ServiceNow LCO, Harvey, and Legora: an honest look at the enterprise legal AI stack

chiaragiuli · an hour ago

The conversation every Legal Operations team is having right now

Harvey and Legora are everywhere. If you work in legal technology, you have seen the press releases, the funding rounds ($11B and $5.6B respectively as of May 2026), and probably fielded questions from your GC or CPO asking why you are not using them.

The short answer: for many enterprise in-house legal teams already running on ServiceNow, you already have something more powerful, and most teams have not fully realised it yet.

This article explains exactly what that is, why it matters, and how to think about it when Harvey or Legora comes up in a platform conversation.

What Harvey and Legora actually are

The honest starting point: Harvey and Legora cover one layer of Legal Operations — document intelligence. ServiceNow LCO covers the full stack. Whether that difference matters depends entirely on what your team actually needs.

Before comparing, it is worth being precise about what these tools do.

Both Harvey and Legora are legal AI point solutions. They are excellent at document-intensive tasks:

High-volume contract review and bulk clause extraction
M&A diligence across large document sets
AI-assisted drafting and redlining
Legal research (Harvey via LexisNexis; Legora via its Qura acquisition)

What they are not: enterprise Legal Operations platforms. Neither has contract lifecycle management, legal service delivery intake, obligation tracking, GRC integration, or any connection to the broader enterprise.

Critically: both run on the same frontier models you can access directly. Harvey's multi-model system routes tasks across Claude Sonnet/Opus 4, GPT-5, o3, and Gemini 2.5 Pro. Legora hot-swaps between GPT-5, Claude, and GPT-o1 via Azure OpenAI. Neither has a proprietary model that is unavailable elsewhere.

The real question: what happens after the contract?

Harvey and Legora answer the document question. A lawyer uploads a contract, gets clauses extracted, risk flags surfaced, and a redline suggested. That is genuinely useful.

But in enterprise legal operations, the document is the beginning of the work, not the end. Consider what happens next:

Who routes the contract for approval?
What SLA governs the review?
Which obligations need to be tracked over the contract lifecycle?
Which suppliers carry risk that needs to surface in GRC?
How does the GC see throughput, cycle time, and deviation rates across the whole portfolio?

None of those questions are answered by Harvey or Legora. They are document tools, not Legal Operations platforms — and for an in-house team, that gap is the entire job-

The heterogeneous agentic flow: ServiceNow's structural advantage

Here is where the conversation gets technically interesting, and where the gap widens significantly.

Harvey and Legora offer agentic workflows — but they are homogeneous. Every step in the flow is AI-driven. The vendor controls which model runs at each step. You cannot insert your own business logic, your own approval rules, or your own enterprise use cases as native flow nodes.

ServiceNow CM Pro with agentic workflows and BYOM (Bring Your Own Model) lets you build what I call a heterogeneous agentic flow: a single governed workflow where each step gets the right type of intelligence for that step.

Two types of steps work side by side:

AI steps: probabilistic, generative intelligence where you need reasoning:

Clause extraction using Claude Opus 4 (BYOM)
Obligation summary using GPT-5 (BYOM)
Risk narrative generation using a fine-tuned domain model

Governed logic steps: precise, auditable, rule-based execution where you need certainty:

Contract value threshold check ("if > €1M, route to CFO")
SLA enforcement and escalation
GRC control triggers
Supplier risk score lookups from your existing ServiceNow data integrated in S2P
Approval chain routing using your organisation's delegation rules

The critical insight: governed logic steps are not workarounds or integrations, they are first-class nodes in the same Flow Designer workflow as the AI steps. They run in sequence, in parallel, or conditionally, all within a single governed, auditable process.

A concrete example: M&A contract intake to obligation tracking

Here is what a heterogeneous agentic flow looks like in practice:

Intake (governed logic) — employee submits via ServiceNow portal; contract type, counterparty, and deal value captured; routing rules applied automatically
Clause extraction (AI — Claude Opus 4 via BYOM) — bulk extraction of key clauses, non-standard terms, and risk indicators across the full document set
Risk threshold check (governed logic) — business rule: if identified risk score exceeds threshold, flag for senior counsel; no AI involved, fully auditable
Obligation summary (AI — GPT-5 via BYOM) — generate plain-language obligation summary per contract for business stakeholders
CFO approval trigger (governed logic) — if contract value > €1M, initiate approval workflow; deterministic, traceable, compliant
Contract record update (governed logic) — obligations, renewal dates, and counterparty data written to contract record; links to supplier and GRC records created automatically
GRC control check (governed logic) — verify regulatory clause requirements; flag gaps to compliance team
GC dashboard update (governed logic) — Performance Analytics updated with cycle time, risk score, and SLA status

Harvey and Legora can assist with steps 2 and 4. ServiceNow runs all eight — with a single audit trail, a single vendor, and a single system of record.

BYOM: why model flexibility matters more than model quality

A common assumption is that Harvey or Legora have better AI than ServiceNow because they are "legal AI specialists." The data does not support this.

