Home / Customers
Customer stories
Agents at work, on real operations.
How organisations use Camaan and Relevance Lab to put agents to work on real data, under rules, with proof.
Supply chain, public sector, aquaculture, lighting, insurance and a non-profit.
- NexusNow
- Action Lab 2050
Reality Scale
- The Maze
Supply chain
Redefined Logistics, New York
A supply-chain decision platform, built entirely on Camaan.
Redefined Logistics built DropSkip, its decision layer for direct-to-consumer brands, on Camaan. Connectors bring the data in, scenarios turn it into reports the team can act on, and agents propose what to buy, move and when.
The challenge
Brand operators decide what to buy, how much, when and where, with the facts spread across their store, ERP, warehouse, carrier and customs broker. A bestseller runs short in one warehouse while the same product sits in another, and a new purchase order commits cash a transfer could have saved.
What was built
Connectors read each system into one shared model. Scenarios produce the reports: the same question asked in November runs the calculation it ran in August, on November's data. Planning and forecasting agents go from a signal to a proposed action, a transfer, a purchase order or a markdown, that a person approves, adjusts or dismisses with the reasoning in view.
The result
One connected view of demand, stock and cash, and recommendations a planner can explain. Camaan runs inside DropSkip's own server as a library, so the product ships as one process.
Public sector
A national wildlife sanctuary, India
Fire watch across a forest reserve, from camera towers to the control room.
For a state forest department, we built and run the system that spots fires and smoke from camera towers, places them on the map and puts every alert in front of a ranger before anyone is sent.
The challenge
A large reserve of forest and hills, watched by a handful of staff. Fires have to be found early, placed accurately on difficult terrain and sent to the right patrol, and night-time fog and glare mean a camera alone cannot be trusted to decide.
What was built
AI cameras on towers and trap cameras across the reserve feed a vision pipeline. Each detection is placed on the map by line of sight over the terrain, matched to its patrol cell and forest beat, and queued for a ranger to confirm or dismiss. Field teams get tasks; the control room sees towers, network devices and patrols on one map. It all runs on the department's own servers.
The result
Every alert is reviewed by a person, and every decision is on record. The detection model is tuned on what rangers reject.
Aquaculture
Oceanova Exports
Prawn grading by vision, not by hand.
Oceanova built an app on Camaan that classifies prawns with vision and vision-language pipelines, replacing thousands of hours of manual sorting.
The challenge
Prawns are sorted by species, size and quality by hand, tray after tray. It is slow, tiring and inconsistent between people and shifts, and the grade decides the price.
What was built
Cameras on the line capture each tray. A vision model finds and measures every prawn and a vision-language model checks quality, with workflows that route each batch and record its grade. Built as a Camaan app, so the pipeline, the people and the records share one set of rules.
The result
Grading that is quicker and the same on every shift, with a record for every batch.
Lighting
Traxon e:cue
One product catalogue, many jobs: web pages, answers and quotations.
Traxon uses Camaan to manage its products, turn product details into web pages, answer questions from clients and leads, and draft quotations.
The challenge
Thousands of fixtures, each with detailed specifications, kept current across a website, sales conversations and quotes. Every question from a lead means someone looking it up.
What was built
Agents keep the product data in one place, generate web pages from it, answer client and lead questions from the specifications, and draft quotations for the sales team to review and send.
The result
Product information that is right everywhere it appears, and quotes that start from the catalogue rather than a blank page.
Insurance
An insurance company, India
Outbound calls by voice agents, in English, Hindi or both.
An Indian insurer uses Camaan Voice for outbound calls to motor and health customers, started by its CRM and recorded back to it.
The challenge
Renewal and follow-up calls take a lot of people, in more than one language, and the result of each call has to reach the CRM so the next step happens.
What was built
The insurer's CRM starts each call. A Camaan voice agent speaks with the customer in English, Hindi or a mix of both, works from the insurer's own call scripts, and writes the outcome to the CRM during the call. Guardrails keep it to what the script allows.
The result
Calls placed as soon as the CRM asks, in the customer's language, with every outcome recorded.
Non-profit
Action Lab 2050
From public reports and private files to tables a policy team can defend.
Action Lab 2050 runs a data platform on Camaan that collects unstructured public and private data, extracts the tables inside it for analytics, and has agents write reports from the result.
The challenge
Policy work runs on census tables, survey reports, departmental spreadsheets and PDFs with the numbers buried in prose. None of it lines up, and every figure has to be traced to its source.
What was built
Upload a PDF, spreadsheet or report and extraction agents route its figures into the right tables, one, several or none. Nothing enters a table until a data steward accepts it. A geography spine from country to village keeps places comparable. Agents then answer questions and write reports from both the tables and the documents, with every change on a hash-chained audit trail.
The result
Comparable, sourced observations that can be queried and charted, and reports that show their working. One example from national health survey data: anaemia rose in 14 of 21 states between two rounds.