Skip to main content
AGENCYSCRIPT
CoursesEnterpriseBlog
đź‘‘FoundersSign inJoin Waitlist
AGENCYSCRIPT

Governed Certification Framework

The operating system for AI-enabled agency building. Certify judgment under constraint. Standards over scale. Governance over shortcuts.

Stay informed

Governance updates, certification insights, and industry standards.

Products

  • Platform
  • AI Scripts
  • Certification
  • Launch Program
  • Vault
  • The Book

Certification

  • Foundation (AS-F)
  • Operator (AS-O)
  • Architect (AS-A)
  • Principal (AS-P)

Resources

  • Blog
  • Agency Archetype Quiz
  • Free Live Training
  • Build AI Agents Masterclass
  • Build with AI Challenge
  • OS Plugin Install
  • Verify Credential
  • Enterprise
  • Partners
  • Pricing

Company

  • About
  • Contact
  • Careers
  • Press
© 2026 Agency Script, Inc.·
Privacy PolicyTerms of ServiceCertification AgreementSecurityCookies

Standards over scale. Judgment over volume. Governance over shortcuts.

On This Page

The Three ApproachesAdopt a commercial platformAssemble your ownBolt a generative layer onto existing databasesThe Axes That DecideControl over the corpusTime to first valueTotal cost over three yearsResponsibility for accuracyWhere Each Approach WinsWhen commercial winsWhen building winsWhen the bolt-on winsA Decision Rule for UncertaintyPrefer the approach you can exitDefault to buy unless you have a specific reason not toRunning the Decision in PracticeA mid-sized litigation firmA firm with a validated niche databaseFrequently Asked QuestionsIs building my own system ever worth it for a mid-sized firm?How do I compare costs fairly across the three approaches?Can I start with one approach and switch later?What if different practice groups want different approaches?Which approach has the lowest risk?Key Takeaways
Home/Blog/Build, Buy, or Bolt On: Choosing a Legal Research Approach
General

Build, Buy, or Bolt On: Choosing a Legal Research Approach

A

Agency Script Editorial

Editorial Team

·September 2, 2016·7 min read
ai legal research platformsai legal research platforms tradeoffsai legal research platforms guideai tools

When a firm decides it wants AI in its research workflow, the decision is rarely framed correctly. People argue about which vendor to pick when the deeper question is which approach to take — and the approaches are genuinely different, with different cost curves, control surfaces, and failure modes. Picking a vendor before picking an approach is how firms end up locked into a model that fights their actual needs.

There are three broad approaches in play: adopt a commercial platform wholesale, assemble your own retrieval system over licensed or public corpora, or layer a generative tool on top of databases you already pay for. Each is defensible. None is universally right.

This piece lays the three approaches side by side, names the axes that distinguish them, and gives you a decision rule that holds up when the future is uncertain. The aim is to make the choice deliberate rather than accidental.

The Three Approaches

Naming the options precisely is half the work, because much confusion comes from comparing things at different layers.

Adopt a commercial platform

You buy a finished product — corpus, retrieval, synthesis, citations, and workflow in one package. You trade control for speed: the vendor handles currency, accuracy, and infrastructure, and you accept their choices about how research works.

Assemble your own

You license or ingest corpora and build retrieval and synthesis yourself. This gives maximum control and the ability to tune for your exact practice, at the cost of real engineering investment and ongoing maintenance. It only makes sense at scale or for unusual requirements.

Bolt a generative layer onto existing databases

You keep the authoritative databases you already trust and add a generative interface for speed and natural-language querying. You get a middle path — modern interaction without abandoning sources you have validated over years. The appeal is that you change the experience without re-litigating the foundation, which both reduces risk and shortens the time before attorneys see a benefit. Its limit is that you are constrained by what your existing databases expose and by how well a generative layer can sit on top of them.

These three are genuinely different commitments, not three flavors of the same purchase. Confusing them is the most common mistake in early conversations, where someone compares a finished commercial product against a hypothetical homegrown system and reaches a conclusion that an honest, like-for-like comparison would reverse.

The Axes That Decide

Once the approaches are clear, a handful of axes separate them cleanly.

Control over the corpus

Commercial platforms decide what is in scope; you live with their coverage. Assembling your own gives total control but total responsibility. The bolt-on path keeps you on corpora you have already vetted, which matters for practices with idiosyncratic source needs.

Time to first value

The commercial approach delivers value in days. The bolt-on path in weeks. Building your own in months or quarters. If you need results this fiscal year, the assembly path is effectively off the table regardless of its other merits.

Total cost over three years

License fees look expensive until you price the engineering, maintenance, and opportunity cost of building. Run the comparison over three years, not one, because the build path front-loads cost and the buy path spreads it. The discipline here mirrors Putting Numbers Behind a Legal Research Investment. The trap is comparing a license fee against an engineering estimate that conveniently omits ongoing maintenance, corpus updates, and the attorney time spent supervising a homegrown system. When those are included honestly, the build path's apparent cost advantage often inverts.

Responsibility for accuracy

A commercial vendor owns the correctness and currency of its corpus; when you build, you own it. That responsibility is easy to underweight at the decision point and impossible to ignore once a homegrown system surfaces an overruled case nobody flagged. The question of who is accountable when the research is wrong is a real axis, not a footnote, and it weighs heavily toward buying for most firms.

