Key facts
| Access control | Scoped API keys with rotation; enterprise SSO and RBAC options |
| Engagement separation | Separate keys and retrieval indexes per client engagement |
| Deployment | Cloud, VPC, on-prem or air-gapped to match client contracts |
| Auditability | Request, model and response logging for engagement review trails |
| Data grounding | Embeddings and RAG are live for methodologies and deliverables |
| Models | 30+ models behind one OpenAI-compatible API |
| Pricing model | Flat monthly self-serve plans; no per-token billing on self-serve |
| Endpoint roadmap | Files, batch and fine-tuning are coming soon |
TL;DR
- Classify data per engagement, not per practice area.
- Give each engagement its own keys and retrieval index with an end date.
- Document cross-border processing before staffing work across regions.
- Log requests, sources and reviewers so deliverables can be defended.
- Keep a named engagement owner accountable for AI-assisted output.
How it works, step by step
- Inventory AI use cases across research, analysis, drafting and proposals.
- Classify content by client confidentiality, engagement stage and jurisdiction.
- Choose deployment per engagement: cloud, VPC, on-prem or air-gapped.
- Issue per-engagement keys with expiry dates and central inventory.
- Define log fields and retention: request ID, model, sources, output, engagement owner.
- Approve a model allow-list and ground answers in current methodologies.
- Document subcontractor and cross-border data flows before they start.
Try it yourself
Open the AI API key security checklist →
Engagement confidentiality and classification
Professional services content divides into public methodologies, firm know-how, client working papers and client deliverables. Each class needs a handling rule, and client contracts often add restrictions that are stricter than the firm's default, especially for financial, health or government clients.
Start with methodologies and firm know-how, then extend to client content only where the engagement permits it and the deployment keeps prompts, documents and embeddings inside the agreed boundary.
Keys, teams and least privilege
Issue a distinct API key per engagement, application and environment, with an expiry date that matches the engagement timeline. Store keys in a secrets manager, rotate on a schedule, and revoke when the engagement closes or staff rotate off. Enterprise SSO and RBAC options keep console access aligned with staffing changes, and subcontractor access should be scoped and time-bound.
Never share a key across engagements, and never place client identifiers in prompts where retrieval can supply only the fields a task requires.
Cross-border residency and retention
Decide where processing happens for each engagement and record it. Region selection covers many residency needs; VPC, on-prem and air-gapped deployment covers clients who require data to stay in a jurisdiction or inside their own environment. Retention applies to prompts, outputs, logs and retrieval indexes, and engagement records follow contractual and professional schedules.
Log enough to defend a deliverable: request ID, model and version, retrieved source identifiers, output and the engagement owner. See AI audit logs for a schema.
Model governance and account-level judgment
Keep an approved model list with evaluation evidence and re-test when versions change. Ground analysis in current methodologies and cited client materials so reviewers can verify quickly. Professional judgment stays with the engagement team, and the checklist should name who reviews AI-assisted work before it reaches the client, including how subcontractors and cross-border team members are covered.
Honest comparison
| Control area | Plugsky capability | Common gap | Owner |
|---|---|---|---|
| Engagement separation | Separate keys and indexes per engagement | One workspace for all clients | Risk and quality |
| Identity | Scoped keys with expiry, SSO and RBAC options | Keys outliving engagements | IT security |
| Data boundary | Cloud, VPC, on-prem or air-gapped deployment | Cross-border flows undocumented | Client contracting |
| Retention | Configurable logging under contract terms | No defined retention | Risk and records |
| Audit trail | Request, model and source logging | Deliverables not traceable | Quality review |
| Review | Citations and structured output for teams | AI drafts sent to clients unreviewed | Engagement partner |
Frequently asked questions
Does using Plugsky make us compliant?
No. Compliance is your program. Plugsky provides deployable controls - scoped keys, engagement separation, private deployment, logging - that you document and audit against client and professional obligations.
How do we keep engagements separate?
Use a distinct key and retrieval index per engagement with an expiry date, and ensure no shared workflow can query across clients.
What should we log?
Request IDs, model names and versions, retrieved sources, outputs and the engagement owner, retained under contract terms so deliverables can be defended.
Can we run AI on client sites?
Yes. VPC, on-prem and air-gapped deployment keeps prompts, documents and embeddings inside the client environment, and the API stays OpenAI-compatible.
How do we handle subcontractors?
Issue separate scoped keys per subcontractor with expiry, keep their retrieval access limited to the engagement, and revoke on completion.
Is fine-tuning available for our methodology?
Fine-tuning, files and batch endpoints are coming soon. Today, use retrieval over approved methodologies with JSON mode for consistent output structure.
Where should a pilot start?
Pilot on public methodologies and internal know-how with the free plan, prove citations and logging, then extend to engagement content on private deployment.