About Services Ethics Method Did you know Use cases Blog FAQ Contact
AI Strategy for Impact Organizations.

You have the expertise. AI has the power.
I build the link.

Je construis l'architecture qui les relie, pour transformer votre savoir institutionnel en intelligence hybride.

Scroll
About

Impact isn't held back by will.
It's held back by architecture.

I'm an AI strategy consultant for impact organizations. I come from the field. I know what an 85% collection rate means, how a donor project cycle works, what a MEAL report looks like. That domain knowledge is what makes the architecture I design reliable in practice, not just on paper.

I help impact organizations turn their data and institutional knowledge into usable intelligence. AI is the means. Data architecture, context engineering and automation are the tools. I audit, design, prototype and hand over. My work ends when your organization can carry it forward without me.

More about my background →
What I do

Three ways to build hybrid intelligence.

I work with humanitarian and development organizations, microfinance institutions and impact finance structures. Wherever complex data and deep professional expertise are ready to be connected.

Make your data usable

Data in silos, disconnected systems, knowledge in people's heads. I connect your internal sources, create a single source of truth and document what your organization knows before it walks out the door.

Data architectureInstitutional memoryContext engineering

Make your organisation smarter

Actor reports, sector projections, public datasets, geopolitical and security data. I monitor, cross-reference and synthesize what your organization doesn't have time to read alone, to turn scattered information into actionable intelligence.

Contextual analysisSecurity monitoringPublic datasets

Make your teams faster

Donor reporting, project proposals, compliance analysis, blended finance deal structuring, document synthesis. Your teams decide. AI handles the heavy lifting.

Donor reportingProject proposalsBlended financeMEAL
Ethics charter

AI as a tool for human augmentation.

The AI I design must be trustworthy. For your teams, your clients, your beneficiaries and your partners. Ethics isn't a constraint added after the fact. It's a design requirement.

This charter is mine. I also write yours: a custom AI ethics framework, grounded in your data and your decisions, enforceable with your partners and readable by your board.

Read the full charter

Sensitive data protection

Systematic Data Processing Agreements (DPAs), AI architectures designed to minimize exposure of sensitive data and reduce the risk of leaks or misuse.

GDPRDPANDAAI Act

Human control & transparency

No tool goes live before the teams using it understand how it works, where it can go wrong, and what its limits are. The final call always stays human.

ExplainabilityHuman decisionTeam training

Digital sobriety & environmental impact

Every solution is sized to actual need. I favor lightweight architectures and energy-efficient models, and I won't deploy tools that are oversized relative to the actual need.

AI sobrietyCarbon footprintResponsible tech
Approach

Five steps. One rule:
prove before you invest.

Understand your data before picking a tool. Prove the use case before investing. Hand over the knowledge before stepping away.

1

Audit & mappingUnderstand first

Understanding your data, systems, processes, information flows and constraints. Mapping key processes. Identifying high-value use cases. Ruling out those that don't hold up in your reality. → Data and process map, prioritized opportunity matrix.

2

DesignContext engineering

Designing the architecture and documenting the business context: definitions, management rules, procedures, edge cases. Turning tacit knowledge into context the machine can use. In liaison with your IT department where needed. → Target architecture, roadmap, context documentation.

3

PrototypeThe proof of concept

One use case, a concrete tool tested with the people who will use it. Lightweight automation, dashboard, small application, security monitoring tool, proposal writing assistant, beneficiary tracking dashboard. Not a trade show demo. A prototype fed by the business context designed in phase 2 and validated in your real conditions. → Working prototype, documentation. 3 to 6 weeks.

4

ValidationGo or no-go

Testing relevance, output quality, reliability, traceability. Measuring user adoption. Identifying risks and limitations. Clear recommendation: move forward, or stop. And why. → Evaluation report, go / no-go recommendation.

5

TransferYour autonomy

Training your teams on tool use, critical reading of outputs and context engineering. So they can evolve the tool when procedures change or a donor updates its requirements. Handing over documentation, methods, governance. Drafting your AI ethics charter if needed. The goal: your organization decides what comes next. → Training materials, documentation, charter, procedures. 2 to 4 weeks.

I design, prototype and hand over. I don't build production systems and I don't bill for maintenance. Your organization stays sovereign over its technology.

In numbers

AI is already at the core of impact.

The numbers say it best. AI isn't a promise anymore, it's a measurable operational lever, already adopted by the organizations that moved first.

26–31%
in cost savings across supply chain, finance and HR thanks to AI
MasterOfCode Cross-Study, 2025–2026
75%
of companies plan to adopt AI within five years. Fewer than half have restructured their data to receive it
World Economic Forum, Future of Jobs Report 2023
–40% / +25%
of time saved processing ESG data, plus better risk predictions thanks to AI
Mezzi ESG Screening, 2025
FAQ

What you often ask me.

