Production Set 12 // In-House Guide
eDiscovery for in-house legal operations, build vs buy, hold, procurement.
VERIFIED 21 APR 2026 // INDEPENDENT REFERENCE // NOT LEGAL ADVICE
Most eDiscovery content targets outside counsel. The in-house audience, legal operations directors, general counsel, and litigation support teams embedded in F1000 corporations, faces a different set of decisions. This page addresses those decisions directly: build versus buy, outside-counsel coordination, legal hold, and the procurement questions that matter most in the AI era.
Section 01 // Decision Matrix
Build vs buy: the central in-house decision
The build-versus-buy decision is the most consequential one for an in-house legal ops team evaluating eDiscovery. Building in-house platform capability means subscribing to and operating a platform (Relativity, Everlaw, DISCO, or a competitor) directly, staffing a litigation support team to run it, and coordinating data-in-place access for outside counsel. The benefit: cost control, institutional knowledge retention, and consistent data handling. The cost: platform subscription, staffing, and operational overhead.
Buying means forwarding matters to outside counsel with platform access through their lit-support vendor (HaystackID, Consilio, Lighthouse, Epiq), or retaining a managed-service provider to run review on an in-house-selected platform. The benefit: lower fixed cost, access to expert lit-support project management. The cost: lower visibility, higher per-matter cost, and dependency on outside counsel platform choice.
| Scenario | Recommended Model |
|---|---|
| F1000 with 20+ matters/year, recurring custodian population | In-house platform (Relativity or Everlaw) |
| F1000 with 5-20 matters/year, variable data | Hybrid: in-house hold, outside counsel review |
| Mid-market (2,000-10,000 employees), occasional litigation | Managed service or per-matter (Logikcull) |
| High-growth tech (frequent regulatory, IP, employment) | Everlaw (integrated, fast, no intermediary) |
Last verified Apr 2026
Section 02 // Legal Hold
Preservation and legal hold: the in-house-only obligation
Legal hold is an in-house obligation that outside counsel cannot fully discharge. The moment the company anticipates litigation or a regulatory inquiry, the duty to preserve relevant ESI attaches under FRCP 37(e). This means identifying relevant custodians, issuing hold notices, and confirming that data preservation controls are in place, all before outside counsel is involved in reviewing documents.
Modern in-house hold tools (Zapproved, Exterro Fusion, Mitratech) automate hold notice issuance, custodian acknowledgment tracking, and custodian data preservation confirmation. These tools integrate with O365, Google Workspace, Slack, and other enterprise data sources to suspend auto-deletion policies on hold data. Confirm that your hold tool's custodian roster is current and that your O365 in-place hold or equivalent is activated before any document review begins.
Section 03 // Six Questions
The six procurement questions for the AI era
When evaluating any eDiscovery platform with AI capabilities, ask these six questions before signing. The first three are about what the AI actually does; the last three are about what happens to your client's data.
- Q1
Is the relevance scorer a classical classifier or an LLM?
Ask for specifics. 'AI review' can mean a 2015-era SVM classifier or a 2025 foundation model. The answer affects accuracy, validation requirements, and the explanation you can provide to opposing counsel.
- Q2
Which LLM, and is it general-purpose or legal-tuned?
OpenAI GPT-4 architecture (Relativity), Everlaw-proprietary, or Nuix-tuned are different risk profiles. General-purpose models may be less accurate on legal-specific privilege signals. Legal-tuned models may be less available for benchmarking.
- Q3
Can you export per-document reasoning traces for attorney QA?
If the AI makes or influences relevance and privilege determinations, can you show the court what the AI 'thought'? Partial trace export (Relativity partner API) is different from no trace export (most platforms).
- Q4
Is data processed in a tenant-isolated environment?
Tenant isolation means your client documents and prompts do not cross into another client's compute environment. Required under ABA 512 Rule 1.6 confidentiality and GDPR Article 28. Ask for the architecture diagram.
- Q5
Is there a contractual zero-retention commitment on prompts and outputs?
Zero-retention means the vendor does not retain documents, prompts, or AI outputs beyond the immediate API transaction. Essential under ABA 512. Ask for the specific contract language, not the privacy policy summary.
- Q6
What are the SOC 2 Type II, GDPR Article 28, and HIPAA BAA status?
For matters involving EU data or health records, these certifications are prerequisite. Ask for the current SOC 2 Type II report date and the processor agreement (Article 28) before sending any EU personal data to the platform.
Section 04 // Coordination
Outside-counsel coordination
In-house legal departments that own their platform can give outside counsel review access without surrendering data custody. The two primary architectures are data-in-place (in-house retains the data and outside counsel reviews through the in-house platform, typically with an attorney-access role rather than a full administrator role) and data-at-outside-counsel (the in-house team sends collected data to outside counsel who loads it into their platform and runs the review).
Data-in-place architectures are preferred for in-house teams that want to maintain chain of custody and avoid the cost of re-collecting and re-processing data for each outside counsel engagement. They require platform-level outside-counsel access permissions and clear protocols for privilege review, coding decisions, and production sign-off.
Section 05 // FAQ