VocxAI — From Customer Feedback to GitHub Pull Requests

What is VocxAI?

VocxAI is a web platform that turns customer feedback and product usage data into shipped software. It sits at the intersection of product management and engineering: the system ingests signals from tools such as Gong, Zendesk, Slack, Salesforce, G2, GitHub, Nolt, and Amplitude, classifies that feedback into themes automatically (with deduplication, spike alerts, and PII scrubbing), and then ranks opportunities against a goal the team sets — ARR, adoption, churn risk, or a custom metric. From there, AI agents do the building: they analyze the codebase through a semantic code graph (the site says only metadata is indexed and "source code never stored"), generate PRDs and task decompositions with confidence scores, and write code with Claude-based "cloud agents" that stream live logs and track per-job cost against spend limits. The output of every build is a GitHub pull request — the site is explicit that "nothing merges itself." No company, founder, or location is named; the footer reads "© 2026 VocxAI. Voice of Customer × AI."

How VocxAI Works

The product runs a continuous loop in five stages. First, set a goal so the system knows what "valuable" means. Second, Listen: connectors pull in feedback and usage signals, which are auto-classified into themes and deduplicated. Third, Prioritise: themes are ranked with evidence against the goal, so product owners see the biggest opportunities rather than the loudest requests. Fourth, Build: agents produce a spec, decompose it into tasks, write the code, and open a pull request; humans approve before anything lands, and extra builds queue automatically when a slot frees up. Fifth, Learn: post-release tracking feeds results back into the next prioritization cycle. Case studies on the site cite ScreenlyyID moving 15 roadmap items to pull requests in about a day and Workforce Directory going from spec to PR in an hour, though these are self-reported.

Main Features

• Feedback ingestion — connectors for Gong, Zendesk, Slack, Salesforce, G2, GitHub, Nolt, and Amplitude with auto-classification, deduplication, spike alerts, and PII scrubbing.
• Goal-based prioritization — ranking by ARR, adoption, churn risk, or custom goals, with evidence attached to each theme.
• Semantic code graph — codebase analysis that indexes metadata only, with source code never stored.
• AI PRDs and task decomposition — generated specs with confidence scores for each task.
• Claude-based build agents — live log streaming and per-job cost tracking with spend limits.
• GitHub PR automation — pull requests with human approval gates; nothing merges itself.

Pricing

VocxAI uses a credit model where credits meter AI work across feedback analysis, data chats, PRD drafting, and build agents. The Free plan grants 500 one-time credits that never expire, includes the full platform, allows 1 build at a time, and supports 2 connectors plus GitHub — GitHub never counts toward connector limits — with no credit card required. Pro costs $49 per month (labeled "Most popular") for 4,900 monthly credits, 2 concurrent builds, and 5 connectors plus GitHub; a credit card is required. Team costs $99 per month for 9,900 monthly credits, 5 concurrent builds, and unlimited connectors. Unused credits on paid plans do not roll over; when credits run out, the free plan pauses AI features while Pro and Team get a small buffer followed by automatic $10 packs of 1,000 credits. Payments run through Stripe, upgrades are prorated and immediate, and downgrades take effect at the end of the billing period. No refund policy is mentioned.

Common Use Cases

• Startups without product ops — teams that collect feedback across Slack, Zendesk, and sales calls but lack a process for prioritizing it.
• Product managers and CPOs — controlling what ships with goal-ranked themes, evidence, and cost visibility per job.
• Teams needing extra engineering capacity — using build agents as embedded capacity for a whole department.
• Feedback backlog reduction — converting months of scattered requests into ranked, spec-ready opportunities.
• Post-release validation — tracking whether shipped items actually moved the goal metric.

Limitations

VocxAI is early-stage: the site addresses "early adopters" and demo dashboards use sample data, so real-world reliability is not yet publicly proven. Its headline statistics — 73% of features unused, $2.1M ARR loss, 14-week cycle times — are unverified marketing claims, and the case studies are self-reported. The workflow requires human approval at each step, so it accelerates but does not automate shipping. Paid-plan credits expire monthly without rollover, and overages trigger automatic $10 charges, which teams should budget for. Connector limits constrain smaller plans (2 on Free, 5 on Pro), and real use demands connecting multiple data sources plus adopting the PR-approval workflow. No refund policy is published, and no company details are disclosed.

Summary

VocxAI sketches a complete, unusually honest loop: listen to customers, rank opportunities against a stated goal, draft the spec, build with AI agents, and stop at a pull request a human must approve. Transparent credit pricing, a non-expiring free grant, and per-job cost limits make it easy to evaluate. The reservations are its pre-GA maturity, self-reported results, monthly credit expiry, and the organizational buy-in required to wire it into real feedback and review pipelines. It is best suited to lean product teams ready to treat AI-built PRs as first-class candidates in their existing review process.