For most of its history, Teravision Technologies won business by being the company that told clients the truth. If a smaller shop promised a project in eight months for a tenth of the budget, Ricardo Arcia would tell a prospective client it would take twelve months and cost more, because the cheaper version would eventually cost them their business relationship with him instead. For years, that pitch worked. Then, in 2024, it stopped.
This is a story about what happens when a founder's entire competitive advantage seems to become irrelevant overnight, and about the difference between defending an old model and rebuilding a better one. Ricardo had spent twenty years teaching clients why quality software takes time. He was not prepared for competitors who could plausibly claim that AI had erased that tradeoff.
This story moves through four stages: Build, Break, Breakout, and Breakthrough. Each one shows up in how Ricardo actually ran Teravision through the year that AI stopped being a talking point and started being a threat to his pipeline.
By 2024, Teravision had spent more than two decades building a specific kind of reputation in the nearshore software development market. Early on, the company worked mostly with startups building minimum viable products, but Ricardo watched roughly eighty percent of those clients run out of capital regardless of how good the engineering was. Teravision shifted toward mid-size companies instead, typically businesses generating ten million to five hundred million dollars in revenue that needed serious engineering teams embedded in their operations.
The pitch that won those deals was consistency. When a client asked for an eight-month timeline at ten times a lowball competitor's budget, Ricardo would counter with twelve months and fifteen times the budget, and explain exactly why. "If you go to a smaller shop that is going to tell you yes and yes and yes and everything you want to hear before you start, you're going to get burned with that shop," he said. "But if you want to go with a serious company that is really going to deliver what you need, this is going to happen." That pitch kept winning. Teravision built its business on being the company that would not lie to a client about what serious software development actually costs.
In 2024, the deals started slipping away. "I typically kept winning those deals," Ricardo said. "But in 2024, I started losing them. It looks like the AI hype is starting." Clients who had always responded to his honesty about timelines and budgets were suddenly finding competitors willing to promise both AI-accelerated delivery and a fraction of the cost, and some of them were taking the bet.
Ricardo did not accept the premise on faith, and he did not dismiss it either. He built an internal lab and ran a direct test, taking a project that had taken seven months the previous year to see whether three separate teams could complete it in eight weeks using AI tools. The results were not what a simple AI success story would predict. Junior engineers moved fast but produced work Ricardo described as lower quality than what they used to produce without AI at all. Senior engineers were more careful and produced better work, but inconsistently. And at every stage of the process, from requirements writing to testing, some part of the pipeline accelerated while the rest of the team was not ready to receive the output, creating new bottlenecks that had never existed before.
That test forced a recognition Ricardo did not want to reach. "We felt that fear of, well, first it's obvious that it's obsolete what we're doing," he said. "We have to switch the way we've been doing software. It's not the way that software is going to be built in the future for sure." The company that had built its identity on knowing exactly how serious software gets built was looking at proof that it no longer fully understood its own process.
The reflection that followed was not about whether AI worked. The lab had already shown that it did, unevenly and inconsistently, but it worked. The real question was how Teravision's engineers needed to think differently in order to use it well, and Ricardo's answer became what he now calls a cognitive engineering framework, a system for deciding where AI belongs in the software development process and where it does not.
Building the framework meant working within a real constraint. Teravision's engineers already had live client obligations while Ricardo was asking them to unlearn habits that had made them good at their jobs for years. He addressed that directly rather than pretending the transition would be painless. "I need you to jump into this new wagon and do it the way I'm telling you," he told his team. "And I want you to make mistakes. I know that it's going to take more time. I know that at the beginning it's going to be harder, and I agree that it's going to be slower. That's fine. But I need you to do it."
The decision that made the framework possible was how Ricardo reframed the threat itself. Rather than telling his engineers that AI would replace parts of their job, he told them plainly what he believed was actually true. "AI is not going to replace your work," he said. "What is going to replace your work is another engineer that choose to harness AI with us in the proper way." That reframe turned adoption from a threat into a condition of staying on the team, without making it about AI outperforming people.
From there, Teravision built what Ricardo calls a zero team internally, tasked with answering one question, how the company would build software from now on. The answer became a formal cognitive engineering framework paired with a certification and training program, built in levels. The first level was a change of mindset, teaching engineers to think of themselves as strategists orchestrating a tool that works faster than they do, rather than executors doing the work by hand. Later levels covered specific tools and then role-specific applications, so a project manager, a developer, and a quality assurance engineer each learned the framework differently, matched to their actual job.
The clearest external result came from Teravision's longest-running client implementation, where the company measured a thirty to thirty-five percent increase in software development productivity within roughly a year. In parts of the business built around new, unencumbered projects, the pricing shift was more dramatic still. Work that would have been quoted at a hundred and fifty thousand dollars five years ago now runs closer to ten thousand dollars in that segment, a fifteen-fold change Ricardo attributes directly to the framework.
The internal shift went further than the numbers. Ricardo spent the following months interviewing more than a hundred and fifty chief technology officers, discovering that nearly every one of them was asking some version of the same question about how to use AI in their own software development process. That research became the basis for a book, and it reshaped how Ricardo sees his own role. He now describes the goal as becoming what he calls a cognitive leader, someone who deliberately connects what AI can do with what a team can do, rather than assuming either one will simply take care of the other. "You have to do it intentionally," he said.
Nothing about Ricardo's story suggests that AI made his engineers less necessary. If anything, the opposite happened. The engineers who lost ground were the ones who tried to keep working exactly as before, treating AI as a tool to bolt onto an unchanged process. The ones who kept their place on the team were the ones willing to be slower for a while, willing to make mistakes in public, and willing to let Ricardo tell them the truth about what needed to change.
That is the same instinct that built Teravision's reputation in the first place. Ricardo won business for twenty years by telling clients the truth about what quality software actually costs, even when a competitor offered a cheaper story. When AI made that same truth-telling instinct feel obsolete for a moment in 2024, the response was not to protect the old pitch. It was to go find out, faster than his competitors did, what had actually changed and what had not.
