Sandstone Logo

Stop Being the Bottleneck: AI Intake and Triage for In‑House Legal That Actually Scales

Legal teams spend up to 30–40% of their time chasing context and routing requests. Here’s how AI-powered intake and triage—grounded in your playbooks and positions—unclogs workflows, improves SLAs,…

Jarryd Strydom

September 29, 2025
You don’t need another survey to know it’s true: a surprising 60%+ of legal requests hit your team incomplete—missing context, documents, or the right business owner. Across mid-market and enterprise teams we speak with, 30–40% of legal time disappears into intake triage and status chasing. That’s not legal work; that’s operations drag.
The fix isn’t another webform. It’s an AI-powered workflow that turns your playbooks into a living triage layer—so every request arrives complete, routed, and decision-ready.

The Hidden Tax of Unstructured Intake

Email and Slack are great for speed but brutal for structure. Requests arrive with half the facts, legal must follow up, and cycle time balloons. Meanwhile, leadership judges legal on responsiveness, not on the complexity you’re untangling behind the scenes.
Common signals you’re paying the intake tax:
  • First-response SLAs slip during quarter-end crunches.
  • NDAs and low-risk vendor reviews pile up next to revenue-critical deals.
  • Lawyers rewrite the same guidance weekly because it lives in heads, not systems.
  • Business users don’t know where to go, so they try everyone.
Without a knowledge layer, every intake is a snowflake. With one, it’s a pattern you can automate.

Design AI Triage That Mirrors Your Policy

AI triage isn’t magic—it’s policy, structured and enforced by an agent. Start with one high-volume workflow (e.g., NDAs, vendor onboarding, or marketing review) and capture your “source of truth” in three layers:
  1. Required facts: What must the business provide to start work? Example for NDAs: counterparty name, jurisdiction, term, mutual/unilateral, use case.
  2. Positions and thresholds: What’s pre-approved vs. what needs escalation? Example: mutual NDA, 2-year term, Delaware law = auto-approve; unilateral or 5-year term = route to queue B.
  3. Actions and outputs: What should happen next? Generate a redline from your standard, open an approval task, or respond with self-serve guidance.
With these layers, an AI agent on a platform like Sandstone can:
  • Collect missing facts in-channel (Slack, email, or portal) with smart follow-ups.
  • Classify requests by risk and route using your playbooks—not gut feel.
  • Generate a first draft or redline from your clause library.
  • Surface exceptions and tag approvals to the right owner.
  • Log every question-and-answer back into your knowledge layer so guidance compounds.

What “Good” Looks Like: Metrics That Matter

If you can’t measure it, you can’t scale it. Define KPIs upfront and share them with stakeholders so value is visible:
  • Intake completeness rate: % of requests that arrive with all required fields.
  • Auto-resolve rate: % resolved without attorney touch (e.g., standard NDA).
  • First-response SLA: median time to acknowledgment by request type.
  • Cycle time: request-to-close, segmented by risk tier.
  • Rework rate: % of matters reopened due to missing info or misrouting.
  • Escalation mix: how many requests require senior review—and why.
Teams running AI intake on Sandstone often see 2–3x faster acknowledgments, 30–50% auto-resolution on low-risk items, and cleaner queues for complex work. The win isn’t just speed; it’s confidence in where legal effort goes.

A Concrete Starting Point: Vendor Onboarding

Vendor reviews are ripe for automation because they cross legal, privacy, security, and procurement. Here’s how an AI agent can orchestrate the flow:
  • Intake: Collect data category, data subject type, hosting region, subprocessor use, contract value, and timeline.
  • Policy check: Map responses to your data processing thresholds and security guidelines.
  • Routing: Auto-route low-risk vendors to standard DPAs and privacy review; flag high-risk to counsel with a prebuilt risk memo.
  • Drafting: Generate a DPA or redline from your approved positions; attach a tailored DPIA questionnaire if needed.
  • Decision record: Capture approvals, exceptions, and rationales so the next, similar request moves faster.
This is Sandstone’s sweet spot: layered data, modular workflows, and decisions that build on each other.

Actionable Next Step: A 2‑Week Pilot Plan

You can stand up a credible pilot in 10 business days:
  • Days 1–2: Pick one workflow (NDAs or vendor onboarding). Document 8–12 required fields, 5–7 standard positions, and 3 escalation criteria.
  • Days 3–5: Configure intake (Slack form or portal), map routing rules, and connect your template library.
  • Days 6–8: Train the agent with examples of “good” and “edge” cases. Define SLAs and owners by risk tier.
  • Days 9–10: Launch to a friendly business group. Track completeness, auto-resolve, and response time. Hold a 30‑minute retro; tune prompts and thresholds.
If the metrics move, expand. If not, you’ll know exactly where the friction lives.

The Payoff: Trust, Speed, and a Stronger Foundation

When intake is AI‑driven and policy‑anchored, your team stops firefighting and starts operating. Business users get instant clarity. Lawyers focus on the work that actually needs judgment. Knowledge compounds rather than evaporates.
That’s the promise of a modern legal ops platform and knowledge layer like Sandstone: not another tool to manage, but the bedrock where business and law move in harmony. Start with intake. Make it actionable. Let every request strengthen your foundation—and watch speed, alignment, and trust follow.

About Jarryd Strydom

Jarryd Strydom is a contributor to the Sandstone blog.