Sandstone Logo

Stop Treating Legal Intake as a Help Desk: Make It a Knowledge Flywheel

Most legal delays aren’t from redlines—they start at intake. Here’s how AI agents and layered workflows turn every request into reusable knowledge and faster decisions.

Jarryd Strydom

September 1, 2025
Half of your legal work starts life as a Slack or email with missing context. That’s where velocity disappears—not in redlines, but in the back‑and‑forth to gather basics. The fix isn’t another form; it’s a system that turns every request into structured, reusable knowledge.
The Real Bottleneck Is Intake, Not Review
When intake is unstructured, your team becomes a switchboard. Legal spends cycles asking who the counterparty is, which template to use, whether procurement has blessed the vendor, and what the commercial terms are.
That friction compounds:
  • Work hides in channels you can’t see or measure.
  • Approvals stall because routing is unclear.
  • Institutional knowledge lives in people’s heads—or long email threads.
Modern legal ops treats intake as a data problem. Capture the right fields once, route based on policy, and let decisions leave a trail you can audit. Do this well, and review time drops because you’re starting from clarity, not chaos.

Turn Intake Into a Knowledge Flywheel

A knowledge flywheel means every request makes the next one easier. Playbooks, positions, and approvals don’t just sit in a wiki—they power decisions.
Here’s the model:
  • Structured capture: Requestors answer targeted questions in Slack/email/web forms. Required fields dynamically adapt to the workflow.
  • Policy‑based routing: Rules route to sales ops, procurement, privacy, or legal based on risk and dollar thresholds.
  • Reusable positions: Pre‑approved fallbacks auto‑apply (or suggest) where the fact pattern matches prior decisions.
  • Audit and learning: Each decision updates the operating record, strengthening guidance and reducing re‑work.
Platforms like Sandstone operationalize this with AI agents that sit where work already happens. They triage, gather context, suggest the right template, and surface your positions—so knowledge compounds instead of disappearing.

A Workflow You Can Automate Now: Sales NDAs and Vendor MSAs

Start where volume is high and risk is known. Two great candidates: NDAs (non‑disclosure agreements) for sales and vendor MSAs (master services agreements) in procurement.

How an AI agent runs the play:
  1. Intake in the flow: The requestor triggers “New NDA/MSA” in Slack or via a link in Salesforce/Procurement. The agent collects counterparty name, jurisdiction, data types, deal value, and timeline.
  2. Template decision: Based on responses, the agent selects the correct template (mutual vs. one‑way NDA; standard vs. data‑processing addendum), autofills party info, and proposes standard terms.
  3. Policy checks and routing: If data includes personal information, privacy review is auto‑added. If annual spend exceeds threshold, procurement and finance are routed. Otherwise, it’s auto‑approved or sent to legal with a clean packet.
  4. Position suggestions: When the counterparty pushes back, the agent suggests pre‑approved fallback clauses with rationale and links to your playbook.
  5. Record and handoff: Every step—intake fields, decisions, fallbacks used, approver and timestamps—writes back to your knowledge layer for search, reporting, and reuse.
The result: requestors get self‑serve where possible, approvers see only what requires judgment, and legal’s guidance shows up at the exact moment of need.
Metrics That Matter (So You Can Prove It Works)
Measure the flywheel, not just the finish line:
  • Cycle time: Intake‑to‑signature for NDAs/MSAs. Segment by self‑serve vs. legal‑touched.
  • First‑time quality: Percentage of requests submitted with all required fields.
  • Approval SLAs (service level agreements): Time in each queue—privacy, procurement, finance, legal.
  • Deflection/self‑serve rate: Share of requests completed without attorney time.
  • Reuse rate: How often pre‑approved positions resolve an issue without escalation.
Baseline for two to four weeks, then pilot. You’ll see where the bottleneck lives and which rules move the needle first.

Common Pitfalls and How to Avoid Them

  • Boiling the ocean: Don’t automate everything at once. Pick one workflow, one region, one business unit.
  • Rigid forms: If every user sees every field, adoption drops. Use conditional questions.
  • Shadow channels: Keep Slack/email entry points, but route them into the same structured workflow.
  • Unlabeled knowledge: Playbooks without metadata can’t power decisions. Tag by clause, risk, and use case.
  • No feedback loop: Review exceptions monthly. Promote frequent fallbacks into standard positions to reduce friction.
  • Security/privacy gaps: Ensure data handling, permissions, and audit logs meet your standards—especially for AI‑assisted suggestions.

Actionable Next Step: Run a 30‑Day Pilot

  • Choose one workflow (Sales NDA or Vendor MSA) with clear templates and known thresholds.
  • Define required intake fields and approval logic with business owners.
  • Embed the intake trigger in Slack and the system of record (Salesforce/Procurement).
  • Turn your playbook into structured fallbacks with rationale and risk notes.
  • Set targets for cycle time, first‑time quality, and self‑serve rate; review weekly.
If you can’t measure it, you can’t scale it. Prove value on one workflow, then layer on the next.
Legal shouldn’t be a help desk—it should be the connective tissue that speeds deals with clarity and confidence. By turning intake, triage, and decisions into a living operating system, you build strength through layers: structured data, modular workflows, and positions that improve with every request. That’s how platforms like Sandstone help lean teams move from reactive support to a proactive foundation for trust and growth.

About Jarryd Strydom

Jarryd Strydom is a contributor to the Sandstone blog.