Skip to content
es
Menu

Sebastián Delsalto

I build and run companies around AI.

I'm Sebastián Delsalto. From Quito, I connect strategy, software and operations through the Daleki Global ecosystem.

Let's work togetherSee the companies
Black and white studio portrait of Sebastián Delsalto against a black background.
From Quito, Ecuador.Building without borders.

Founder · Builder · AI-Native Operator

Explore my work

The evidence

The thesis is operated, not explained.

These companies exist, have public sites and operate in industries that look nothing like each other. They are the proof that this is not a theoretical framework.

Let's work together

Three ways to start a conversation.

  • Projects

    Designing and putting your company's operating system to work. Delivery happens through Daleki Lab.

    Explore a project

  • Partnerships

    Building an opportunity with operators, partners and companies that share a long-term vision.

    Propose a partnership

  • Speaking and media

    Conversations about AI-native companies, entrepreneurship and the decisions behind the operation.

    Invite me to a conversation

The thesis

Don't add AI to companies. Build companies AI-native.

Almost everyone is doing the first thing: an existing company, with its usual processes, plus some tools on top. The result is a layer. It works a little, it shows a little, and it changes neither the cost structure nor the speed of decision.

An AI-native company is not a company with agents. It is a company where artificial intelligence, software, data, workflows, agents, automation, deterministic systems and people were designed together, as a single operating system.

The difference shows up in where the decision sits. In the first, AI assists a human process. In the second, the process was born assuming it, and whatever stays in human hands is there because someone decided it should be, not because it was left over.

The full definition

How I build

Workflows, Agents, Tools.

This is the framework I use to organise the operation of an AI-native company. Three layers, and nothing else. The temptation is always to add a fourth one that is really just documentation.

  • Workflows

    Reproducible processes that turn an input into an output. They have a trigger, steps, validation, and defined behaviour for when they fail. If a process cannot be described that way, it is not ready to be automated.

  • Agents

    AI systems that inspect, reason, propose or execute within explicit permissions. Each has a scope, allowed tools, prohibited actions and an escalation rule. An agent without written limits is not autonomy, it is exposure.

  • Tools

    The services, APIs, scripts and libraries workflows and agents use. Each documents its interface, its auth model, and what happens when it fails.

Human gate

What I don't automate.

An AI-native company is not a company without people. The hard part is not deciding what to automate. It is deciding what not to, and holding that line when automating would be cheaper.

In the companies I run, these things pass through a person before they ship:

  • Anything published under the brand's name.
  • Any claim about results, figures or performance.
  • Pricing, terms and commitments to a client.
  • Decisions that affect someone else's money.
  • Hiring, letting go, and anything that touches a team.

This is not decorative caution. A system that can publish without supervision can be wrong at the speed it produces, and that speed is precisely why it was built.

Build log

What I'm building.

Each entry is a real build with its problem, its decision, and what was deliberately left in human hands. No figures that cannot be verified.

Let's talk on WhatsApp+593997854652