We help vendors of specialist and public administration software bring AI features into their existing products: from model selection and architecture to prototype, without building an in-house AI specialist team and designed for sovereign operation.
You bring the product, the customers and the public-sector expertise. We bring the AI architecture.
We design an architecture that fits your customer's hosting and data requirements.
We bring the AI architecture and work with your developers instead of building a parallel team.
We find the gaps between prototype and operation and close them with you.
We work out options, assumptions and risks that hold up in your proposal and in customer conversations.
You start with a validated architecture instead of months of experiments. Your developers keep building the product, not AI infrastructure.
You can answer your customers' AI requirements on solid technical ground, even when hosting and data requirements are strict.
You make model, platform and hosting decisions based on a comparison, not on a vendor pitch.
The first AI feature creates building blocks and knowledge in your team that further features build on, instead of a custom solution per customer.
The intro call is free. After that, you receive a fixed-price offer for the check or the sprint.
We clarify whether your AI project is technically feasible under the given requirements, which architecture options are viable and what the next step should be.
Based on the agreed documents and conversations. The check does not replace a full security or compliance review.
For an AI feature that is to go into your product. Designed so that further AI features can build on the same foundation.
Covers one prioritised use case and one agreed integration environment. Further systems, security testing and formal compliance reviews are not part of the standard scope.
We develop with your team and support architecture reviews and technical decisions. The goal is that your team can then evolve the foundation on its own.
An AI search or assistant only sees what the requesting person is allowed to see in the specialist software. The permissions of the source system apply to every query.
The AI creates drafts and suggestions. Binding decisions and critical actions stay with people, with defined permissions, approvals and logging.
Answer quality is checked with traceable tests, including when the model or hosting changes. This keeps model selection a decision you can change for good reasons.
Zapote Ventures is led by Oscar Morillo Victoria. He has worked in software engineering and AI for more than twelve years and has taken AI products from idea to operation. More about Zapote
We are not public administration experts, and we don't pretend to be. That is why we work with vendors who know the public sector.
Your developers know your product best and should keep working on it. We bring experience with AI systems and take responsibility for the architecture decisions, not just for individual tasks. Your team works alongside us and builds knowledge along the way.
We are technology-independent and receive no commissions from model, cloud or platform providers. The choice depends on your customer's requirements, from open models on your own infrastructure to commercial services in the EU.
We do not provide operations, legal advice or certification. We flag early in the design, from a technical perspective, whether a system could be considered high-risk under the AI Act; you agree the legal classification with qualified experts.
Whether it's a customer request, a new product feature or a stalled pilot: in 30 minutes we assess your requirements and clarify whether working together makes sense.
Information on how your data is processed when you book an appointment or write to us is in the privacy policy.