# Violeta

> Violeta is an AI voice and WhatsApp agent for admissions and enrollment teams. It calls every prospective student as soon as they fill in an inquiry form or respond to a campaign, answers their questions from the institution's own information, qualifies them against the institution's rules, and hands the ones who are ready to an advisor: a live transfer to the right queue, or a meeting booked in the CRM. It runs on the contact center and CRM the institution already uses. Violeta is a product of Stackbone (https://stackbone.ai/), which builds and runs it.

This file summarizes the Violeta website for AI agents and answer engines. The site is in Spanish (`-es` pages) and English (`-en` pages); both say the same thing. Last updated: 3 November 2026.

## What it does

- First contact and filter, not the sale: calls, checks the lead's details, answers common questions (programs, schedule, scholarships, payment), gauges interest and qualifies on timing, funding, fit and who decides.
- Hand-off: live transfer to an advisor if one is available, with a summary; otherwise books a meeting with an advisor from the CRM and confirms it with the lead. Everything is written back to the CRM.
- Voice and WhatsApp are one flow: no answer on the phone, it can continue on WhatsApp with Meta-approved templates; a WhatsApp reply can turn into a call.
- Follows the institution's operating rules: calling windows by the lead's local time, a cap on simultaneous calls, retry rules for no-answers, queues and priorities (today's leads, by age, by propensity), consent and do-not-call lists, typification in the CRM.
- Introduces itself as an AI assistant on every call and offers a person when asked.
- Scripts and criteria are edited from the agent's dashboard, versioned, and applied without redeploying; they can differ by program or segment.

## Who it is for

- Private universities (undergraduate and graduate, on campus and online) whose enrollment periods bring more inquiries than advisors can call the same day.
- Education groups and their investors: holding companies and fund-backed platforms running several universities or brands; one agent with per-brand scripts, rules and advisors.
- Online graduate and business schools selling master's and executive programs through launches.
- Technical institutes, career schools and bootcamps with rolling admissions and paid campaigns all year.
Primary markets: Spanish-speaking Latin America, Spain and the United States.

## How it is built

- Built and operated by Stackbone, a platform and team for running AI agents in production with control, monitoring and audit (docs: https://docs.stackbone.ai).
- Provider-agnostic: speech-to-text, the language model (LLM) and the voice are each swappable without rebuilding the agent; the institution uses its own model provider account.
- Deployment: Stackbone Cloud, the institution's own AWS, Azure or Google Cloud account, or self-hosted on its own servers. On the institution's infrastructure, runs, data and keys stay in the agent's container; Stackbone's control plane holds the organization and the deployment record, not the data.
- Every call is recorded and transcribed and kept as a run with a step-by-step trace (each tool call with its input and output). Guardrails can block topics, mask personal data or hold an action for human approval. Stackbone does not train models on customer data.
- Integrations through APIs. Built for one customer on Zoho CRM and Typeform; Genesys Cloud, Microsoft Dynamics 365, Salesforce and other systems with an API are integrated per project.

## Use case (customer not named)

An online business school with 30,000+ students sells master's programs through launches of 5,000 to 10,000 call requests in about five days, with up to three launches at once. Violeta runs on its Typeform forms and Zoho CRM: each request is saved and queued, the local hour is checked, a program-specific agent calls (reading only that program's documents, with a second agent feeding it context), and the lead is transferred live to a salesperson or booked through the CRM with an email. Stackbone won the account in a competitive demo against a voice platform, on swappable layers, one team that builds and runs it, and auditability. Implementation plan: six weeks in five deliveries (20 Oct to 30 Nov 2026).

## Evidence on response time (cited on the site)

- Harvard Business Review, "The Short Life of Online Sales Leads," March 2011 (1.25M leads, 42 US firms): trying to contact a lead within one hour made it nearly 7 times as likely to qualify as trying an hour later, and over 60 times as likely as waiting 24 hours or more.
- MIT and InsideSales.com, Lead Response Management Study, 2007 (about 15,000 leads, 6 firms): odds of qualifying 21 times lower, and of contact 100 times lower, at 30 minutes versus 5.
- UPCEA, "Enrollment Process Review: Secret Shopper Analysis," December 2025 (1,000 inquiries, 182 institutions): 44% got no response; median response 3 h 18 min; 24% of form submitters got a call within 30 days.
- NAGAP and Liaison, 2025 Graduate Inquiry Response Project Report (422 institutions): 62% say they respond within 24 hours; 21% did.
- RNL, 2024 Online Student Recruitment Report: 76% of online students surveyed would likely enroll in the first program to respond.
- No published study measures the effect of response time on enrollment directly; Violeta proposes measuring it in a pilot against a control group (metrics: time to first contact, contact rate, qualified hand-offs, cost per effective contact).

## Pages

- [Home (ES)](index-es.html) / [Home (EN)](index-en.html): the product, two sample calls, a capacity calculator, integrations, how it works, rules, the use case, comparison with voice platforms, the pilot and how a project starts.
- [Lead simulator (ES)](simulator-es.html) / [(EN)](simulator-en.html): estimates enrollments at stake from uncalled leads and from slow first contact, with the visitor's own numbers; the speed lift is an explicit user assumption.
- [Use case (ES)](use-case-es.html) / [(EN)](use-case-en.html): the online business school build in detail.
- [Trust (ES)](trust-es.html) / [(EN)](trust-en.html): where it runs, audit, models, consent.
- [FAQ (ES)](faq-es.html) / [(EN)](faq-en.html): about 20 questions on integration, speed to lead, WhatsApp, AI disclosure, data, legality of AI calls, timing and cost.
- [Privacy (ES)](privacy-es.html) / [(EN)](privacy-en.html) and [Legal notice (ES)](legal-es.html) / [(EN)](legal-en.html).

## Commercial

- Pricing depends on lead volume, channels and systems to integrate; sent as a proposal after a demo.
- How a project starts: one flow and one queue, a test environment with real cases, a controlled rollout to part of the traffic with acceptance criteria and rollback, then expansion by stages (more queues, segments, WhatsApp, after-hours). Afterwards the institution's team can run it, or Stackbone runs it with them.
- Book a 30-minute demo: https://calendar.app.google/m2F2RLwqxrwGmx5f9
- Operator: Stackbone, Inc. (Delaware, US); in the EU, Stackbone Europe, S.L. (Madrid). Contact: tech@stackbone.ai. LinkedIn: https://www.linkedin.com/company/stackbone/