Legalweek 2026 Pre-Show Rundown: Relativity's Brand Refresh, Epiq's Agentic AI Push, and What Legal Teams Should Do Now
Legalweek 2026 is kicking off in New York, and the early signals are clear: product launches, fresh partnerships, and a sharper focus on AI accountability. Two headlines stand out-Relativity's brand refresh and Epiq's expansion of its agentic AI suite.
If you're responsible for litigation, investigations, privacy, or firm operations, this week is a live test of what's real, what's ready, and what will move the needle this quarter.
Why this week matters
Budgets are set, matters aren't slowing down, and boards are asking for measurable impact from AI. Vendors will promise speed and accuracy. Your job is to translate that into procurement terms: risk, control, and ROI-backed by proofs, not promises.
Relativity: Brand refresh-what to watch
A brand refresh usually signals focus. Expect clearer product positioning, an updated story around AI and trust, and hints at how the company sees cloud, on-prem, and hybrid deployments fitting together. Look for how this ties to review speed, quality control, and admin load.
Start with first principles. What changes for your team on day one? What stays the same for security, data residency, and pricing? If the visuals are new, the value should be practical and quantifiable.
- Ask for a simple map of the product lineup: what's core, what's optional, and what's being retired.
- Confirm deployment parity: features and performance across cloud vs. on-prem, and documented migration paths.
- Pin down governance: audit logs, model transparency, encryption, and admin controls by role and matter.
- Pricing clarity: SKUs, consumption triggers (documents, users, compute), and caps to prevent overage surprises.
- Integration depth: APIs, event hooks, and supported workflows with your DMS, identity provider, and archival tools.
Epiq: Expanding an agentic AI suite-what to evaluate
"Agentic" AI points to step-wise, semi-autonomous workflows-think matter intake triage, privilege QC suggestions, ticketing handoffs, and compliance checks. Useful, if the guardrails are real and the cost is predictable.
Your north star: control. You want auditable actions, human-in-the-loop checkpoints, and clear error handling. Also, a way to see and cap spend before a pilot becomes a budget leak.
- Workflow design: which steps are automated vs. assisted, and where approvals are enforced.
- Evidence trail: full activity logs, inputs/outputs retained by matter, and reproducibility for defensibility.
- Model posture: vendor-managed vs. bring-your-own model, prompt/response retention, and PII handling.
- Data controls: residency options, data minimization, and opt-outs for training on client data.
- Cost controls: rate cards by action (classification, summarization, retrieval), throttling, and budget alerts.
Partnerships and integrations worth your time
Expect a wave of "now integrates with" announcements across CLM, DMS, and comms platforms. Treat every connector claim as a hypothesis. Ask for latency numbers, permission mapping, and exactly what fields or objects sync both ways.
- Security: SCIM/SSO maturity, least-privilege access, and redaction at export.
- Data scope: which objects, which metadata, and whether hidden or private channels are exposed.
- Ops reality: who maintains the connector, update cadence, and documented break-fix SLAs.
Your Legalweek procurement playbook
- Define a 30-day pilot: 1-2 workflows, clear owners, and a sandbox with production-like data.
- Pick hard metrics: hours saved per custodian, first-pass precision/recall, cycle time to review complete, cost per matter.
- Security baseline: SOC 2 Type II, ISO 27001, data residency choices, retention defaults, and incident response timing.
- Controls: role-based approvals, export restrictions, watermarking, and immutable audit logs.
- Pricing hygiene: document, user, and compute thresholds; unit prices; discount triggers; and termination terms.
- References: 2 customers with similar scale and stack, plus 1 who didn't renew-ask why.
Governance and risk questions to ask every AI vendor
- Quality management: evaluation datasets, error rates by task, and how regressions are detected and rolled back.
- Model updates: change control, release notes with risk impact, and opt-out windows.
- Content controls: PII scrubbing, profanity/toxicity filters, and policy enforcement logs.
- Legal risk: IP ownership of outputs, indemnities, and third-party component disclosures.
For small and mid-size teams
Start narrow. Pilot privilege log drafting, first-pass triage, or matter intake summaries. Prove value with 1-2 measurable tasks before expanding to broader review or CLM automation.
If bandwidth is tight, compare managed services to in-house deployment. The right choice is the one you can secure, measure, and maintain with your current team.
Keep learning
For deeper briefings and practical playbooks, see AI for Legal and AI Agents & Automation.
Bottom line: enjoy the show floor, but make vendors work. Ask for receipts-benchmarks, logs, and pricing transparency. Then buy what you can govern and prove.
Your membership also unlocks:
