TwinLadder Weekly

Issue #10 | June 2025

Editor's Note

I want to talk about something unfashionable: rules-based document automation. Not generative AI. Not large language models. Plain, deterministic, if-then-else logic that assembles documents from templates.

Last month, a colleague at a 30-person estate planning practice in Amsterdam told me she had been pressured by her managing partner to "get on AI" for document generation. She had been using Gavel (formerly Documate) for two years with excellent results — consistent output, client-facing intake, zero hallucination risk. Her managing partner had read about Harvey and wanted something that sounded more impressive.

This is the problem with the current conversation. We have become so fixated on generative AI that we have forgotten the tools that already work. Not every document needs intelligence. Most documents need consistency. That distinction matters more than any technology trend, and it is the subject of this issue.

I have watched this pattern before, across firms in Riga, Stockholm, and Berlin. The newest technology is not always the most appropriate technology. Sometimes the right answer has been available for years, quietly doing its job without a press release.

When Templates Beat Intelligence: Document Automation in 2025

Two Approaches, Two Problems

Two approaches dominate legal document creation, and they solve fundamentally different problems.

Rules-based document automation (Gavel, HotDocs, Woodpecker) uses conditional logic to assemble documents from templates. If the client is married, include spousal provisions. If asset value exceeds a threshold, trigger complex trust language. The output is predictable, auditable, and deterministic. The system does exactly what you told it to do, every time.

AI-powered contract lifecycle management (Ironclad, Icertis, Agiloft) manages contracts from creation through expiration. AI assists with drafting, redlining, extraction, and analysis. Broader scope, significantly more complexity, and substantially higher cost.

Gavel's core proposition is straightforward: lawyers save up to 90% of time previously spent generating legal documents. Having watched firms implement it across Europe and North America, I can confirm that claim holds for template-driven documents with known structures. Estate planning, residential conveyancing, standard corporate formations, routine leases — these are documents where the structure is settled, the variations are finite, and the value lies in eliminating human error, not generating novel text.

The Decision Framework

In December 2025, Gavel announced Gavel Workflows, transforming the platform from document automation into comprehensive workflow automation. CEO Dorna Moini stated something I wish more legal tech founders would say: "Not every legal document should be created by AI. When the document's structure is known, rules-based automation is faster, safer, and more accurate."

She is right. And the implications for mid-market firms are significant.

Approach	Best For	Cost	Hallucination Risk
Rules-based (Gavel, HotDocs)	Predictable structures, known variations	$83/month	Zero — deterministic output
CLM (Ironclad, Icertis)	Portfolio management, obligation tracking	$25,000+/year	Low — structured workflows
Generative AI (Harvey, CoCounsel)	Novel research, uncertain structures	Enterprise pricing	17-34% hallucination rate (Stanford)

The decision framework is simpler than vendors want you to believe. If your document structure is predictable and your variations can be expressed as if-then rules, use rules-based automation. If you need to manage thousands of active contracts with obligation tracking, renewal management, and cross-department access, consider CLM — but expect to spend $25,000+ annually with months of implementation. If you are generating documents where the structure itself is uncertain and requires judgment, that is where generative AI earns its place.

Many firms need elements of all three. A 40-person firm I work with in the Baltics uses Gavel for client intake and standard documents, a lightweight CLM for their commercial contract portfolio, and Harvey selectively for novel research questions. The systems coexist. The mistake is treating any single approach as universal.

The Governance Problem Nobody Plans For

Where I have seen firms go wrong: template maintenance. Building 50+ conditional templates is satisfying. Maintaining them across law changes, partner preferences, and jurisdictional updates is a governance problem most firms do not plan for. By year three, templates drift. Different people update different templates with different approaches. The consistency that justified automation erodes without dedicated ownership.

Liga Paulina, who advises on EU regulatory compliance, raises an additional concern for European practices: "Template maintenance is not just a quality issue in Europe — it is a compliance issue. When directives are transposed into national law, template language must be updated across every jurisdiction you serve. A firm operating across Latvia, Lithuania, and Estonia with outdated templates is not just inefficient. It may be producing documents that do not reflect current law." [HIGH CONFIDENCE]

For firms navigating the EU AI Act, there is an irony worth noting. Article 4's AI literacy requirement applies to AI systems — but rules-based document automation is not AI. It is deterministic logic. A firm that replaces generative AI with rules-based automation for appropriate tasks reduces both its hallucination risk and its Article 4 compliance burden. Sometimes the smartest AI strategy is knowing when not to use AI.

The tool that matches your actual problem — not your aspirational one — is the right tool. For most mid-market practices generating routine documents, that tool has existed for years. It does not require a press release.

The Competence Question

Picture a five-person estate planning team in Rotterdam that adopts Gavel. Within six months, the time per will drops from four hours to fifty minutes. Client-facing intake eliminates the information-gathering loop. Consistency is perfect. Everyone is pleased.

Now picture the same team eighteen months later. A junior associate has never drafted a will from scratch. She understands the intake form. She knows which template to select. But she has never confronted the underlying decisions — why a particular clause exists, what happens when facts do not fit the template, how to draft for a situation the template does not anticipate.

This is the competence question that automation of all kinds poses, not just AI. Rules-based systems are transparent — you can read the logic, trace the conditions, understand the output. But transparency does not guarantee that anyone is actually doing that work. If the system produces correct documents 99% of the time, the incentive to understand why it produces them diminishes with each successful generation.

The firms using automation well are the ones that treat template-building as a training exercise. Junior lawyers do not just use the templates. They build and maintain them. They articulate why clause A triggers when condition B is met. That process builds the judgment that no template can encode.

What To Do

Audit your document types. List every document your firm generates regularly. For each, answer: is the structure predictable? Can variations be expressed as rules? If yes to both, you have an automation candidate that does not require AI.
Evaluate Gavel or equivalent before Harvey. For template-driven practices, rules-based automation at $83/month may solve 80% of your problem at 2% of the cost of enterprise AI. Test the cheaper solution first.
Assign template ownership. Every automated template needs a named owner responsible for annual review, law change updates, and quality control. For multi-jurisdictional European practices, include a cross-border review process to ensure templates reflect each jurisdiction's current law.
Use template building as training. Require junior lawyers to build or substantially modify at least one template annually. The process of encoding legal logic forces the kind of structured thinking that routine document use does not.
Map your automation against your AI Act obligations. Distinguish between true AI systems (which trigger Article 4 literacy requirements) and rules-based automation (which does not). This distinction affects your compliance burden, your training requirements, and your risk profile. Document it clearly.

Quick Reads

Gavel Workflows launches — comprehensive workflow automation that extends Gavel beyond document assembly into full matter workflows.
Gavel expands to 60 languages — relevant for firms with cross-border practices or multilingual client bases, particularly across the EU's 24 official languages.
Sharable AI Playbooks — Gavel's interesting move into firm-client AI sharing, letting law firms sell productised legal tools to corporate clients.
Gavel integrates with Clio — the Gavel-Clio integration eliminates double data entry for firms already in the Clio ecosystem, which includes a significant share of the mid-market on both sides of the Atlantic.

One Question

If a junior lawyer has never drafted a document that a template now generates automatically, can she competently supervise the template's output — and how would she know if she could not?

TwinLadder Weekly | Issue #10 | June 2025

Helping European professionals build AI competence through honest education.