I’m Maor David-Pur. I’ve built my career moving between technology, customers and business - from software and product to presales, data, AI and technology leadership. I like turning a real problem into something that can be understood, built, adopted and scaled.
MY VIEW ON AIThe AI demo is becoming a commodity. Everything around it isn’t.
LinkedIn ↗
I started in software and kept moving toward the places where technology has to meet reality: product decisions, customer conversations, presales, data platforms, AI and eventually the challenge of building teams and organizational capabilities around all of it.
Software, architecture, platforms, data and AI - not as isolated domains, but as parts of a system that has to work.
Discovery, presales and solution shaping taught me to ask what someone is actually trying to achieve before deciding what to build.
As my roles became more senior, the question shifted from “what should we build?” to “how do we build the organization that can keep doing this well?”
I care less about launching another project than about leaving behind something an organization can reuse, operate and improve.
I lead a technology organization of approximately 50 people, connecting multidisciplinary capabilities around shared outcomes.
Turning an interesting idea into a capability with real data, architecture, ownership, security, measurable value and a path to adoption.
Moving between customer needs, executive priorities and technical trade-offs without losing the substance in translation.
Models are becoming easier to replace. Organizational context is not. That shift changes what we should build, what we should govern and how we should think about the cost of intelligence.
The winner of this month’s benchmark is not a durable architectural dependency. Models will improve, prices will fall, capabilities will converge and the best choice will vary by task.
A replaceable model layer. Keep business logic, organizational context, evaluation and security outside the model. The durable asset is the system around it.
A catalog used to help people find data. Agents need more: meaning, ownership, quality, permissions, lineage and a reliable way to discover what organizational data they can use.
Metadata stops being housekeeping and becomes machine context. The catalog may become part of how an organization tells AI what its data means - and what it is allowed to do with it.
In AI systems, model choice, context, routing, frequency and agent loops create cost while the system runs. More intelligence is not automatically more value.
How much intelligence is this decision worth?
The goal is not to minimize tokens. It is to spend intelligence where it changes the outcome - and use smaller models, less context or deterministic software where it does not.
If models become increasingly interchangeable, competitive advantage moves outward: into the context an organization owns, the architecture that can use it safely, and the ability to apply the right amount of intelligence to each decision.
AI Weekly is where I interpret what is changing now. These are the longer arguments I’m developing from the ideas above.
A weekly briefing focused on the developments that may actually change architecture, economics, products and management decisions.
Read AI Weekly ↗I’m based in Herzliya, Israel. Outside work, I’m usually experimenting with some piece of technology, doing a DIY project at home, reading about history, or trying to make something work a little better than it did before.
I’m always interested in a thoughtful conversation with people building ambitious things.