Harvey explicitly runs Claude, GPT-5, and Gemini — the same models available via BYOM on ServiceNow. Legora's own CEO describes their platform as a "Ferrari chassis that can hot-swap AI engines." Neither has a proprietary model that delivers meaningfully better results for general legal operations tasks.

What BYOM on ServiceNow actually gives you:

Right model for the right step: use a small, fast model for classification; a reasoning model for complex clause analysis; a fine-tuned model for obligation extraction
Your own fine-tuned models: if your organisation has trained a domain-specific model on your contract history, you can bring it directly into a Contract agentic flow
No vendor lock-in on AI: as frontier models improve (and they improve every few months), you swap the model at the BYOM layer — your workflow logic and governed steps are unaffected
Cost control: you are not paying Harvey or Legora's margin on top of the underlying model costs

What Harvey and Legora still do well

Intellectual honesty matters here. There are scenarios where Harvey or Legora remain compelling:

Very high-volume M&A diligence where Legora's tabular review at thousands of documents and Harvey's LexisNexis integration deliver speed and citation quality that ServiceNow BYOM does not yet replicate out of the box
Law firms (not in-house teams) where the primary buying criterion is document AI depth, not enterprise workflow integration
Legal research requiring jurisdictional case law with verified citations — Harvey's voyage-law-2 embedding model and Legora's Qura research database are purpose-built for this

For in-house enterprise legal teams already on ServiceNow, these are edge cases, not the daily workload.

The protocol layer: one more reason this matters in 2026

A2A (Agent-to-Agent) protocol reached v1.0 in April 2026. MCP (Model Context Protocol) has crossed 97 million monthly SDK downloads. The emerging enterprise architecture is an open, composable agent fabric where agents across platforms can discover, delegate to, and receive results from each other.

ServiceNow is built toward this architecture. Harvey and Legora have not announced A2A or MCP support. Their agents are self-contained: they can receive API calls but cannot participate in open multi-agent orchestration.

This means that as enterprise agent fabrics mature, ServiceNow legal agents will be able to interoperate with procurement agents, HR agents, finance agents, and GRC agents across the platform. Harvey and Legora agents will remain islands.

How to position this in a customer conversation

When Harvey or Legora comes up, the most effective reframe is not to dismiss them but to change the question:

"Harvey and Legora are excellent at making lawyers faster at documents. The question for your in-house team is different: how do you connect legal work to the enterprise? How do you enforce the right business rules at the right point in a workflow, with a full audit trail, connected to GRC and procurement and finance? That is what ServiceNow does — and no amount of document AI replaces that."

The one-sentence version for each:

vs Harvey: "Harvey answers the document question. ServiceNow answers the operations question — and for in-house teams, that is the question that matters."
vs Legora: "Legora is great for M&A diligence at law firms. You are an in-house team. You need intake, CLM, obligation tracking, and GRC integration. Legora cannot deliver any of that."
On AI models: "Harvey and Legora run Claude and GPT-5. So does ServiceNow via BYOM — except on ServiceNow you choose the model per step, you can bring your own fine-tuned model, and the AI steps sit alongside your governed business logic in a single auditable workflow."

Summary

	ServiceNow LCO	Harvey	Legora
AI models	BYOM — any model, per step	Claude/GPT-5/Gemini (auto)	GPT-5/Claude/o1 (Azure)
Model you control	✅ Full	❌ Harvey controls routing	❌ Legora controls routing
Governed logic in flows	✅ First-class Flow Designer nodes	❌ AI-only flows	❌ AI-only flows
Mix AI + governed logic	✅ Native	❌	❌
Your enterprise use cases as flow nodes	✅ Any SN use case	❌	❌
CLM / obligation tracking	✅ CM Pro native	❌	❌
Legal intake + SLAs	✅ LSD native	❌	❌
GRC integration	✅ Native	❌	❌
A2A / MCP open protocol	✅	❌ Not announced	❌ Not announced
Bulk document review	✅ CM Pro + BYOM	✅ Vault (100k docs)	✅ Tabular review
Legal research DB	⚠️ Via integrations	✅ LexisNexis	✅ Qura (27 jurisdictions)

What to do next

If you are evaluating Harvey or Legora, the right first step is to map your legal team's actual workflow — not just the document review step, but the full journey from intake to obligation tracking to GRC to reporting. Then ask which platform governs that entire journey.

If you are already on ServiceNow LCO and want to explore Contract agentic workflows with BYOM, the ServiceNow LCO documentation and the Legal and Contract Operations community board are the right starting points.

Have you built a heterogeneous agentic flow on ServiceNow Legal and Contract Operations? Share your use case in the comments — this is exactly the kind of real-world pattern the community benefits from seeing.

This article reflects the state of the market as of May 2026. Harvey and Legora are evolving rapidly — always verify current capabilities directly with those vendors before making purchasing decisions.