Skip to content
Engagements

Three ways to bring Glasshouse into your decision.

Advisory when you want us in the room. Cloud when your team wants to run it themselves. On-premise when the data cannot leave your network.

Engagement modes

01 / Advisory

We run the rehearsal. We sit in the room.

For boards, executive committees, and crisis teams who want footnoted evidence in front of them by Tuesday, without operating the system themselves.

What's included
  • Scenario design with your decision team
  • Up to four rehearsal runs per engagement
  • Findings briefing: markdown report, PDF archive, board-ready deck
  • Written guidance on which framing to take and why
  • Glasshouse reports cited as evidence in your record
What's not
  • System access for your team
  • Ongoing seat licenses
To start
  • One conversation. NDA. Decision brief.
  • Lead time: 5–10 working days from kickoff to first findings.
€500K–€5M per engagement, scoped on decision criticality and rehearsal count.
Start an advisory engagement
02 / Cloud

Your team. Our infrastructure.

For comms, strategy, and policy teams who want to rehearse on their own cadence, against their own dossiers, without standing up infrastructure.

What's included
  • Full Glasshouse access for your seats
  • EU-region inference (Vertex AI, Frankfurt)
  • Audit log per run, retained for the contract term
  • Onboarding workshop for the first cohort
  • Quarterly office hours with the engineering team
What's not
  • Multi-tenant data sharing, every tenant is isolated
  • Dossier retention beyond the contract: your data, your call
To start
  • Discovery call. DPA. Tenant provisioning.
  • Lead time: 2–3 weeks from contract signature.
From €120K / year, by seat count and rehearsal volume.
Talk about Cloud
03 / On-premise

Your network. Your hardware. Your auditors.

For regulated decision-makers: finance, insurance, legal, government, where the dossier cannot leave the perimeter and every line of the engine has to be inspectable by your security team.

What's included
  • Self-hosted deployment, air-gap-capable
  • Full source access to the engine for your security review
  • Local-only inference: llama.cpp on your hardware, no outbound calls
  • Quarterly engineering touchpoint and patch cadence
  • Optional on-site implementation sprint
What's not
  • Hosted compute: you provide the workstations or cluster
  • Telemetry: none, by design
To start
  • Architecture review. Security questionnaire. Hardware sizing.
  • Lead time: 6–10 weeks from contract signature, depending on review depth.
Annual license, scoped on cluster size, audit support, and patch SLA.
Talk about On-premise
At a glance

How they differ.

Advisory Cloud On-premise
Who runs the rehearsal Creative Human Your team Your team
Where data lives Our EU tenant for the engagement Isolated EU tenant Inside your network
Source-code access — — full
Air-gap capable — — yes
Lead time 5–10 days 2–3 weeks 6–10 weeks
Commercial shape Per engagement Annual subscription Annual license

Not sure which mode fits? Tell us the decision; we'll tell you the shape.

Start a conversation