Why retention starts with a better understanding
If the client doesn't feel that you understand them, they will leave. Even if the product works, even if the price is fair. The reason is simple: today, it's not so much the features that compete as attentiveness to the user's context. And this attentiveness starts not with hypotheses, but with precise, deeply developed customer insights.
Customer retention is not the job of the support department. It is the result of how accurately the product and communication hit real pain points and motivations. That is why the best marketing agencies today start not with visuals, not with channels, but with audience research. Not just "age and position", but how exactly a person makes a decision, what they hope for, what they are afraid of, what barriers hold them back from buying.
The problem is that manual collection of such data takes tens of hours. Interviews, tables, frameworks, validation - this whole process is complex and does not always provide systemic conclusions. Previously, you could spend 150 hours just to get a preliminary picture. Now, with tools like Elsa AI, you can get a structured ICP and marketing strategy in half an hour — and then validate, clarify, and work out the nuances.
The best retention starts when you stop “guessing” who your client is and start working with proven insights: what slows them down, what criteria are important when choosing, what words inspire trust. These details become the basis of the offer, landing page structure, call script, and email chain. When a person sees that you speak their language, they stay.
What agencies get wrong about ICPs
The idea of ICPs has long been mainstream in agency practice. Almost every client brief begins with the question, “Who is your audience?” And almost everyone gets an answer that sounds confident, but does not provide support for decisions: “men aged 30–45, entrepreneurs,” “B2B marketers,” “startup founders.” The problem is that such definitions say nothing about behavior, motivation, pain, and decision-making processes.
The main mistake is to confuse the description of the audience with its understanding. A real ICP is based not on demographics, but on context: what a person does before buying, what alternatives they consider, what they are afraid of, how they learn about solutions, what is “good enough” for them. Without this data, the offer loses its edge, the content loses its relevance, and sales lose their pace.
Agencies often try to fill in the gaps with intuition. Sometimes it works. But more often it leads to the team testing dozens of creatives, manually sorting through segments, and the result does not change. Why? Because the foundation itself is weak. Without a clear ICP, you don’t know who you are trying to convince and which arguments will work.
Bozhidar Batev, one of the users of Elsa AI, directly calls the old approach “shooting in the dark”. His agency used to conduct dozens of interviews with different segments, just to feel out where there is a response at all. Now, after implementing Elsa, the process begins with an accurate hypothesis: who is the target group, what barriers it has, how they think. And casting development, proposal, strategy are laid on this basis.
At M1-Project, we believe that ICP is not a template. It is a research artifact. It can be compact or detailed, but its value is not in the form, but in how much it helps make decisions: in creative, in landing pages, in email, in a call. Everything starts with precision. And accuracy is the result of systematic work with data, not beautiful slides.
How Elsa AI works: from ICP to action
Usually, collecting data on the target audience means months of interviews, dozens of hypotheses, complex tables and semi-structured chaos. Elsa AI automates this process. Not replacing thinking, but enhancing it. It takes on the first, most difficult part - turning disparate input data into a coherent map of the segment, needs, barriers and motivations.
How does it work? It all starts with you specifying the client's website or a brief description of the business. Then the system automatically analyzes the context, compares it with known patterns and segments, generates a Buyer Persona and selects the most relevant Job-to-be-Done, pain points, goals, decision triggers and communication channels.
As a result, you get not a “generative fantasy”, but a working hypothesis with which you can go into the field, validate, clarify, launch offers. The entire report is structured by key sections:
- Customer overview
- Goals, barriers and decision-making triggers
- Buying behavior
- Communication patterns
- Market positioning and marketing strategy (if the corresponding function is enabled)
It is not only important that this data exists. It is important how they are related to each other. For example, if the ICP states that the key trigger for switching to a new solution is the loss of control over processes, you immediately understand that the offer needs to emphasize “controllability” and “transparency”. If a segment avoids solutions with long implementation, you focus in advance on “quick start” and “low entry threshold”.
Bozhidar Batev describes it this way:
“Instead of randomly testing 3-4 segments, we immediately know where there are more matches. And we do not waste resources in vain. We shorten the path from hypothesis to validation by at least 40 interviews.”
What else does Elsa do? After building the ICP, you can automatically generate a marketing strategy and positioning - with a choice of channels, offers, price logic, and even the tone of communication. This is a whole stack: from customer research to messaging.
Client story: From 4K to 40K MRR with Elsa insights
Bozhidar Batev has been working with consultants, coaches, and infobusinesses who want to grow for years — from $5-10K per month to $50K and more. In the past, his team built a strategy intuitively: brainstorming, several castdevs, trying to “feel” a response. At best, it worked. At worst, it burned weeks on inappropriate segments.
