so you see the robot's decisions before it acts alone.
How long should I stay in Copilot mode?
Usually about four weeks, or until you approve roughly 90 percent of proposals unchanged for two weeks in a row. Some businesses in regulated fields keep content in Copilot permanently while letting budgets run on Autopilot.
What happens when an action falls outside my guardrails?
In Autopilot it goes to the Approvals inbox instead of running. For example, if your shift limit is 15 percent and the robot wants to move 25 percent, you see the proposal with its reasoning and decide.
Can I go back from Full Robot to Copilot?
Yes. The autonomy level is a setting you can change at any time, and the kill switch pauses the robot instantly. Past actions stay in the activity feed and can be reverted.
Which plan do I need for Full Robot?
Full Robot mode is part of Pro and Scale. Pro costs $1,499 a month or $749 a month billed yearly and includes 3 brands, 900 robot tasks a month and 5 seats with roles.