Almost every conversation about software starts with the price and ends somewhere else. Someone asks for a quote, gets it, and the doubt that surfaces is not whether the number is high: it is whether the person on the other side actually understood what they need, and whether they will still be there in six months.
In the first episode of Escala sin Miedo, the MasterDragon podcast, Alejandra Zuluaga interviewed Marcel Verand, founder of Business Clarity, who has already walked the full path. This article collects what he said, organized by the three moments where the fears show up: before hiring, during the build, and after delivery [1]. The episode is in Spanish and the quotes below are translated.
What three paths exist when you decide to solve something with software?
When an operation hits a ceiling, exactly three options appear. They are not equivalent and each one fails differently.
| Path | What it gives you | What it costs you |
|---|---|---|
| In-house team | Full control and knowledge that stays with you | Hiring, managing and waiting before the first result |
| Outside vendor | Experience accumulated on similar problems | Depending on them to understand you and deliver |
| Off-the-shelf tool | Speed and a low entry price | Working the way the tool wants, not the way your process needs |
Marcel evaluated all three. He had been using AI agents built inside existing platforms, assembled by himself, and they worked. The problem was not that they did not work.
"You can have those AI agents that work well because they are very good, but they live inside a platform. It does not give me the freedom to create a subscription, for example."
That sentence is the practical distinction between the first two rows of the table and the third. An off-the-shelf tool solves the task and keeps the platform. If what you want to build is your own product on top of that task, someone else's tool is a ceiling, not a shortcut.
Why does an off-the-shelf tool fall short when the process differentiates you?
The rule is short: if your process is part of what makes you distinct, a generic tool will fall short.
Marcel's case shows it well because his business is high-ticket one-to-one mentoring. What he built was not a support chatbot, it was an extension of his own judgment for clients who already pay for his time.
"This tool replaces Marcel Verand 24/7, though not completely. It gives my clients the option of an immediate answer at any moment."
Here it is worth being honest about the data, because this is the part a case study usually hides. Marcel said on camera that he has not yet made subscription sales, and that the return so far is not new revenue but the value it adds to the mentoring he already sells. That is a real return, and it is different from the one a pitch promises.
And why not build it yourself with AI?
This is the question that barely existed two years ago. Today you ask a tool to build you an application and something working appears in minutes.
The honest answer is that you can, and that the point where it stalls is predictable. AI hands you the visible 80% fast: the screen, the flow, the demo. The missing 20% is your complete process with your real data and your real customers, and it is precisely the 20% that decides whether the thing is useful.
There is one figure worth holding in mind before deciding to go it alone. Gartner projects that more than 40% of agentic AI projects will be canceled before the end of 2027, driven by escalating costs, unclear business value and inadequate risk controls (2025) [2]. None of those three causes is a model problem. All three are judgment problems about what to build and how to sustain it.
How do I know they will understand my business?
This is the most human of the three fears, and the one Marcel describes most precisely. Nobody knows your operation the way you do: the procedures, the exceptions, the details you learned the hard way and that are written in no manual.
What he did is simple and repeatable: ask a lot before signing.
"I was very rigorous in the meeting I had with Neo, asking him a lot of questions."
The signal is not what the vendor shows you, it is what they ask you. A team that will understand your business asks about your exceptions, about what happens when the case is not the normal one, and why you solve it today the way you solve it. A team that only receives a requirements list will build the list, not the business.
What if I pay and they do not deliver?
The blow many have already taken. They paid, nobody answered, or they got something half-built, and after that trusting again is expensive.
What Marcel says gave him confidence was not a contract or a certification.
"A person, a company can have a good reputation, but it is the people inside the company who give you confidence. You end up buying people more than companies."
It is an uncomfortable answer for anyone who wants a checklist, and it is the answer of someone who already got it wrong once. Institutional reputation tells you the company exists. Direct contact with whoever will answer for the project tells you whether they will answer.
What happens after they hand it over?
The most overlooked fear, because it arrives once the project already went well. You have the application running and four questions appear: who provides support, where it is hosted, who owns the data, and what happens if tomorrow you want it to evolve.
In the episode, Marcel says he asked for a change that meant new work and the response was fast. He also says something worth more than the praise: about where the application is hosted, they explained it to him and he cannot repeat it.
That is a common gap and it is homework for any vendor, us included. A client who cannot say where their data lives does not have a technical problem, they have a documentation problem. The four questions about the "after" are worth settling in writing before signing, while you still have alternatives, and not later, when it is a negotiation from a single position.
So what is the real risk?
The episode's closing line sums it up better than any analysis: most of these fears are not protecting you, they are slowing you down. The real risk is not choosing wrong, it is that your competitor decides while you are still hesitating.
It is worth marking the honest limit of all of the above. This is one case, told by one client, and one case is not statistical evidence. What transfers is not his results, it is the three questions he asked himself before signing: is this process part of what differentiates me? Do the questions they ask show that they understand? And what happens the day after delivery?
How we apply this at MasterDragon
We start with the "after" questions, not the "before" ones.
In practice that means the first conversation of a project covers where the application will live, who owns the data and how a change is requested, before we talk about screens. It is the least persuasive part of a proposal and the one that prevents the uncomfortable conversation six months later.
If you are in any of the three moments, watch the full episode and keep the questions, not the conclusions. You can start with how we build your software and review our portfolio. The argument for why this is an asset and not an expense is developed in custom software is an asset, not an expense, and why having the tool is not enough in AI adoption is not adaptation.
References
- MasterDragon.AI. (2026, August 10). The truth about hiring custom software. Escala sin Miedo, episode 1, with Marcel Verand. In Spanish. https://www.youtube.com/watch?v=Vw9fgxg4c98
- Gartner. (2025, June 25). Gartner predicts over 40% of agentic AI projects will be canceled by end of 2027. https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027