The Mistake I Almost Made in CRM by Relying Solely on AI

In the modern world, we increasingly rely on artificial intelligence, and I am no exception. While developing the CRM for formspace.design, I encountered a situation where I almost made a serious mistake by fully trusting AI's recommendations. This experience clearly showed me that despite the power of neural networks, the human factor, deep contextual knowledge, and critical thinking remain indispensable.

How I Came to Automation and CRM

My journey in landscape design and construction began in 1997, and from 2000 to 2022, I managed the company 'Dom i Sad' (House and Garden), which designed and built cottages, landscaped plots, and maintained them. Throughout these years, I keenly felt the lack of effective tools for project and client management. Off-the-shelf CRM systems often failed to account for the specifics of the landscaping business: its seasonality, multi-stage projects, the need for visualization, and in-depth detail. This led to manual record-keeping, loss of some information, and inefficient planning.

That's why, starting to code at 61 with the help of AI, I decided to create formspace.design — a tool I so desperately needed. My goal was not just to automate routine tasks, but to build a system that would help landscape designers and property owners make informed decisions, structure information, and interact effectively. I have previously detailed why ready-made Western CRMs are not suitable for the CIS landscaping business, and this became the starting point for creating my own solution.

The Trap of Complete Trust in AI

When I was designing the deal management module for my CRM, I turned to AI for help in structuring the sales funnel stages. I gave it general input about the landscaping business and asked it to suggest standard stages. The AI generated a perfectly logical sequence: 'New Lead,' 'Qualification,' 'Proposal,' 'Negotiation,' 'Closed (Won/Lost).' At first glance, this seemed reasonable and aligned with general CRM principles.

However, delving into the details, I realized that key stages specific to our business were either missed or combined. For example, for landscape design, 'Site Visit and Measurements' is critically important, followed by 'Concept Development,' and only then 'Detailed Design and Estimate.' These stages are not just intermediate steps; they are often paid services or require significant time and resource investment. If I had blindly accepted the AI's structure, my CRM would not have reflected the real state of affairs, and I would not have been able to effectively track progress, issue invoices, and analyze profitability at each project stage.

Why AI Isn't Always Right: Context and Nuances

This incident became a vivid lesson for me. AI is trained on vast datasets and can offer statistically probable and generally accepted solutions. But it lacks a deep understanding of the context of a specific niche, particular business processes, and unwritten industry rules. It doesn't know that a site visit to a 15-acre plot isn't just a checkbox, but hours of work involving measurements, photo documentation, and client communication, which can cost 5000-10000 rubles and is a separate service.

AI does not possess 'common sense' or intuition, which are developed through years of practical experience. It doesn't understand why, for a landscape project, it's important to consider soil characteristics, utility locations, sun exposure, or climatic zones. All these nuances, which are obvious to me as a practicing specialist, remain outside the AI's current context unless explicitly detailed in the prompt.

Principles for Working with AI in Automation

After this incident, I developed several key principles for working with AI, which I also apply when developing formspace.design:

1. Always double-check: Never accept AI results as the final truth. Use them as a starting point or a draft, but always cross-reference with your own experience and knowledge.
2. Provide maximum context: The more detailed and precise the prompt, the more relevant the answer will be. Don't hesitate to describe details, industry specifics, and your expectations.
3. Use AI as an assistant, not a decision-maker: AI is an excellent tool for generating ideas, automating routine tasks, writing code, or analyzing data. But strategic decisions and final verification must be carried out by a human.
4. Iterate and refine: If the first AI response isn't satisfactory, don't despair. Refine the query, ask leading questions, change formulations. Working with AI — is a dialogue.
5. Break down complex tasks: Instead of asking AI to solve an entire problem, break it down into smaller, manageable parts. This will make verification and correction easier at each step.

Common Mistakes When Implementing AI in Business Processes

My experience shows that many make similar mistakes when relying on AI. Here are the most common ones:

• Blind trust in generation: Accepting AI-generated answers or solutions without critical analysis and verification against business specifics.
• Lack of context in prompts: Expecting high-quality, specialized results from general or overly short queries that lack key details.
• Ignoring the human factor: Attempting to automate absolutely all processes, including those requiring empathy, unconventional thinking, or personal contact.
• Lack of iteration and feedback: Unwillingness to refine queries or correct generated data, leading to the use of suboptimal solutions.
• Overestimating AI capabilities: Believing that AI can solve all business problems without expert involvement and a deep understanding of the subject matter. I have already written about the gap between 'I know how the business should work' and 'I don't know how to code it', and AI, while helpful, does not completely eliminate it.

How to Plan in formspace.design

At formspace.design, we strive to use AI as a powerful tool for creativity and accelerating work, always leaving humans in control of the final result and the ability to contribute their expertise. Imagine you own a plot of, say, 12 ares and want to create a unique landscape design, but don't know where to start.

You can start with Module 2.1 AI Concept. Upload photos of your plot, specify style preferences (e.g.,

FAQ

How to get started with formspace.design?

Open a plot project, define boundaries and zones, then specify objects. This way, you immediately see the structure, not just a 'pretty picture'.

How does formspace.design differ from a one-off image from a generator?

In formspace.design, you work with a plan, zones, and scenarios (concept, photo-design) that can be edited and saved within the project.

Can one do without a staff of programmers?

Yes — if you have subject matter expertise and discipline. I am building formspace.design precisely with the help of AI, relying on practical experience in the landscaping business.

What if AI made a mistake in plants or scale?

Check climate, sizes, and proximity based on the actual plot. The tool speeds up the draft, but the decision remains yours.

How to link the plan with the estimate and the client?

After the basic plan, it's easier to agree on volumes and stages: fewer 'forgotten' tasks and rework on site.