How a UK Property Management Company Cut AI Spend by 64% and Gained Full AI Governance in Less Than 30 Days
Residential Property Management
15+ Employees Across Operations, Sales, and Support
Birmingham, United Kingdom
J2 Lettings Group, a digitally-forward property management company based in Birmingham, had started adopting AI tools across its tenant lifecycle — from answering FAQs and screening applicants, to generating tenancy agreements and processing maintenance requests.
Their stack included OpenAI's GPT-4 and GPT-3.5, Anthropic's Claude, and Google PaLM 2 — accessed via multiple teams using disparate tools, often with little coordination.
However, within 6 weeks of adoption, their AI usage had become a financial and operational liability. Monthly spend ballooned to £22,000, with no visibility into how that money was being spent or if it was driving meaningful value. Prompts were being fired off by ops, lettings agents, and back-office teams — without governance, consistency, or accountability.
"We weren't against spending on AI — we just needed to know what it was doing. But we had zero insight. The cost was growing faster than the value."
Trace Optic's engagement with J2 Lettings started with a bootstrapping phase. Our engineers ran a passive, zero-code deployment that instrumented their AI activity via proxy endpoints and access logs. This observational window allowed us to gather granular telemetry across their most active AI workflows — identifying patterns, bottlenecks, and areas of waste.
We captured and classified 5,000+ prompt events over a 7-day period and mapped them to:
Task type
Model used
Token load
Outcome latency
Redundancy score
This data laid the foundation for Trace Optic's automated optimisation engine to take action.
Trace Optic dynamically routed prompts in real-time based on complexity, urgency, and context. Basic queries were pushed to cheaper models while more nuanced legal drafting used GPT-4.
The system applied intelligent debouncing logic for high-frequency, low-variance prompts — cutting redundant calls by nearly 50%.
Department-level budget ceilings, usage caps, and rate-limiting controls with real-time alerts and custom escalation workflows.
Every AI interaction logged with tamper-proof, cryptographically hashed records including actor identity, purpose, model used, and tokens spent.
Trace Optic integrated with internal CRM and ticketing systems to tie AI usage to actual outcomes like lead conversion and SLA adherence.
Within the first month, they cut AI spend by 64% and reduced unnecessary token usage by nearly half. Most importantly, they could finally trace AI usage back to real outcomes — from tenant satisfaction to response resolution times — while generating compliance-ready logs for audit.
What began as an experiment in visibility quickly became a core operational layer for financial control and risk mitigation.
| Metric | Before Trace Optic | After Trace Optic |
|---|---|---|
| Monthly AI Spend | £22,010 | £7,985 (↓64%) |
| Redundant Prompt Traffic | Unmonitored | Reduced by 49% |
| Smart Routing Efficiency | None | 85% of prompts optimised |
| Visibility Into Usage | Absent | Real-time dashboards |
| Governance & Audit Readiness | Manual, patchy | Full compliance logs |
| Business KPIs Attribution | None | Mapped to 6 metrics |
| Time to Deploy | — | < 24 hours |
With Trace Optic, J2 Lettings didn't just cut costs — they transformed AI from an uncontrolled expense into a governable, attributable asset.
They now review AI spend weekly at director-level meetings with automated usage reports by department. AI usage is tied directly to business goals, and further rollouts are pre-approved based on model performance and cost metrics, not guesswork.
"It's not just about reducing spend — it's about knowing what AI is doing for the business. Trace Optic gave us that control, with zero friction."
J2 Lettings is now piloting Trace Optic's agentic orchestration features to extend usage into scheduling viewings, triaging maintenance with vendors, and automating routine back-office workflows — all while maintaining cost controls, usage observability, and audit compliance.
"We started this to save money. We ended up with AI governance."