The problem
The operator was distributing content across several channels, and the work of tracking, scheduling, and reviewing it lived in four different places: spreadsheets for the schedule, separate threads for review, manual logs for what shipped, and individual memory for the rest. Nothing was the source of truth, nothing was auditable, and review was slow because reviewers had to assemble context from scattered tools before they could judge a single item.
The brief
Give us one place to track, schedule, and review everything before it goes out. Score quality automatically so reviewers start from a signal instead of a blank slate. Role-based access so the right people see the right things, and an audit trail so we can answer what happened and when.
What we built
- A single internal dashboard for tracking, scheduling, and reviewing every item across channels.
- Role-based access control so each role sees and does only what it should.
- An audit trail that timestamps and attributes every meaningful action, so the review history is reconstructable from the platform, not from email.
- An AI agent for automated quality scoring, so each item arrives at review with a quality signal already attached.
What it delivers
- Four prior workflows (separate spreadsheets, separate review threads, manual logging, and individual memory) collapsed into one tool.
- Reviewer time per item roughly halved, because the context and the quality signal are in one place.
- A complete audit trail on every item, replacing reconstruction-from-email.
The approach
- Model the real workflow first. We mapped how the operator actually moves an item from draft to published, then built the tool around that, not around a generic project-management shape.
- One source of truth. The dashboard replaced the scattered tools rather than sitting alongside them, which is the only way the audit trail stays honest.
- AI where it earns its place. The quality scoring is an agent because the judgement is fuzzy; the scheduling and access control are plain deterministic software because they should be.
The stack
- Web: a Next.js frontend with a clean, fast, operator-facing interface.
- Data: PostgreSQL with Prisma, role-based access enforced at the data layer.
- AI: a model-agnostic agent for quality scoring, integrated into the review flow.
- Infrastructure: AWS in ap-southeast-2, AU-region by default.
What this engagement says about how Bedstone works
- We collapse tool sprawl into one source of truth rather than adding a fifth tool to the pile.
- We put AI only where the judgement is genuinely fuzzy, and keep the rest deterministic.
- We build the audit trail in from the start, because a review tool without one is just another inbox.
Want a reference call?
If you are evaluating Bedstone for a similar engagement, we can arrange a direct conversation with the relevant client under appropriate confidentiality. Start a brief and we will scope the right reference for your situation.