Unless you're MacGyver, your data is probably stuck in your AI tool forever
Everything your company pours into its AI — chats, files, memory, skills, custom assistants — mostly stays with the vendor. On some business plans there is no export button, and retention clocks are destroying history as you read.
Discover the Data Portability AgentWhich AI tool does your company run on?
The differences between plans are bigger than the differences between providers — and “we’re on the business plan” often means less exportable than personal, not more.
How much of your data do you actually control?
Sounds absurd? It is. On Claude Team, the paid business plan gives you less export than a free personal account. Counterintuitive — but true.
Percentages are indicative weightings of the asset classes, not vendor numbers — the agent replaces them with a measured coverage report per client
Underlying export and API facts audited against vendor documentation, re-verified 28 August 2026
How the agent works in your case
Your data lives in three places, so the agent runs one loop per place. Loop 1 empties the provider’s server. Loop 2 empties the model’s head — the memory and routines no export contains. Loop 3 empties the laptops.
Everything lands in storage you own. The data store is the archive: plain, readable files holding everything the loops capture. Next to it sits your Git repo, the source of truth for skills, prompts and agent configs, which deploy from Git into the AI.
What each loop concretely does on the plan you picked:
| Loop | Route on this plan | Cadence | What you get |
|---|
Moving to another tool
Your current setup carries over from step 1. Pick where you’d go:
| Asset | Leaves the old tool as… | Lands in the new one as… |
|---|
What Eyetea actually builds
Everything above is what you get. Here is the thing that does it.
The agent lives on your box — nothing on ours
The agent runs inside your own perimeter — a box on-premise or a machine in your own cloud — as a stateless pump: playbooks and schedules, never a copy of your data. What it collects lands in your storage and Git, and seeds the next tool when you switch.
That settles GDPR the clean way: agent and data stay inside your perimeter, and Eyetea ships playbook updates, never content.
An Eyetea-managed EU cloud instance stays available as a fallback home for the same container
On that variant Eyetea acts as a GDPR data processor: standard DPA, per-client isolation, content-free logs
Runtime credentials live under the strictest contract of the engagement either way
What it pays back before anyone switches
Firm-wide AI search
“Has anyone here worked on this?” becomes a query. On most tiers, nobody can search across seats today.
Leaver continuity
People resign long before companies switch vendors. It pays for itself at the first resignation.
Duty of knowledge
“What has our firm told AI about client X?” — answerable, for GDPR, audits and professional secrecy.
Licence truth
Used seats, recurring prompts, dead weight — the licence audit with live data instead of guesses.