Solution stories
What we have built.
Anonymised stories from our services work. Each one describes the problem, what was built and how it is put together.
BFSI (UK) · Conversational AI and analytics
AI sales intelligence for an insurance broker
The problem
A large, multi-region sales team spent hours assembling client intelligence before every call, with no unified view and no AI assistance.
What we built
- A client-intelligence CRM with a single-screen client view
- Renewal-trigger alerts and client risk scoring
- Retrieval-grounded answers with cited sources
- Automated client briefings
How it's built
- Single-page app
- Role-based views
- REST API
- Portfolio scoring engine
- LLM briefings with fallback
- Data visualisation
BFSI · Conversational AI (VocoMotiv)
Conversation-led loan origination for digital lenders
The problem
Digital lenders lose leads in the hours before a tele-caller reaches them; eligibility waits on bank statements emailed in pieces, and only a sample of calls is checked for mandatory disclosures.
What we built
- A VocoMotiv agent built from the lender's SOP running lead-to-application on voice and WhatsApp
- Account Aggregator consent over WhatsApp, eligibility from the lending engine, KYC and the LOS application
- Relationship manager on exceptions only
- QA agent checks every call for disclosures
How it's built
- ASR and TTS
- LLM router
- Director model
- MCP connectors (Account Aggregator, LOS, CRM, payments)
- A2A hand-off to QA agent
- VPC or on-premise with BYO keys
- Consent, DNC and audit trail
EdTech · AI platform
An academic lifecycle platform with AI answer-sheet grading
The problem
Attendance, fees, exams and grading lived in disconnected systems; students, faculty and parents chased information across platforms while handwritten answer sheets waited to be graded.
What we built
- One platform for the student lifecycle
- AI OCR and subjective grading of handwritten answer sheets, with educator override on every grade
- Real-time academic dashboards
- Role-based access
How it's built
- Cross-platform app
- REST APIs with role-based authentication
- OCR and AI grading engine
- Weak-topic and curriculum-gap analysis
- Audit logs
Healthcare · Regulatory AI
Provider and diagnosis-code validation with MCP servers
The problem
A healthcare provider platform validated provider identities against the national registry and mapped diagnoses to ICD-10 codes by hand, causing claim errors, rework and compliance risk.
What we built
- An NPI MCP server for real-time provider validation
- An ICD-10 MCP server for diagnosis coding
- A claims pipeline: SFTP intake → OCR → code tagging → storage
- An AI agent that cross-checks provider and diagnosis codes and flags anomalies before submission
How it's built
- MCP servers
- Node.js
- REST endpoints
- Exclusion-list checking
- Multi-dimensional search
BPO · Analytics AI
Post-call intelligence for a contact centre
The problem
A contact centre relied on manual call sampling that missed quality and compliance issues, and leadership had no real-time view of agent performance, resolution trends or sentiment.
What we built
- Automatic ingestion of every call
- Transcription with speaker diarisation
- Post-call sentiment, intent, compliance flags and resolution status
- Cross-call trends, repeat issues, agent performance scoring and supervisor alerts
How it's built
- SFTP ingestion
- Real-time processing
- Speech-to-text
- Sentiment and intent analysis
- Compliance flagging
- Drill-down dashboards
Have a workflow like this? Talk to us.
Bring one workflow. Leave with a governed system running in your environment.