Agents that do
the actual work.
I design, harden and run AI agents that take real work off your plate - built on the same pipelines I run my own business with. One operator, a team of agents, agency-grade output.
I don't sell hours.
I sell the machine.
Most shops bill you for people. I run my own company on agents - research, content, ops - so what I sell you is the thing I depend on daily, already battle-tested. And the win isn't hours saved on a timesheet. It's leverage: capacity to take on more, think harder, and stop drowning in the work around the work - without adding headcount.
Agents I built that earn their keep.
EVP Research Engine
Analyses internal video interviews and employee research through a multi-step pipeline, surfacing concrete executive findings - automatically.
Content Pipeline
Takes a brief, produces images, video and posts, and ships to socials autonomously - with human approval gates and the sense to pivot.
Training Brief Generator
Turns compliance-driven needs into full briefs - prompts, VOs, visual direction - and fact-checks itself against real research.
From an idea to a fully managed agent - and everything between.
Three ways in. Every engagement is one of these - the specific builds underneath each are scoped to your process.
Audit
EntryA fixed-scope teardown of an existing agent or process - or a one-week build sprint. Low-risk way to start.
Build
CoreDone-for-you agent or pipeline, built on your infrastructure or mine - integrated, documented, handed over or co-piloted.
Managed
RecurringFully managed agents that live inside your business - kept healthy, tuned and improving while you just see outcomes.
Indicative ranges to set a ballpark - every job is scoped to your process on a call.
Scope. Build. Harden.
Scope
Map the process, define the outcome, and decide exactly what the agent should - and shouldn't - do.
Build
The full pipeline: prompts, tools, memory, model orchestration and human approval gates where they matter.
Deploy & Harden
Live on your infrastructure, then tuned against real usage until it's reliable enough to trust unattended.
Every agent ships with guardrails and human approval gates - by default, not as an afterthought. Agents take the grind, not the judgement.
exec
verify
Every build coordinates multiple AI models - one plans, one executes, one verifies. The output beats any single model, every time.
You choose how much you own.
I build on your infrastructure, integrate it, document it, and hand you the keys. Your team runs it.
I build and run it alongside you, hardening and tuning as real usage exposes the edges. Shared ops.
From nothing to a working, monitored agent that lives inside your business. You see outcomes; I keep it healthy.
A small peek at who you're working with.
I'm Ed. My background is creative - photography, filmmaking, editing - built up over about 16 years. When AI arrived I started folding it into that work, and it's become the thing I do most.
These days I sit somewhere between creative production and development: building the agents and pipelines, then wiring them into the tools and infrastructure a business already runs on. Most of what each project needs I learn as I go - not from a textbook.
I build these for my own business first, then for clients. If an agent can genuinely take something off your plate, I'll tell you - and if it can't, I'll tell you that too. Still learning, always evolving.
The thinking behind the builds.
Not the headline - but if you want to check the work runs deep, it's all here. Open notebooks on agent reliability, hardening and the patterns I keep reusing.
Field notes on making agents reliable
Write-ups on agent hardening, failure modes, guardrails and the orchestration patterns behind production-grade work.
Read the research →A library of what agents can actually do
Concrete, vetted agent capabilities - the building blocks I draw on when scoping a new build for a client.
Browse the skills →