Services
I turn manual workflows into AI-native systems of action — for B2B SaaS and enterprise teams running operations, finance, logistics and support. That's one job, delivered in three layers, and they ship in this order.
I think in systems before I think in screens. The design system is the language that makes that thinking reusable — the same states, permissions, and decision points a workflow needs, expressed as tokens and components engineering builds from directly. From there, workflows get automated to the level a company is actually ready for: deterministic automation where the outcome is fixed (L1), agent-assisted steps with a human approving the result (L2), or adaptive agent loops only where the audit trail and permission scopes make removing that review safe (L3).
How I work: embedded with the product team as the designer on the platform. I was the sole designer across both client engagements through Agentic Dream. On Avanto/Strata, I presented new work daily to the CEO, project managers and engineers, defending each decision in front of the people who could veto it. On Repfabric, I worked near-daily with founder John Mitchell, who reviewed and approved every design decision.
Layer 1 · Define, redesign, and structure critical workflows for B2B SaaS tools
I help teams working under pressure define and restructure execution workflows, especially in operations, finance, logistics, and support.
• B2B SaaS platforms and internal tools.
• Workflows across operations, finance, logistics, and support.
• Dashboards, information architecture, exception handling, and escalation paths.
• Audits to identify friction, failures, manual work, and misaligned workflows.
• Connecting current workflow problems to a realistic product roadmap.
✓ End-to-end workflows teams can actually follow.
✓ Fewer errors and less coordination overhead.
✓ Faster execution and better visibility into next actions.
✓ Clearer direction, stronger alignment, and a realistic path.
✓ A roadmap that balances quick wins with medium-term evolution.
On Strata, restructuring execution into state-based workflows delivered +40% SLA compliance, +40% throughput and -30% operational delays — measured from the platform's own reporting, before and after the rollout.
Product de-scoping — 3 weeks, fixed price:Check the price on Contra
Workflow Clarity Map — 1 week, fixed price:Check the price on Contra
Layer 2 · Define and design AI-native human+agent systems of action (from discovery to delivery)
I help teams define human+agent workflows that support judgment, add guardrails, and keep critical decisions under human control—from early discovery to implementation.
• Human-in-the-loop systems and AI decision interfaces.
• AI-driven patterns for routing, summarizing, and proposing actions.
• Risk-aware workflows where mistakes are costly (finance, compliance, ops).
• Governance rules for when AI should act, suggest, or wait for approval.
• Reusable rules and learnings, not just prompts.
• Translating AI opportunities into roadmap items and phased delivery.
• Workflow routing and escalation in n8n.
• RAG-based workflows that deliver timely, relevant information into the right step.
• API orchestration and webhook integrations across tools.
• Managing tools like Perplexity as digital coworkers for research and automation.
• Computer-use automation for the repetitive work that never fit an API — form filling, local files, desktop steps between systems — running on the user's own machine and with their permission.
• Designing the agent's own behavior — retrieval routing, guardrails, persona, voice — the way this site's assistant is built, with no contact form behind it.
✓ AI that improves decisions and execution instead of creating noise.
✓ Human accountability over automation.
✓ Product teams that treat AI as part of the system, not as a feature.
✓ A maturity framework to sequence AI adoption safely.
✓ Durable human+agent collaboration patterns that evolve over time.
The question a company actually has here is what happens when the model is wrong. On Strata that answer shipped: L2 in production where the decision surface was narrow enough to trust, and L3 designed and deliberately held back until audit trails and permission scopes could remove human review safely.
AI workflow audit — 2 weeks, fixed price:Check the price on Contra
AI agent boundary spec — 4 weeks, fixed price:Check the price on Contra
Layer 3 · Design systems — the contract that makes layers 1 and 2 shippable
I view design systems as a visual API between design and engineering. Semantic tokens map one to one from Figma to code, so engineering consumes that API directly in its own stack instead of re-deriving it from a static spec. I hand the system over as a Storybook the frontend team pulls into its own workflow — Figma is the collaborative space for exploration and client iteration, and Storybook is the versioned source of truth for the implemented components, tokens, and their documented contract. They help teams ship faster, stay consistent, and improve products through safe, incremental changes.
• Figma systems built with primitives, tokens, and component variables.
• Tokens and components mapped to implementation primitives (Tailwind, ShadCN, Radix, etc.).
• API-first UI patterns for configurable enterprise products.
• Clear documentation for appearance and behavior.
• Prioritizing improvements using the design system.
• Systems as Figma libraries and v0/Cursor structures.
• Example flows for faster code integration.
• Storybook as the versioned reference frontend actually builds from, once a component is resolved from Figma exploration.
• Design systems that stay aligned with engineering constraints and roadmap changes.
• Scoping legacy CRM/ERP and cloud-based platforms down to the workflows, roles, and underlying data model that actually drive adoption.
• When engagement scope requires it, delivery can also include a documented npm package/API so the frontend team and approved AI-enabled tools consume the same system directly, not just Storybook.
✓ Faster development and less inconsistency across teams.
✓ Shared mental models across product, design, and engineering.
✓ A design system that evolves with the product.
✓ Clearer handoffs and fewer gaps between design intent and shipped code.
On Strata, the design system existed so the action system could ship incrementally — that's where the +35% delivery velocity came from. A component library on its own doesn't cover the states an AI-assisted workflow needs — when a machine is about to act, whether it's still thinking, where a result came from. The system has to define those states, along with permissions, provenance, and behavior, so humans and agents can operate on the same contract, not just designers and engineers. Live walkthrough: sebastianbalderas.com/storybook.
Design system as code — 3 weeks, fixed price:Check the price on Contra