Run a 4 week governed AI sprint with Cal
Make AI useful in security and risk – safely. In one month, prove value on your data, inside your controls.
How the Sprint works
Week 1 — Define
Pick one high-impact use case (incident triage, audit pack, or threat brief). Set scope and data boundaries.
Week 2 — Connect
Cal links to your authorised sources. Nothing leaves your environment without approval.
Week 3 — Run
Your team queries Cal, compares to today’s workflow, and iterates. Every answer is grounded, cited, and logged.
Week 4 — Review
We measure outcomes, capture lessons, and agree next steps—assist → approve → (optionally) automate.
What you get
-
A working, governed Cal workspace on your authorised data
-
10–15 mapped queries and visualisations for your use case
-
One audit-ready output pack (citations, timestamps, logs)
-
A short playbook: how to use Cal safely in your control room
-
Executive read-out with recommended next steps
What we’ll measure
-
Time-to-brief (mins)
-
Noise reduction / alert precision
-
Report completeness & consistency
-
% answers with citations (0 “black box” outputs)
Why Cal vs generic LLMs
-
Governed: Role-based policies, source allow-lists, human approvals
-
Grounded: RAG over your pre-approved corpora; must-cite policy
-
Visual: Maps, overlays, and schematics – not just text
-
Auditable: Full prompt/action logs and diff view of human edits
Put simply: Cal turns information into actionable, explainable intelligence.
What we need from you
-
A named sponsor and a practitioner lead (1–2 hrs/week each)
-
Access to approved sources (e.g., policies, public datasets, vendor feeds)
-
One 30-minute check-in per week
-
Optional: a redacted example brief/report for benchmarking
Typical team time: ~6–8 hours across the month.
Frequently asked questions
Is our data safe?
Yes. Cal operates with strict role-based access, logs every action, and only reads sources you approve. You can keep everything inside your environment.
How much IT effort is needed?
Minimal. We start with low-lift sources (approved documents and public authorised feeds). No invasive integrations needed for the sprint.
What does a ‘governed’ answer mean?
Every answer cites sources, shows timestamps, and is stored in an audit log. No opaque outputs.
Will we see maps and visuals?
Yes. Cal presents findings on maps/schematics with overlays – great for control rooms and exec briefings.
What if we decide not to continue after the sprint?
That’s fine. You keep the outputs and recommendations. No lock-in.
How much does it cost?
We’ll scope based on your use case and data boundary in the discovery call.
