AI & Process Automation
Production AI, not proof of concept. Agents that replace manual work end to end, with the monitoring to prove it's running and a person to call when it isn't.
Most AI projects stall because the gap between a demo and a production system is bigger than anyone planned for. The demo impresses; the production version is where the edge cases, the data quality problems and the maintenance burden show up. We build automation that runs in the real world, with the guardrails and monitoring to prove it.
We've shipped production AI across our own products. Anonamoose is our self-hosted PII redaction proxy, built so AI processing can happen without sensitive data leaving the environment. Polly pairs Anonamoose with frontier models from OpenAI and Anthropic — built for NZ healthcare, where patient data can't go near an external model. AI only earns its place where it replaces work a human was doing; if a plain script or a rule engine solves it better, we build that.
- Process audit to identify where automation creates real value, not just activity
- AI agent development for repetitive decision-making tasks: triage, classification, extraction, routing
- Document and data processing pipelines that handle real-world inputs, not clean demo data
- Integration into your existing systems — email, CRM, ERP, databases, APIs
- Monitoring and alerting so you know when the automation is working and when it isn't
- Handover with enough documentation that your team can maintain and adjust it without us
Quick questions
How do we know AI is the right tool?
We'll tell you if it isn't. Automation that doesn't involve learning or judgment is often better served by a straightforward script or workflow. AI adds overhead — model costs, prompt maintenance, evaluation — that's only worth it when the problem actually requires it. We start from the problem, not the technology.
What about our data staying private?
We take that seriously, and we've built products specifically around it. Anonamoose is our own PII redaction proxy, designed so data is scrubbed before it touches a frontier model. For sensitive environments, we can keep processing entirely on-premise. The architecture depends on your requirements.
Can you work with OpenAI or Anthropic models, or do you use something else?
Both. We work with frontier models from OpenAI and Anthropic, and with open models where self-hosting matters more than capability at the frontier. The choice depends on the use case, the privacy requirements and what actually does the job.
We've shipped systems still running a decade later.
Custom software, AI agents and automation for real business problems. We design, build and run production systems — not prototypes. We ship working software that scales from a first version to national platforms. All of Build →
Need this sorted?
Tell us what you need. A real person replies — not a bot, not a call centre.