A practice built around one rule: move data as few times as possible.
A data practice, not an agency
xerocopy helps teams migrate, inspect, and prepare data without multiplying untrusted copies. Fewer hops, in-place reads, contracts at the boundary, and AI that works from data someone can defend. A small book of clients, work done slowly and in writing.
What we do
Four services
01 — Data migration & integration
Migrate, consolidate, or connect systems across on-prem and cloud without leaving a permanent trail of staging tables and duplicate truths.
Good fit when a platform is changing, integrations have multiplied, or the cutover needs a defensible path.
Explore migration & integration →02 — Data audits
Follow a critical figure from source to decision, inventory every copy, and price the cost, risk, and disagreement attached to each hop.
Good fit when meetings start with reconciliation or nobody can name the authoritative source.
Explore data audits →03 — AI audits
Evaluate source quality, retrieval, token waste, permissions, tool access, verification, and human oversight as one connected system.
Good fit when an AI pilot works in a demo but its sources, costs, or authority are hard to defend.
Explore AI audits →04 — AI integration
Build retrieval, copilots, automations, and agents on authoritative data with explicit contracts, bounded capabilities, and evidence.
Good fit when a useful workflow has a clear owner and needs to move from prototype to production.
Explore AI integration →On-prem
A container or cron job on hardware the client already owns. No subscription tax.
Hybrid
Warehouse plus SaaS sources. Each workload sits where latency and ownership say it should.
Cloud
Terabytes to petabytes on Kubernetes — scheduled, autoscaled, observed, and costed.
The hidden invoice
What copies cost
In most of the stacks we audit, the same customer row lives in four to seven places: an extract, a staging schema, the warehouse, the marts, a semantic layer, a BI cache, and the spreadsheet somebody exported last Tuesday. Each copy was justified once. Each one is still billing.
Storage — the small bill
Duplicated tables, their snapshots, their backups, and their time-travel windows multiply a storage bill several times over. Storage is usually the smallest of the costs — the canary, not the coal mine.
Compute & egress — the growing bill
Every copy is re-extracted, re-loaded, and re-computed on a schedule, whether or not anyone reads it. Replication jobs, warehouse warehouses-of-warehouses, and cross-cloud egress fees routinely double or triple the compute line — and nobody can say which dashboard each run is actually feeding.
Reconciliation — the invisible bill
When the same number lives in five places, it disagrees with itself in five ways. Analysts spend the first half hour of every review arguing about whose figure is current; engineers maintain reconciliation jobs to paper over the gaps. That's headcount spent litigating copies instead of making decisions.
Risk — the bill you can't see
A stale copy is a decision made on old data. An orphaned copy is a compliance surface nobody owns — customer data outliving its deletion request in a mart no one remembers creating. Every copy widens the blast radius of the next mistake.
Deleting a hop doesn't just simplify the diagram. It deletes the storage, the compute, the meeting, and the risk that came with it.
The thesis
“Every extra copy is a place for drift, cost, and quiet disagreement. Our job is to delete hops.”Read the five claims in full →
Tools
Boring on purpose
Cloud & DevOps
Data infra
Start with what you need to know or what you need to change. Each path leads to a small first engagement with a concrete answer.
I need to understand the problem
I need to build or move something