Horizon LabsHorizon Labs

Insights

Product, design, AI, and engineering perspectives from our team.

ML Proof of Concept to Production: Why Most Stall
AI in production1 Aug 2026

ML Proof of Concept to Production: Why Most Stall

Most ML proof of concepts never reach production — not because the model is wrong, but because reproducibility, deployment, monitoring, evaluation, versioning, and lineage were never built. Here's why POCs stall and what a realistic path to production looks like.

6 min readChris Kerr
AI Incident Response: What to Do When Your Model Fails in Production
AI in production25 June 2026

AI Incident Response: What to Do When Your Model Fails in Production

When an AI model fails in production, the failure is often silent — no error code, just degrading outputs. This guide is a practical incident response playbook for ML and LLM systems: detection, severity classification, rollback, stakeholder communication, and post-incident review, built for technical leaders who need to extend their existing incident processes to cover AI-specific failure modes.

13 min readChris Kerr
Structured Outputs and Function Calling: Making LLMs Reliable
AI in production23 June 2026

Structured Outputs and Function Calling: Making LLMs Reliable

Structured outputs and function calling are the mechanisms that make LLMs viable in production workflows — but reliable implementation requires deliberate schema design, validation layers, and observability from the start. This guide covers how to use both patterns effectively, when to choose each, and the failure modes that catch teams off-guard at scale.

9 min readChris Kerr
AI Inference Optimisation: A Production Decision Guide
AI in production6 June 2026

AI Inference Optimisation: A Production Decision Guide

Shipping a model is the beginning, not the end. Once an AI feature is live, inference cost and latency become real engineering problems that compound at scale. This guide explains the key optimisation levers available to technical leaders — and how to choose between them.

9 min readChris Kerr
Multi-Model Routing: Cut LLM Costs Without Sacrificing Quality
Costs & budgeting6 June 2026

Multi-Model Routing: Cut LLM Costs Without Sacrificing Quality

Multi-model routing sends each AI request to the cheapest model capable of handling it well, rather than routing everything through a frontier model. Combined with semantic caching and fallback chains, it is one of the most effective ways to control LLM costs in production without degrading quality. This post covers the core patterns and what a practical production architecture looks like.

9 min readChris Kerr
AI Agents in Production: Lessons from Real Enterprise Deployments
AI in production22 May 2026

AI Agents in Production: Lessons from Real Enterprise Deployments

Enterprise AI agents require careful orchestration, robust failure handling, and strategic cost management to work reliably at scale. Here's what we've learned from real deployments.

6 min readSarah Mitchell
Production AI Agents: Architecture and Deployment Lessons
AI in production5 May 2026

Production AI Agents: Architecture and Deployment Lessons

Production AI agents require careful orchestration, robust failure handling, and strategic human oversight. The gap between proof-of-concept and production-ready agents is significant — here's what actually works.

7 min readSarah Mitchell
AI Security: Protecting Your Models, Data, and Users
Governance & privacy1 May 2026

AI Security: Protecting Your Models, Data, and Users

Production AI systems face unique security challenges from prompt injection to model theft. Learn how to protect your AI models, data, and users with practical security measures designed for Australian organisations.

6 min readTom O'Brien
AI Guardrails: How to Prevent Your AI From Saying Something Dangerous
Governance & privacy15 Apr 2026

AI Guardrails: How to Prevent Your AI From Saying Something Dangerous

AI systems in production can generate harmful, biased, or sensitive outputs without proper safeguards. Learn how AI guardrails protect your business through content filtering, PII detection, output validation, and human-in-the-loop safety patterns.

6 min readChris Kerr