Where Each Approach Wins

The approaches are not interchangeable; each has a natural home.

When commercial wins

Most firms, most of the time. If your practice resembles other firms' practices, a commercial platform has already solved your problem and amortized the cost across a large customer base. Reinventing that is rarely justified.

When building wins

Very large organizations with unusual corpora, strict data-residency requirements, or a research workflow so specific that no product fits. The bar is high, and the maintenance burden is permanent, so the decision should survive a skeptical review.

When the bolt-on wins

Firms with deep investments in validated databases who want modern interaction without re-vetting their sources. It is the pragmatic middle, and it is more common than the marketing for either extreme would suggest. It also lowers the change-management cost covered in Bringing a Whole Practice Onto New Research Tools.

A Decision Rule for Uncertainty

Because you cannot perfectly predict your needs, bias toward reversibility.

Prefer the approach you can exit

A commercial subscription you can cancel beats a custom build you cannot easily unwind. When two approaches score similarly, the more reversible one wins, because it preserves your ability to react when you learn more. In a market changing as quickly as this one, the value of being able to change your mind cheaply is unusually high, and it is a value that rarely shows up on a feature comparison. Treat reversibility as a first-class benefit rather than a fallback, and many close decisions resolve cleanly.

Default to buy unless you have a specific reason not to

The burden of proof sits on building or heavily customizing. If you cannot articulate a concrete requirement that no product meets, the answer is to buy. This default fails safe.

Running the Decision in Practice

The framework is only useful if you can apply it to a real situation, so consider how it plays out.

A mid-sized litigation firm

Picture a firm with a heavy litigation practice and no unusual data requirements. Its needs resemble those of hundreds of similar firms, time-to-value matters because cases are active now, and it has no engineering team to maintain a custom system. Every axis points the same direction: buy a commercial platform with strong citator coverage. The decision is not close, and the framework simply confirms what the constraints already imply.

A firm with a validated niche database

Now picture a firm that has spent years validating a specialized regulatory database it trusts deeply. It wants natural-language interaction but is unwilling to re-vet a new corpus. Here the bolt-on path is the obvious fit — modern interaction on top of a foundation it already trusts — and forcing a wholesale commercial migration would discard real, hard-won value. The two scenarios use the same axes and reach different answers, which is exactly how a sound framework should behave.

Frequently Asked Questions

Is building my own system ever worth it for a mid-sized firm?

Rarely. The engineering and maintenance cost almost always exceeds the value unless you have a requirement no product addresses — unusual data residency, a proprietary corpus, or a workflow that is core to your differentiation. For most mid-sized firms, buying or bolting on wins.

How do I compare costs fairly across the three approaches?

Use a three-year horizon and include hidden costs: maintenance, internal engineering time, training, and the opportunity cost of attorneys building instead of practicing. The build path looks cheaper on a one-year license comparison and far more expensive over three.

Can I start with one approach and switch later?

Yes, and you should design for it. Starting commercial and migrating to a bolt-on or build later is common. Starting with a build and walking it back is painful, which is another argument for the reversibility rule.

What if different practice groups want different approaches?

That is often the right answer. A high-volume group might use a commercial platform while a specialized group bolts a generative layer onto a niche database. Uniformity for its own sake is not a goal.

Which approach has the lowest risk?

The bolt-on path tends to carry the lowest risk because it keeps you on validated sources while adding modern interaction. You are changing the interface, not the foundation.

Key Takeaways

  • Decide on an approach — buy, build, or bolt on — before evaluating any vendor.
  • The deciding axes are corpus control, time to first value, and three-year total cost.
  • Commercial wins for most firms; building wins only at scale or for unusual requirements; the bolt-on is the pragmatic middle.
  • Under uncertainty, prefer the approach you can exit cheaply.
  • Default to buying unless you can name a concrete requirement no product meets.

Search Articles

Categories

OperationsSalesDeliveryGovernance

Popular Tags

prompt engineeringai fundamentalsai toolsthe difference between AIMLagency operationsagency growthenterprise sales

Share Article

A

Agency Script Editorial

Editorial Team

The Agency Script editorial team delivers operational insights on AI delivery, certification, and governance for modern agency operators.

Related Articles

General

Rolling Out AI Hallucinations Across a Team

Most teams discover AI hallucinations the hard way — a confident-sounding wrong answer makes it into a client deliverable, a legal brief, or a published report. The damage isn't just to the output; it

A
Agency Script Editorial
June 1, 2026·11 min read
General

A Model Behind an API Is Only Potential

Large language models don't do much on their own. A model sitting behind an API is potential, not capability. What converts that potential into something useful—something that drafts, classifies, summ

A
Agency Script Editorial
June 1, 2026·11 min read
General

Case Study: Large Language Models in Practice

Most teams that fail with large language models don't fail because the technology doesn't work. They fail because they treat deployment as a one-time event rather than a discipline — pick a model, wri

A
Agency Script Editorial
June 1, 2026·11 min read

Ready to certify your AI capability?

Join the professionals building governed, repeatable AI delivery systems.

Explore Certification