An AI studio for regulated operations.
We put AI to work in insurance, legal, biotech, and health — the places where "mostly right" isn't good enough.
We sell two things: training that puts a working AI assistant on every desk, and governed automations we build into your operation, one job at a time. A named person approves everything that leaves.
- ~60 cases a day through a system we run in production
- 15 min → 2 min per scanned claims file, measured
- £178k of recoverable leakage found in one review
- 73 hrs/wk of automatable work mapped in one inbox
| 01 The model | You own this | Frontier models where your obligations allow, open-weight leaders served in your region, or fully on-premises when nothing may leave. Swapping the model is a configuration change, not a rebuild. |
| 02 Data | Delegate | Everyone can license the same models; only you have your email, your shared drives, your CRM, your APIs. We connect the model to where your work already lives, read the scans and photographs that never arrive clean, and write results back through your platform's own API. |
| 03 Workflows | Delegate | The machine runs the steps; a named person approves before anything leaves. Every job works probation like a new starter, until measured accuracy on your own samples earns autonomy. |
| 04 Skills | Enable | We map each role against what it actually demands, install the skills worth having, and teach the assistant your voice. |
| 05 Memory | Enable | What you decided, who you work with, what happened last time. Private by default; what the organization needs graduates into shared memory by choice. |
Enforcement wraps all five — human gate · scoped access T0–T4 · audit log · your choice of region.
| Train your people | An assistant on every desk. Starts in weeks, touches no client data. |
| Automate one job | Through the Readiness Audit: fixed fee, half credited into the first build. Mapped, priced in hours and money, on probation. |
Jobs chain into routines — each new one reuses the rails the last one built. Start with one assistant. End with a department.
- The map of AI jobs — an interactive map of the work AI can take across an organization, every job graded by access tier and build depth.
- Simplified AI Framework — the framework behind our team training.
- Academy — hands-on tutorials for non-developers who want to ship production-grade apps.
- Blog — field notes on building, shipping, and funding AI.
XY Space is Cho Yin Yong, a University of Toronto lecturer with two AI patents and nine years of regulated engineering leadership, and Mahmoud Halat, employee two on the platform behind roughly a quarter of Canada's daily COVID vaccinations and winner of the $100k Lovable Shipped 2025 grand prize against 5,800 builders. A decade of complex automations in the most regulated industry there is. Both of us stay on the work after the sale.
Website · Case studies · Training · Contact · hello@xyspace.dev
