Fractional AI Lead

The agent deployer your team needs before you can hire one. We embed with teams who are bringing agents into their production workflows — for climate-tech scale-ups and the institutions building alongside them.

What we do Pricing Get in touch
3-month minimum · 2+ days/week embedded · From €8K/month · Onsite or remote
The Problem

From AI tools to AI systems.

Most teams have learned to prompt. The next bar is operational. Production agent systems need somebody whose job is to map the workflow, design the data flows, engineer the context, and run the evals after every model change. The harder gap is the willingness to change the workflow itself — teams will adopt new tools, but they will not, on their own, change how the work gets done. Someone has to drive that, and most organisations do not have that person yet.

McKinsey · State of AI 2025

23% / 39%

of enterprises are scaling agentic AI systems / experimenting with them. Most are mid-transition from chatbots to multi-agent systems and have no internal owner for the architecture.

mckinsey.com

IBM · 2025

26%

of enterprises now employ a Chief AI Officer, up from 11% two years ago. Two-thirds plan to appoint one within two years. Full-time hiring is six to twelve months out, and the candidates who can actually do the job are not on the market yet.

IBM IBV, via Slayton Search

BCG · The Widening AI Value Gap

22%

of companies have moved beyond proof-of-concept. 60% generate no material value. The difference is curation — picking the right architectures, the right vendors, the right trade-offs. We make those calls every day, on systems we run.

bcg.com
Scope

Six things an embedded
agent operator does.

We come in, find the work where AI changes the answer by orders of magnitude, and ship the systems that get you there. Half of the job is operational. Half is having the courage to question the workflow itself.

01

Workflow leverage analysis

We sit with each team and find the work where throwing 100× compute at a task changes the answer entirely. Lead enrichment, contract review, client onboarding, knowledge retrieval, support escalation. Not every workflow is an agent workflow. We tell you which ones are — and which ones should not exist in the first place.

02

Question the job itself

Sometimes the job is to change the job. We will not just automate what you already do. We will challenge whether you should be doing it at all, and we will force the conversation if nobody else on your team will.

03

Data flow architecture

Structured and unstructured data, mapped and connected. Where it lives, where it moves, what an agent needs to act on it without making a mess. The unglamorous work that makes everything else possible.

04

Context engineering

The hard problem of agent work. Designing what the agent reads, what tools it has, what scope it operates in, what prior decisions it inherits. Most agent failures are context failures.

05

Evals, reviews, and the willingness to rip things out

Every model change, every data change, every prompt change needs a review pass. We build the eval harness, run the cadence, and pull the plug on systems that are not working — even ones we built.

06

Production operations & KPIs

Running agents in production. Tracking cost per task, error rate, escalation rate, time to recovery. Iterating on them. Reporting them upward in language the organisation actually uses.

Engagement

Three months. Embedded. Re-chartered.

Fractional AI Lead is a recurring retainer engagement, not a workshop and not a project. We embed with your team for a minimum of three months and shape the engagement around your context. The structure below is the default starting point.

Term

3-month minimum

Long enough to set a roadmap, ship against it, and decide together whether to extend. Renewable on either side.

Cadence

2+ days/week embedded

Weekly standing meeting with the executive sponsor, embedded time with the working team, and async availability between.

Delivery

Onsite or remote

Decided per engagement. Onsite weeks are common at the start; ongoing operations run remote. We are based in France and travel for the work.

Working language: English or French. Onboarding begins within two weeks of contract signature.

Past Work

We build what we lead.

A short selection of agent systems we have shipped publicly — built and run ourselves, not advisory deliverables.

Built & shipped

Anima — Synthesis hackathon, 2nd place

Autonomous AI characters that earn their own income, manage their own treasuries, and create digital art from their experiences. The most concrete demonstration we have built of an agent that lives, transacts, and operates on its own — the same operational discipline a Fractional AI Lead engagement applies inside your team.

Visit anima.cards

Interspecies Parliament — PL_Genesis 1st place, two tracks

An interactive “parliament” of AI agents that argue on behalf of ecosystems — wetlands, forests, marine zones — over how environmental projects should be funded. A multi-agent system with real coordination, eval cycles, and human-in-the-loop interfaces.

Read the blog post · Watch the demo

Windfall

Our production AI inference platform. Routes requests across three locations to balance speed, cost, and carbon footprint. We run it ourselves, and we can deploy it into your stack as part of an engagement.

Visit windfall.ecofrontiers.xyz

Written & published

The Green Crypto Handbook — Taylor & Francis, 2026

The academic reference for climate and crypto institutional design. A procurement-grade credential for regulated and public-sector buyers.

AI Institutional Design Atlas

An interactive map of how AI systems coordinate inside organisations and where they break. We use it to spot which deployments are going to fail before they fail, and to choose architectures that survive contact with production.

Alyra — blockchain & AI school

Course design and external examination for one of Europe's most respected blockchain and AI programmes.

Credentials

The principals behind Ecofrontiers.

The two of us share leadership and delivery, and your scoping call will explain how the engagement is shaped around your context.

Patrick Rawson

Patrick Rawson

Builds and runs Windfall, a production AI inference router. Has shipped six onchain agent systems in 2026. Ex-CEO of Curve Labs (Berlin), a Web3 venture studio.

Louise Borreani

Louise Borreani

Co-author of The Green Crypto Handbook (Taylor & Francis, 2026). Course designer at Alyra. MA Environmental Policy, Sciences Po Paris.

Pricing

Retainer-based.
One number. Three months minimum.

Less than the loaded cost of a senior MLE hire. Faster to deploy. Accountable for the system in production. The price scales with the size of the organisation and the depth of the embed — but the floor is fixed.

Fractional AI Lead

From €8K / month

3-month minimum · 2+ days/week embedded · Onsite or remote

What you get · 01

Workflow & systems design

· Workflow leverage analysis

· Data flow architecture (structured + unstructured)

· Context engineering for the agent layer

· Build-or-buy and vendor selection

What you get · 02

Embedded operations

· Weekly standing meeting + async availability

· Eval harness, review cadence, model-change reviews

· Production KPI tracking and iteration

What you get · 03

Operational handover

· Documented playbooks, eval harnesses, and architecture decisions you keep on day one

The starting floor is €8K/month. Final quote depends on organisation size, regulatory exposure, onsite vs. remote balance, and whether the engagement includes hands-on infrastructure deployment alongside the operator work. Renewal is quarterly.

Next Step

Book a scoping call.

Scoping calls run thirty minutes and decide fit on both sides. If anything in the Scope slide does not sound exactly like the role you need, we will tell you so within the first ten minutes. No preparation required.

Book a scoping call Get in touch
ecofrontiers.xyz X LinkedIn

© 2026 Ecofrontiers SARL