The First 14 Days With an AI Setter: What to Check Every Day

In the first 14 days with an AI setter, you check one thing every day: where conversations stop. Block 45 minutes a day, read the new conversations, find the points where leads get confused, frustrated or simply go quiet, and make two or three changes to your instructions. Start with the cold stage at the top of the funnel. Only change something when at least 30 conversations break at the same point, test each change three times, and leave A/B tests until the 14 days are over.

That is the whole plan. Below is how I run it with clients, step by step. In the video above I go through the same process in the app.

Block 45 minutes a day in your calendar

Do it now, before you read further. Put a recurring 45 minute slot in your calendar for each of the next 14 days. This is the part people skip, and it is the part that decides whether the setter ends up running on its own or stays a half-finished experiment.

What you do in those 45 minutes is simple:

1. Open the conversations the setter ran since yesterday.
2. Read them and note where people stopped replying, misunderstood a question or got annoyed.
3. Pick two or three changes and make them in your instructions.
4. Test each change before it goes live.

The conversation list only shows threads where someone replied to the welcome message. You are not reading through hundreds of empty threads, only the ones with something to learn from.

Two or three changes a day is the right pace. Fewer and two weeks is not enough to get through the problems. More and you lose track of which change did what.

Start at the top of the funnel: cold before warm

Every contact in the panel has a stage: cold, warm, hot and converted. Spend most of your attention on cold, then warm, then hot, then converted, in that order.

The reason is arithmetic. Everything that happens at the cold stage affects everything after it. If you lose half your leads in the first few messages, a better booking message helps only the half that got that far. Fix the start of the conversation first, and every later fix works on more people.

In practice, the first week is mostly about the opening and the first two or three questions. If you have not worked on the first message yet, start with the Instagram DM opener guide.

What friction looks like in a conversation

Friction is any point where a conversation gets harder for the lead than it needs to be. You will see three kinds:

• People do not understand the question. They ask what you mean, or answer something else.
• People get frustrated. Short, irritated replies, or a direct complaint about being asked too much.
• People stop replying. This is the most common one, and the easiest to miss.

Pay particular attention to one pattern: the lead answers the first message, answers the second, and then the conversation stops. When that happens again and again at the same question, the question is usually the problem.

The 30 conversation rule

Here is the rule I hold myself to: one dropped conversation is not a pattern. Neither are two or three. You might see three conversations stop at the same question in one morning and want to rewrite it. Don’t.

Draw a conclusion only when you see at least 30 conversations that stopped at the same point. Before you count one, check that the person had enough time to reply. A lead who has not answered in two hours may still answer in the evening.

Once you have 30 or more, you have probably found a bottleneck. A common cause is a question that goes too deep, too early. Think about it from the lead’s side. You are a stranger from the internet, you have exchanged two messages, and you ask:

What’s your biggest struggle with your weight right now?

For many people it is easier not to reply than to answer that honestly. When 30 conversations stop right after a question like this, the question is causing friction.

Two ways to fix a question that stops conversations

Soften it. Ask the same thing with less pressure. Instead of asking for the biggest struggle, ask whether they have noticed any challenges:

Have you noticed any particular challenges with your goals lately?

Ask an A or B question. Suggest possible answers, so the lead only has to pick one. For example:

A lot of the people I talk to say they eat about the same as before, but it’s harder to keep things where they want them, or they snack more in the evenings. How is it for you?

Or, more directly:

Is your goal right now more about losing weight, or more about building muscle?

The same works in any niche. A business coach might ask whether someone wants to tighten their sales process or bring in more clients first.

The answers you suggest should come from your conversations, not from your head. Read what people say most often, and when the same problem shows up in 30 different conversations, it is a good candidate for an option in an A or B question. The 30 conversation rule applies here too.

Write down what you see

Keep one document for these observations. For each one, add a screenshot or the lead’s Instagram username and a sentence about what you noticed. Each day, write down your two or three main observations.

