AI assistants that formulate international financing projects.
Formulating a project for the IDB, CAF or the European Union takes months of specialised work. Six assistants build the causal logic, the indicators and the economic analysis according to the funder's methodology — with documentary evidence behind every claim, and the final decision in a person's hands.
The economics are computed by a module, never written by a language model. A person signs.
Built on the frameworks and public data of
Reference base loaded today: IDB — 27,329 operations, 507 pages of standards in force, 1,030 debarred firms. The rest are on the roadmap.
The project is viable. The formulation is what leaves it out.
Most proposals are not rejected on merit: they are rejected because they do not meet what the terms demand. The gap is technical, and it is expensive to close.
Months of specialised work
Theory of change, logical framework, indicators, economic analysis and safeguards. Each component demands a different technical profile.
14% approval rate
That is the success rate of programmes such as Horizon Europe. The rest of the technical and financial effort is lost.
10% of the loan
The standard fee of a specialised consultancy. For a large share of organisations it makes the attempt unviable.
Objections that arrive too late
The committee objects when there is no longer time to correct. An internal evaluator assistant objects before submission, while it can still be fixed.
The assistants
Six assistants, one case file.
Each one does one job, in order, and hands the next one what it produced. The result is the document the funder expects — not a summary of it.
01
The Strategist
Reads the project and 27,329 comparable operations, and writes the Theory of Change.
02
The Formulator
Turns it into a Results Matrix, with the call's terms in view.
03
The Evaluator
Objects to the matrix as the committee would, up to three times, before anyone else sees it.
04
The Economist
Structures the cash flow; NPV, IRR and B/C are computed by a module, never by the model.
05
The Manager
Assembles the complete case file with every section the funder requires.
06
The Auditor
Checks it against the standards in force, safeguard by safeguard, citing the page.
Documentation
The system's rules are public
Every rule the system applies is written down and available: which methodological framework governs, what each assistant may assert and what it is forbidden to do. And they are not enforced only by the screen, but by the system itself: a rule that only lives on the screen can be bypassed another way.
Frequently asked questions
What comes up most often in the first conversation.
Does it impose a methodology?
No. It is configured from each call's own terms: the terms are the standard. What the system does require is that every judgement carries its evidence.
How long does a formulation take?
A complete formulation with the six assistants takes between 40 and 45 minutes. The internal quality loop is the slowest stage, and the one that justifies the wait.
What happens if the AI is wrong?
It is designed to be detected: every claim carries the page it came from, so it can be checked. And nothing is issued without a person's signature.
The project already exists. What is missing is the formulation that makes it compete.
An invitation, the project's information, and the six assistants build the case file.