Merivant

Practitioner-led AI consultancy. We design and build AI systems for teams who need tools that work, not demos that impress.

Everything we ship starts from the same question: what is this person actually trying to do? Then we clear the path between that impulse and the outcome. The less you think about the tool, the better we did our job.

How We Think

Think about the best workshop you've ever walked into. Nobody greeted you at the door. The saw was where your hand reached. The clamps were near the workbench. You just started working — because everything was arranged for the way you actually move.

That's what we're building with AI. Not a smarter chatbot. A workshop organized around the shape of your thinking. We audit workflows, design the control layer, and build local-first agents that disappear into your existing tools.

A system organized around how you actually think, decide, and fail gets more valuable every day you use it.

Open loops that keep your unfinished thoughts alive across sessions. Memory that digests experience into instinct. Context that loads what's relevant, not everything. And always — a hard override that keeps authority where it belongs. Because the best tools disappear into the work.

We measure this with what we call the Instinct Cost Index — how much cognitive overhead a tool imposes between impulse and action. A claims team that opens four screens to triage a case has a high index. One that glances at a summary and acts has a low one. The best systems phase from scaffolding to trust to near-invisible.

Read The Instinct Tax →

The Book Series

Building for AI is a four-book practitioner series on AI governance architecture. We establish the standard, find the money, build the thing, and take care of the people.

What we'd build if we were doing it again.

Free PDF downloads, plus a practitioner toolkit with eight templates that bridge from the books to Monday morning.

Explore the series & download →

Research & Field Notes

Control & Governance

For CTOs and engineering leads running agents in production or regulated environments.

Simple Override Beats Elegant Architecture 2026 A single human override is more valuable than any sophisticated AI governance circuit. The Access Layer Pattern 2026 — technical whitepaper Why reversible typed-placeholder redaction beats traditional DLP for AI governance. The Core Instinct 2026 Local-first AI that runs on your machine, learns on your terms, and answers to you.

Memory & Learning

How AI systems get better from use — not just bigger from data.

Building an AI That Digests Its Own Experience 2026 — field note Most agents act and forget. We built one that prunes what didn't work, archives what did, and gets faster. Crystallization: When Questions Find Their Own Answers 2026 — practitioner observation Questions that sit in the stream and let answers precipitate when enough evidence sticks.

Interfaces & Adoption

For product leads and operators designing how people actually use AI tools every day.

The Instinct Tax 2026 — practitioner observation Every step between impulse and value is a tax on human behavior. Design from the instinct backward. We Organize AI Around People, Not Humanize AI for People 2026 Structural organization compounds. Build AI around human intent, not human comfort. I/O Channels: The Public API Is a Feeling 2026 — practitioner observation Three dots. Glance, drill down, done. Every surface is an I/O channel. Remodel, Gut Rehab, or New Construction 2026 — long-form Three ways enterprises build for AI. Three risk profiles, three honest conversations.

What We've Built

These are living systems we use and adapt inside client work. Each one started as a problem we solved for ourselves.

Core runcore.sh The local OS underpinning your internal agents. Memory, context, and intent — on your machine, under your control. Open source. Grey Matter Reports gmreports.com Your tech stack scored by five frontier LLMs against industry signals. See where your confident assumptions might be off. BragBin bragbin.com Text your wins. AI turns them into career evidence, competency maps, and the story you'd forget to tell at review time.

How We Work

01

Assess

Walk the jobsite. Map your workflows, data, and existing tools. Identify where AI creates value and where it creates friction.

02

Architect

Design the control stack: what stays, what changes, where the human override lives. Spec it before building it.

03

Build

Ship agents that run on your infrastructure, learn from your data, and disappear into the daily work. Measure the instinct cost. Iterate.

Request a working session