After a week, this document is worth more than your memory. You can go back, see which problems keep coming back and check whether yesterday’s change actually helped.

Test each change three times before it goes live

When you are confident an observation is real, take the conversation where it happened into the test chat. Go back to the message where things went wrong, change the instructions and replay that moment as the lead.

Do this three times for each change. If the setter handles the moment the way you want three times in a row, the change is done and you move to the next problem. If it does not, keep going. Sometimes a fix takes a few attempts.

When a fix will not hold, three things help, and I cover them in more detail in how to write instructions for an AI DM setter:

• Use AI to diagnose why the setter behaves the way it does.
• Use AI to find which parts of your instructions are not precise enough.
• Check for contradictions: two rules that pull in opposite directions, or a gap the setter fills with a guess.

Over 14 days, these small fixes add up. Each one closes one gap, and by the end there are far fewer gaps left.

Why conversion drops after the first few days

On the dashboard, you will probably see conversion start high and then fall. I call this the beginner’s peak. It does not mean the setter worked well and then got worse.

Two things cause it. First, the setter starts with your newest followers, who followed recently because they have a problem they want to solve. They are your warmest leads. Second, a conversation only appears once someone replies, and the most engaged people reply first. So the first days are skewed upward.

In my experience, conversion then settles at around 50 to 70% of where it started. That is your real baseline. Over time, as the setter writes to people who followed you longer ago, it can keep easing down slowly. Follower number 5,000 is often less engaged than follower number 1, and on large numbers you will see it. If new followers keep coming in faster than the setter writes to them, the line will more likely stay flat.

So do not panic on day four. Judge the setter on where conversion settles, not on the peak.

After 14 days: start A/B tests

Do not start A/B tests before day 14. A test only makes sense once your instructions are polished, when you read the conversations and no longer see much to fix.

Then the process looks like this:

1. Your current instructions are version A.
2. Duplicate them as version B and change one meaningful thing. For example, the first two questions, a more direct approach to selling, a more empathetic tone, a few more example messages, or emojis.
3. Test version B in the test chat, then put it live next to version A.
4. New leads are split between the two versions, and each lead stays on their version to the end of the conversation.
5. Wait for at least 200 to 300 conversations on each version before you compare.

Say version A converts at 5% and version B at 7.5%. Turn A off. Version B is now your new baseline. Duplicate it, change one more thing and start the next test.

That jump means that for every 100 calls the old version booked, the new one books 150, from the same followers. You did not spend more on ads or post more content. The process just got better.

Once the setter is tuned, the work drops to about 30 to 45 minutes a week. Check whether the running test has reached 200 to 300 conversations per version, and when one wins, start the next test right away.

The 14 days at a glance

FAQ

How long does it take to optimize an AI setter? Plan for 14 days at about 45 minutes a day. After that, the setter needs roughly 30 to 45 minutes a week to check A/B tests and start the next one.

How many changes should I make per day? Two or three. That is enough to make progress and few enough to know which change did what.

When should I start A/B testing my AI setter? After the first 14 days, once you no longer see much to fix in the conversations. Collect at least 200 to 300 conversations per version before you pick a winner.

Why did my conversion rate drop after the first week? The setter writes to your newest, warmest followers first, and the most engaged people reply first. Conversion usually settles at around 50 to 70% of the early peak, and that is your real baseline.

Where the numbers come from

• 45 minutes a day, two or three changes, 14 days, 30 conversations, three tests per change, 200 to 300 conversations per A/B version, 30 to 45 minutes a week: my own practice with Setor clients, not a study.
• Conversion settling at 50 to 70% of the early peak: what I usually see after launch. It depends on how many new followers come in.
• 5% versus 7.5%: an example to show the arithmetic, not a client result.
• Each lead staying on one version during an A/B test: how Setor’s script tests work.

Want to see your own conversations by stage? Start now and let the AI appointment setter run your DMs. Before day one, read how to write instructions for an AI DM setter.