Products and services
Grounded in research
We build each product around a use case and a business need. Many of their core capabilities started as experiments. We validate each experiment, and only the ones that work go into a product. We can reuse a product's features and capabilities to make new products and deliver new services.
Products
Tessa Van Wyk
Three follow-ups, no reply
James Coetzee
Asking about renewal terms
Aisha Khan
Two paid invoices, fast
Tessa Van Wyk
Sentiment trending down · 9d silent
AI recommendation
Pattern matches 73% of churn risks last quarter. Call her before Friday.
Relationship intelligence
RapportIQ
Reads every conversation your team is already having. Surfaces who is going cold, who is ready to buy, and who is at risk. Recommends the action, with the reasoning. Plugs into any AI agent through the Model Context Protocol (MCP).
Naledi Sithole
Agent platform
Agent Platform
Build and run AI agents across every channel your customers already use. The contact centre comes built in: WhatsApp forms and voice notes, inbound and outbound calls over SIP or WebRTC, embedded web chat, email triage and case routing, and agent-assist for the human team on the floor. Behind it sit workflows with escalation branches, identity verification, knowledge bases with retrieval-augmented generation (RAG), MCP tool access, guardrails, and observability, all from one console, self-hosted on your node.
$/feature add user-auth
✓feature.md created
$/plan
✓7-step plan written
$/build
✓claude takes the keyboard
$/memory save
✓context preserved
$
Way of working
Agentic WoW
Practices, templates, and tooling for teams building software with AI agents. Plans become first-class artefacts. Memories carry context across sessions. Reviews catch what the model missed.
Memory and context
Neuro
A private memory and context platform for AI agents. Codex, Claude Code, Copilot, and your own agents share one team memory, with each record scoped to a user, team, product, or environment. Retrieval is governed, handovers are durable, and secrets and sensitive memories stay in Vault. One context gateway serves every agent over MCP, across repositories, services, sessions, and machines. Answers carry their source and lineage, and retention runs past 5 years without re-prompting. Self-host it, or run it on your NPU Labs node.
Customer message
“I was charged twice for my subscription.”
Flash
- Classify
- Screen
- Score
- 0.97
Classification
Billing
- 0.99
Guardrail
Pass
- 0.96
Escalation
Handover
- Billing0.97
- Pass0.99
- Handover0.96
Your policy
Thresholds in your software
- Billing ≥ 0.90Billing queue
- Guardrail passesContinue
- Handover ≥ 0.90Human agent
Decision model
Flash
A small decision model for the bounded decisions inside an AI system: which category a request belongs to, which route to take, and whether another system should take over. Flash returns a score with its confidence. Thresholds, escalation rules, and business policy stay in your software, so they change without retraining the model. Flash is in its baseline and validation phase. It targets ONNX Runtime, with CPU deployment as a first-class requirement.
Common questions
Are these products, or are they custom builds?
Both. You get the finished product, configured to your environment, and we can reuse its features and capabilities in a custom build.
Do they run on our own infrastructure?
Yes. RapportIQ, the Agent Platform, Agentic WoW, and Neuro run on a private NPU Labs node in your environment, and they connect through the same MCP and workflow primitives. Flash is in its baseline and validation phase, and is being built to run on ordinary CPUs as well as private compute.
Can we buy one product without the rest?
Yes. They share a stack, so running two is cheaper than running two separate systems, but each stands on its own.
What if we need something none of these do?
Tell us the workload. We will tell you whether one of these products fits it, or what we would need to build.
How does pricing work?
Software sits on top of the node you own, so there is a build and configuration cost and a support arrangement, rather than a per-seat subscription that grows with your headcount.
Do you offer services as well as products?
Yes. We offer AI testing and evaluation as a service in its own right: golden scenarios, release quality gates, safety and privacy testing, and continuous evaluation, run on your AI or used to assess it against our ISO/IEC 42001-aligned framework.
Not the question you came with?
Ask us about your setup