With Elsa, the process changed. In the first ICP he created for himself, the report clearly highlighted two insights: clients desperately need control over their lives and suffer from impostor syndrome. These two signals became central to his proposal: he stopped “training from above” and gave people tools for independent movement. A less directive approach means more involvement. The result was not long in coming.
One of his clients, who had been stuck at $4K per month for years, reached a stable $40K+ 6 months after launching a new strategy built on Elsa. And another one — from $20K to $160K in 10 months.
“We used to do 10 interviews for each segment to understand where there is a response. Now we start with a clear hypothesis and immediately validate it. This saves at least 120-150 hours of work.”
But it’s not just about saving. When you get a finely crafted client profile, you build an offer differently. You write a landing page differently. You don’t try to “explain the product” — you get into real fear, expectation, block. You know which words inspire trust, which ones repel. This is evident in his approach: each new ICP is used as a basis for sales scripts, advertising videos, email structures, and entire training guides for other AI tools.
Bozhidar doesn’t limit himself to one part of the report. He uses everything from segment and barriers to tone of voice. Within his agency, this information becomes the basis for a whole stack of AI solutions. And each element of this system works towards one goal — to speak to the client in an understandable language.
Tangible outcomes: Saved time, clearer targeting, faster content
When you implement AI in customer research, the main metric is not the report, but the result: how much time you saved, how much more accurately you hit the segment, how much faster you started acting. In Bozhidar’s case, the answers to these questions are clear:
Before Elsa: 40–50 interviews, up to 150 hours for preparation, casting and analysis.
After Elsa: 20–25 interviews, at least 2 times less time, more confidence at the start.
And that’s just research. The second front is content creation. When you already have insights on pain, triggers, motivation and tone of voice, you don’t generate ideas from scratch. You just compose them into specific formats. Inside the Bozhidar team, Elsa reports became input data for generating:
- landing page copyright
- video scripts
- email chains
- Instructions for other AI tools
“We create detailed guides on how to talk to this audience: what words to use, what topics to raise, what length of content will work best. This saves not hours, but days.”
A separate point is the message and the offer. Most infobusinesses and agencies confuse the product and the offer. But it is on the basis of the ICP that a strong offer is built: not an abstract “access to the platform”, but a specific solution to a specific pain in an understandable form. This allows you not only to improve conversion, but to make a sale on the first call - without warming up, without additional content.
“I sold a cold call for $300K simply because I knew which triggers work. Without Elsa, this would not have happened - because I would have spoken to the person in my language, not his.”
When insights become actions, systems are born. Bozhidar calls it a predictable business machine. Each report is another step toward a repeatable process that produces results.
How to validate AI insights with real people
Any data is an assumption until you check it against reality. This is where Bozhidar’s approach differs from most: he doesn’t blindly rely on AI, but he also doesn’t waste resources on haphazard research. Instead, he has a structure. After receiving a report from Elsa AI, the team immediately moves on to validation. And they do this in two stages.
The first stage is qualitative interviews.
25 calls, 1 on 1. No scenarios in the spirit of “what would you like from the product”. Instead, specific hypotheses from the report, which are tested for reaction. How does a person react to the formulation of pain? Do the words they use match the language from the ICP? Does they feel that they are truly understood?
Bozhidar puts it simply:
“If you don’t know what to ask, interviews will not give you anything. Elsa helps you ask the right questions. We don’t ask blindly — we validate hypotheses.”
The second stage is quantitative validation.
After the interview, a questionnaire is created. Based on the patterns identified in the first round. This provides not only confirmation, but also prioritization: what pains are the most common, what selection criteria are critical, what insights are most common. The result is an understanding of what is really important to the market, and what can be omitted from communication.
This approach is very different from typical inspirational marketing. There is no place for guesswork. There is a hypothesis, a test, and adaptation. This does not replace creativity. It gives it support.
AI does not cancel contact with real people - it makes it more accurate. You do not start from scratch. You do not spend weeks trying to find the wording. You come with a version that already matches reality by 70%, and the remaining 30% is a matter of adjustment, not saving the project.
Smarter inputs, stronger outcomes
Customer success is not the result of inspiration. It is the result of precise work with expectations, barriers and needs. When you start with the right data, you do more than just save time. You eliminate chaos in strategy, content and communication.
Elsa AI does not do magic. It provides support: a point from which to build a system. The insights you receive turn into specific actions - and the more structured the team, the faster they produce results.
The saved hours are only a superficial gain. The main thing is focus. Instead of wasting effort on dozens of wrong directions, you focus on one, the most accurate. And this is what gives growth. Fast. Consistent. Predictable.
If you are building an agency or developing clients in the infobusiness, everything starts with the right ICP. And with the decisions you rely on it.