AI in impact organizations raises real questions. Here are honest answers to the most frequent ones, no jargon, no detours.

This is almost always the first question teams ask, and it is a legitimate one. The honest answer: no, not in the way most people imagine.

AI augments, it does not replace. According to the McKinsey Global Institute (November 2025), available technologies could technically automate approximately 57% of working hours — this is a theoretical potential, not a forecast of job cuts. What disappears with AI are the tasks no one enjoyed: manual data entry, compiling tables, formatting reports.

What remains — and strengthens — is what AI cannot do in your place: direct engagement with beneficiaries, field judgement, ethical decision-making. And this is precisely where augmented AI comes in: AI does not decide, it informs. It synthesises field data in real time, flags anomalies in a loan portfolio, generates a situation report ahead of a coordination meeting. Your team arrives better informed, faster, and makes better decisions.

The real threat to employment in your sector is budget cuts, not AI. AI is not the problem — it is part of the solution for doing more with less, without sacrificing your teams.

Source: McKinsey Global Institute, November 2025.
A generalist AI consultant doesn't know PAR scoring, SPI4, donor funding cycles, or field realities. They adapt a standard solution to your context, which costs time and generates misinterpretations. An IT developer delivers solid code but doesn't understand your operational challenges: they execute what they're asked, not diagnose what you actually need. The value is in the combination: deep sector knowledge, the ability to identify the right problems, and delivery of operational tools without putting the sector training burden on your side. In practice, this avoids imprecise scoping, reduces the risk of delivering a misfit solution, and saves you the cost of an intermediary translating your domain to a technician.
I am the only person with access to your organization's real data. Any external contributors work on test data only, unless explicitly and formally agreed otherwise. A Data Processing Agreement (DPA) is signed before any data processing, in line with GDPR (Article 28) and the AI Act. Accesses are limited to the strict minimum and revoked at the end of the engagement. For truly sensitive data, beneficiary, financial, or health data, I systematically favor architectures that keep your data in your own infrastructure: anonymization, sovereign European cloud providers, or on-premise open-source model deployment (Llama, Mistral, Phi). In the latter case, no data ever leaves your systems.
I work on fixed-scope mission contracts. My reference day rate is €500 excl. VAT / day. All figures are indicative and exclude VAT.

Retainer engagement over 3 months: −10 to −20%

Audit and mapping
On-site or remote session
€350 to €700
Complimentary 45-min scoping call, then a 2 to 6h session depending on complexity. Data and process mapping, priority use case identification.
Design and roadmap
Written diagnostic and proposal
€700 to €1,400
Target architecture, context engineering, prioritized roadmap, business context documentation.
Light prototype
Automation, dashboard, monitoring tool
€1,500 to €4,000
Delivery 1 to 3 weeks. One use case, one working tool tested with your teams.
Advanced prototype
Application, document assistant, multi-source monitoring
€4,000 to €12,000
Delivery 3 to 6 weeks. Multi-source tool, automated monitoring, writing assistant, dashboard.
Training and transfer
Handover and context engineering
€700 to €1,400 / session
Half-day to full day. Custom materials included. A handover session is included with every delivery.
AI ethics charter
Custom framework for your organization
€1,400 to €3,500
Enforceable charter, data governance, human oversight rules, operating procedures.

* All fees are indicative and exclude VAT. Tool costs vary and are detailed in every project quote.
Because AI isn't arriving anymore. It's settling in.

Generative AI is becoming a new work infrastructure: for searching, analyzing, writing, comparing, producing and deciding. The movement is happening now.

In two years, the question won't be 'should we use AI?' but 'how does your organization use it, on which data and with which rules?'

Today is when the foundations are built: structured data, documented business context, governance and relevant use cases.

Don't let AI enter your organization by default. Build the architecture that will let you control it.
A prompt gives AI an instruction. Context engineering gives it the elements needed to correctly understand that instruction: institutional documentation, business rules, glossaries, past decisions, data lineage. It's the discipline that turns organizational knowledge into context the machine can use.

In practice: a prompt asks AI to summarize a report. Context engineering explains what an M&E report means in your organization, which indicators matter to your donors, which terms carry specific meaning in your sector, and which cases should trigger a human alert. The difference between a generic answer and a reliable one.
Ready to start?

Let's start by understanding your need.

A 45-minute conversation, free, no commitment.

I send a short questionnaire beforehand, so the conversation focuses on your data, your context and your real challenges. Not on generic topics.

1
Book a slot
online, takes 2 minutes
2
I send the form
under 15 min to fill in
3
45-min call
focused on your stakes
4
Proposal in 5 days
if it's the right fit