Spectral
AI and automation agency. Agents, workflows, and the sites, apps, and internal systems around them, from concept to production.
facts
- role
- Co-founder · Engineering
- year
- 2025–present
- discipline
- AI · Automation · Design · Development
- stack
- n8n, OpenAI, Claude, OpenRouter, MCP, Next.js, Supabase, Vercel
brief
01 problem
Most of the hours in a business go on moving information between tools: a form to a spreadsheet, an email to a CRM, an invoice to the books. Teams know AI could take that work off them, but between vendor demos and pilots that never ship, very little of it reaches production.
02 constraints
Clients keep the systems they already run, and everything is deployed in their own accounts. Automations have to be reliable enough to trust without watching them, cheap enough to run every day, and understandable by the people who inherit them. A small studio, so every build has to reuse what the last one proved.
03 solution
Spectral maps where the hours actually go, then replaces those paths with workflows that run on their own. AI agents handle the fuzzy steps: reading, classifying, drafting, deciding. Where a workflow needs an interface, it gets an internal tool or app on the client's own data. The build plugs into the existing stack and ships with documentation, so the automation belongs to the business, not the vendor.
highlights
- 01AI agents and business automation
- 02End-to-end delivery from concept to prototype to production: sites, apps, dashboards, and internal systems
- 03Deployed in the client's own accounts, built on n8n, LLMs, MCP, APIs, Supabase, and Vercel
- 04Brand identity, creative direction, and product and UI design
- 05Local and embedded AI, voice interfaces, and computer vision and browser-control experiments
- 06Developer tooling, LLM tooling, and generative design
- 07Client consulting, technical strategy, and fractional CTO work
notes
What we build
AI agents for the steps that need judgment. Workflow automation across the tools a business already runs. Sites, apps, dashboards, and internal systems where a workflow needs a screen, with the brand and interface design to match. Integrations that stop people copying data from one system to another.
How it works
Start with where the time goes, not with a model. Map the manual paths, pick the ones worth automating, and take each one from concept to prototype to production, measured. AI goes in where the decision is fuzzy enough to need it, and nowhere else.
My part
I co-founded Spectral and lead engineering: the agents, automations, and integrations, the apps and infrastructure they run on, and the technical side of client work, from strategy to fractional CTO roles. If it moves data between systems or makes a decision on its own, it comes through me. My co-founder leads design; I bring 20 years as a designer to the product and interface work.
Own products
Alongside client work, Spectral builds its own products: BotFriendly, a directory of bot-friendly services for AI agents, and AgentFinance, crypto wallets for AI agents. More are in development.
results
- Automations in production, not pilots
- Repetitive work removed from teams across operations, sales, and support
- Agents and workflows running inside the tools clients already use
- Clients own the full system: workflows, code, and documentation