“AI Is Like Magic — but It Also Creates a Lot of Organizational Chaos”
29 January, 2026
Shay Friedman, CodeValue's CTO, on the organizational pressure to adopt AI, the data bottleneck, and the need for new development practices to turn models and agents into stable production systems
For more than a decade, software development services company CodeValue has been accompanying organizations and technology firms at moments when transformative technologies reshaped the rules of the game. It began with large-scale cloud migration projects, continued with enterprise-grade application development, modern architectures, and advanced development methodologies. Today, as AI transitions from an experimental layer to a core business tool, the company once again finds itself at the forefront — this time facing an even deeper shift.
CodeValue employs more than 400 developers, architects, and technology professionals, working with a wide range of clients: large enterprises, startups, banks, insurance companies, and defense organizations. Its activity spans development services, consulting, and training — a combination that takes on new significance in an era where organizations are trying to understand not only how to use AI, but how to build stable, sustainable development and operational processes around it.
According to Shay Friedman, recently appointed CTO of CodeValue after co-founder Amir Zucker transitioned into the role of chief architect, there is a growing gap between the pace at which organizations are adopting AI tools and the level of process maturity required to run them at scale. “Companies want to build agents, deploy models, train internal systems — but they still don’t really know what this is supposed to look like in production,” he says.
Building expertise in data curation
Over the past year, CodeValue has been developing new capabilities around supporting organizations in training internal models, setting up enterprise AI infrastructures, and upskilling teams. At the same time, the company is establishing a dedicated data and data-curation practice, aiming to define new development methodologies suited to a non-deterministic world.
“CodeValue is not a product company,” Friedman emphasizes. “We provide development services, consulting, and training. Our product is our people. As CTO, my goal is to make sure everyone is constantly advancing technologically. If our people aren’t at the frontier, we have nothing to sell.”
The AI era sharpens that challenge. “AI changes how work gets done inside organizations. Suddenly you don’t need hundreds of employees to perform certain tasks — one person with the right tools can do the work of several. That creates fear, but it also reinforces one thing: strong, experienced people will always be needed. Our advantage is that we live the technology while it’s still taking shape.”
Pressure, constraints, and chaos in the enterprise sector
From his experience in recent months, Friedman describes an enterprise sector under constant pressure. “Even the largest organizations understand that AI is something they must adopt. Those who don’t will fall behind. But the path forward is anything but uniform.”
In traditional organizations such as banks, insurance companies, and defense bodies, the barriers are sometimes very concrete. “There’s no external internet access, you have to build internal infrastructure, and everything slows down. Even when they try to deploy an internal chatbot, it’s trained only on organizational data. The dataset may be large, but it can’t compare to the open web — which leads to more hallucinations and less stable models.”
Even in connected organizations, the situation is far from orderly. “It’s often pure chaos. When it’s hard to define a single strategy, every division and every team does whatever seems right to them — like mushrooms after the rain. The problem is that AI tools pull code from the existing codebase. If that code is old, bloated, and poorly maintained, low quality simply propagates. That’s a real risk, and I think we’ll be dealing with the consequences in six months to a year.”
Data, training, and the leap to production
This is where data becomes critical. “We’re now building a data-curation practice internally. It’s still early, but it will grow, because data is the bottleneck — especially when companies train their own models. In that case, data is everything.”
Training is becoming equally essential. “A lot of people work with tools like Claude Code and don’t even tap into half of their potential. Good training turns every employee into a power user, and that can transform an organization from the inside.”
One of the most complex challenges, Friedman says, is the transition from experimentation to production. “AI makes it incredibly easy to generate things. One prompt, and it feels like magic. But AI isn’t deterministic — it’s probabilistic. You really understand that in QA, when a bug appears and no one knows where it came from. This is a new reality.”
As a result, CodeValue is developing training programs and best practices around safety, performance, and cost management. “Some companies rush into AI and then discover that usage costs aren’t sustainable. Everyone is running into this right now. We’re still in the infancy stage.”
Being a CTO when the product is people
For Friedman, the role of CTO in a development services company is fundamentally different from that role in a product company. “My goal is for every employee to have a clear professional horizon — to be fluent in the most advanced tools and technologies.”
That model, he says, also explains the company’s organizational stability. “Anyone can consult with anyone else. People work on a wide variety of projects — it’s like moving between companies without actually leaving. Burnout is lower.”
Looking ahead, Friedman concludes, uncertainty itself is the engine. “It’s exciting to be at the frontier, facing a challenge and then figuring out a solution. These are challenges that didn’t exist a year or two ago — and that’s exactly why we’re here. We’re living in a historic moment.”
Posted in: AI
Posted in tags: CodeValue , Shay Friedman
