MAD HARE

AI consultancy · model-agnostic

We ship AI systems.

Most consultancies sell advice. Mad Hare designs and ships agent systems on any frontier model — Claude, GPT, open weights, or your own — deployed on your infrastructure or ours.

Mad Hare

The problem

Enterprise documentation is scattered, drifted, and detached from the work it describes.

We treat that as the primary problem, not a byproduct. Conflux is built to make a company's corpus coherent enough to operate on.

01

Fragmented

Docs live in five tools. No two teams use the same source. The 'truth' is whoever spoke loudest in the last meeting.

02

Drifted

What's written and what's deployed diverged six months ago. Nobody knows which doc is still right.

03

Detached from work

The docs aren't where the work happens. Engineering ships in one system, knowledge lives in another. The two don't talk.

The difference

Most AI shops sell advice or tools. We sell delivery you can audit.

Two practices combine into one outcome. Consultancy gives us the discipline. System design gives us leverage. The combination is what makes the work auditable from day one — and maintainable past it.

PRACTICE 01

Consultancy delivery

Engagement discipline, scoped milestones, written deliverables. Engineering rigor applied to AI work.

PRACTICE 02

Model-agnostic system design

Agent architectures that work on Claude, GPT, open weights, or custom — without rewriting the system.

OUTCOME

AI that ships, audits, and lasts

Systems your team can defend in front of regulators, debug six months later, and migrate to a new model when the right one comes out.

The platform

We run our delivery on Conflux — our own model-agnostic agent platform.

Conflux is the system we use to ship our own work. Toggle below to see how a real engagement transforms — from scattered, direct coordination to coordinated agents with full traceability.

Flow diagram
Interactive composer used for scoping and capability mapping.
BusinessAgentSystemDataInactive
Details
Business User (Before)

Business teams and individual users must find information, documentation, and data across multiple systems.

This results in things we all recognize:

  • Outdated documentation
  • Incorrect or inconsistent permissions
  • Information that is unsearchable
  • Content that is never maintained

The user is forced to manually stitch together context just to do their job.

01

We built it, we use it.

Conflux runs our delivery work — documentation, audits, project management, code review. Every engagement sharpens it. Yours does too.

02

Model-agnostic by design.

Built on Claude, GPT, open weights, or your own — without rewriting the system. The right model is whichever one fits your constraints today, and tomorrow.

03

You benefit, you don't license.

For most engagements you don't license anything — you just benefit from the platform behind the work. When it does fit, Conflux is available standalone.

How we work

Four phases. Written deliverables at every one.

01

Listen

Interviews, system audit, written intake brief. We come in to understand before we propose anything.

Walk away withINTAKE BRIEF
02

Map

We design the system: agents, data flows, contracts, costed plan.

Walk away withSYSTEM DESIGN
03

Build

Design through ship. Every milestone is a working artifact, not a slide.

Walk away withWORKING SYSTEM
04

Hand off

We're not done until your team can run it without us. Most shops leave; we stay until handoff is real.

Walk away withOperational ownership. The team that runs it day-to-day is yours, not ours.

Services

Four engagement shapes, scoped to your actual need.

From assessment to full implementation — we scope to the work, not to a template.

Assessment & Strategy

2–4 weeks · fixed fee

Small, scoped engagement. We come in, audit your AI landscape, and tell you honestly what's worth building and what isn't.

Best when
  • Exploring AI direction
  • Pre-budget commit
  • Need an honest read
MOST ENGAGEMENTS

Build

8–16 weeks · scoped engagement

Full design-through-handoff for a single use case on existing infrastructure.

Best when
  • Clear use case
  • Existing infra to build on
  • You want it shipped, not advised
FULL IMPLEMENTATION

Foundation

20–30 weeks · scoped milestones

Full platform foundation plus the first use case running on top. Infrastructure, governance, orchestration, and a working system.

Best when
  • Multiple use cases planned
  • Building for the long run
  • You want to own the result, not the vendor

Audit & Rescue

4–8 weeks · scoped

A previous vendor left it in pieces. Costs are spiraling, results aren't. We come in to fix what's there — without a rip-and-replace.

Best when
  • Previous vendor stalled
  • Costs spiraling
  • Need recovery, not restart

Not sure which fits? Start a conversation — we'll route you to the right shape, or tell you honestly if we're not the right partner.

Start a project

Tell us what you're trying to ship.

No form, no funnel. Send a few sentences about what you're working on and we'll route from there — to a conversation, or honestly to a partner who fits better.

Email: 1 business day · Telegram: real-time when online

Maine, USA · remote-first