Aida
Find an administrative workflow worth automating and compare business choices through visible criteria and calculations.
Aida helps a practitioner decide where to act first
A solo practitioner has limited time to evaluate new tools and business changes. Aida offers two paths: find administrative work that an AI tool could reduce, or compare a set of business options against the practitioner’s constraints. Both paths show the basis for the result so the person can review the tradeoffs before acting.
The automation path ends with a usable starting point
A short conversation identifies recurring work such as billing, scheduling, or referral coordination. Aida scores the candidates for time burden, impact, AI fit, risk, and setup effort. It classifies a promising reducer as a skill, plugin, agent, or human task, then produces a paste-ready scaffold for a suitable tool. The aim is to move from “AI might help” to a concrete first workflow.
The decision path exposes the ranking
For a discrete choice such as hiring support or changing a workflow, Aida records alternatives, criteria, weights, and constraints. Hard constraints remove options that do not fit. ELECTRE-style comparisons identify strong dominance; TOPSIS ranks the remaining alternatives by closeness to an ideal option. The result includes a fallback and a trace of the calculations behind the recommendation.
A model helps classify the question and propose short reasons. TypeScript performs the numerical scoring and ranking. The model cannot write a final score and then explain it after the fact. Aida declines questions outside a discrete decision, including diagnosis and continuous optimization, instead of forcing them through the same ranking method.
The architecture keeps intake, scoring, and records separate
Next.js serves the interface and decision API. Groq handles bounded classification tasks, while TypeScript runs the decision math. Neon Postgres and Drizzle store authenticated decisions and an audit of model calls. Guest users can preview an output without saving it. Intake rejects protected health information before the decision enters the pipeline.
The output is a decision aid, not a clinical recommendation. Its trace makes assumptions and tradeoffs visible; whether adopting a recommendation saves practitioner time still needs real-world measurement.