ONGOING SUPPORT & MAINTENANCE
A self-hosted AI system isn't a set-and-forget deployment. New documents need indexing. Retrieval quality drifts as your corpus grows. Better open-source models release every few months. We handle all of this — monthly retrieval audits, document updates, model upgrades, and troubleshooting — so your system keeps delivering accurate results as your Melbourne organisation evolves.
The firms that get the most value from in-house AI are the ones that treat it like infrastructure, not a project. Whether you're running an air-gapped system or a network-connected on-premises deployment, we maintain it the way a good IT partner maintains your network — proactively, on a schedule, and with a focus on what's actually working for your team.
WHAT'S INCLUDED
SYSTEM MONITORING
We track response times, GPU utilisation, error rates, and storage capacity. If something degrades, we catch it before your staff notice — not after.
RETRIEVAL OPTIMISATION
Monthly audits of search accuracy. We test real queries, check whether the system returns the right documents, and tune the embedding model, chunking strategy, and retrieval thresholds to keep results sharp.
KNOWLEDGE BASE UPDATES
New policies, matter files, or records? We ingest and index them. Old documents superseded? We remove or archive them. The system always reflects your current document set — not last quarter's.
MODEL UPGRADES
Open-source models improve fast. Every quarter we evaluate whether a newer release (Llama, Mistral, Gemma) would give better results for your workload. If it does, we test and migrate — zero downtime.
WHAT GOES WRONG WITHOUT ONGOING SUPPORT
STALE KNOWLEDGE BASE
Your firm updates policies quarterly, but the AI still references last year's version. Staff get outdated answers and lose trust in the system. Within six months, adoption drops because people don't believe the results.
RETRIEVAL DRIFT
As your document corpus grows, the retrieval pipeline that worked well for 5,000 documents may struggle with 20,000. Chunking strategies that were optimal for policies don't work as well for correspondence. Without tuning, answer quality slowly degrades — and nobody notices until someone gets a bad answer in a meeting.
MISSED MODEL IMPROVEMENTS
Open-source models are improving rapidly. A system deployed with last year's model is leaving performance on the table. Newer models are faster, more accurate, and better at following instructions — but they don't install themselves.
SUPPORT SCHEDULE
| Frequency | What We Do |
|---|---|
| Monthly | Retrieval quality audit, document corpus refresh, system performance review, access log review |
| Quarterly | Model evaluation against new open-source releases, embedding model assessment, infrastructure capacity review |
| As Needed | Troubleshooting, access control changes, new document type ingestion, urgent re-indexing, user training |
FREQUENTLY ASKED QUESTIONS
What does ongoing support include?
Monthly retrieval quality audits where we check that the system is returning accurate, well-cited answers. Document corpus refreshes when new policies, matter files, or records need indexing. Quarterly model evaluations against the latest open-source releases — if a better model is available, we test and migrate. Troubleshooting, access control changes, and new document type ingestion as needed. Think of it as managed IT for your AI system.
How often are systems updated?
Monthly at minimum. Every month we refresh the knowledge base with new documents, audit retrieval quality, and review system performance. Quarterly, we evaluate whether a newer open-source model would improve results — the landscape moves fast, and a model that was best six months ago may not be best today. We also respond immediately to any system issues, access control requests, or urgent document ingestion needs between scheduled reviews.
Can support be scaled as our usage grows?
Yes — and it usually does. A firm that starts with one practice group often expands to three within six months. We adjust the system configuration, add hardware capacity if needed, configure new access control rules, and ingest additional document sets. The support level scales with the deployment, so you're not locked into a fixed scope that doesn't match how you're actually using the system.
Want to See How This Works for Your Firm?
We'll walk you through a deployment that fits your setup — your documents, your infrastructure, your compliance requirements. No sales pitch.
Request a Consultation →Or email us directly at hello@airgapllm.com.au