The Legal Transformation Institute's Guide To

Forward Deployed
Legal Engineer.

A field guide for people learning the work and organizations building the capability.

You do not need to want the title for these capabilities to matter. Use this guide to develop your own judgment, understand where colleagues contribute, shape a role, build a pod, or interview for the work.

The plain English edition · no AI background assumed

About a twenty minute read

Every firm can now buy artificial intelligence.

The same AI models, the same licenses, the same vendor demos, available to any firm able to pay for them. To be clear: this is machine intelligence, not lateral hires.

If every firm can access it, AI alone cannot be the advantage.

So where does the advantage go?

Into deployment.

R O I A  F O U N D A T I O N A L  M O D E L T H E  S A M E  C A P A B I L I T Y ,  E V E R Y  F I R M FIRM 1FIRM 2FIRM 3FIRM 4FIRM 5

The edge is no longer who has AI. It is where, how, and why a firm uses it.

Someone has to decide where AI belongs inside legal work.

That capability is
forward deployed legal engineering.

Legal Reality

How the work actually happens today: people, systems, exceptions, and the routing logic written down nowhere.

Deployment Judgment

Where AI belongs, and where it does not. What stays deterministic, what gets a model, what stays human.

A Defensible System

Working software carrying real responsibility: an audit trail, evals, and escalation to a lawyer built in.

The role matters because it controls how AI enters the firm, and whether it can be trusted once inside.

The Role

Two kinds of judgment, one capability

Forward deployed legal engineering is a capability before it is a job title. One unusually broad person may carry it. More often, a small pod combines the judgment required to understand the work, build the system, test it, and earn adoption. The operating standard is the same in either form.

Legal and commercial judgment

Matter economics, risk, privilege, professional responsibility. What clients are asking for, and what partners will actually trust and adopt.

Technical judgment

Models, agents, evals, guardrails, integration with the systems a firm already runs. Knowing when AI is the answer, and when it is overkill.

Interactive · The Profile

Two judgments break into six components

Nobody arrives holding all six, and waiting to hire someone who does is not a plan. The feasible path is the pod: three or four of your own people whose individual spikes compose into full coverage. A lawyer spikes on legal judgment and collapses on the technical side. A business professional spikes on workflow and commercial instincts. Toggle the profiles and watch the composition work.

Judgment about the work
Legal judgment. What correct, defensible work looks like. Selected for, then sharpened.
Commercial judgment. Firm economics, incentives, adoption. What is worth building at all.
Workflow intelligence. Decoding how work really happens, task by task, exceptions and all.
Judgment about the system
Technical fluency. Models, agents, integration, guardrails. Knowing when AI is overkill.
Evidence discipline. Evals, golden datasets, measurement. Defensibility and predictability.
The bridge
Translation. Extracting the unwritten from people's heads. Defending the built, twice.
The full hexagon belongs to the capability, not necessarily one person. Three or four people can compose into coverage: a lawyer, a technologist, a process mind. The assessment maps each person's contribution to that shared capability. It does not confer a title or credential.
How To Use The Hexagon

The title is optional. The capability is not.

The same six components can guide individual development, team coverage, role design, and evidence based interviews. The use changes. The operating standard does not.

01 · Develop Yourself

Move from confidence to evidence

Identify the artifacts you can already produce. Choose one thin component and build the next artifact to an objective quality bar.

02 · Map A Team

Look for coverage, not rankings

Ask people to complete their own profiles and share them voluntarily. Combine complementary spikes. Find the component no one currently owns.

03 · Design A Role

Specify the contribution before the title

Select the two or three components that must be strong on day one. State what can develop and what the surrounding team will supply. Do not write a six component unicorn description.

04 · Interview For The Work

Ask for artifacts and judgment

Use the components to structure questions. Ask what the person produced, what failed, what changed, and who reviewed it. Evidence carries more weight than fluency with terminology.

Questions that reveal the capability

  • Show how you mapped a workflow and found an exception.
  • Describe a point where you decided AI did not belong.
  • Walk through an evaluation and what the failures changed.
  • Explain a project you recommended against and the numbers supporting that decision.
  • Describe how you translated expert judgment into a buildable specification.
  • Show where human review, escalation, and an audit trail entered the system.

The Hexagon is a development and team conversation framework. It is not a validated employment selection test. Use it to improve job descriptions and interview questions, not to create score cutoffs or rank candidates.

The Loop

The job has three parts

Embed

learns how the work really happens and finds the workflow worth rebuilding.

Evals

prove the system behaves, with evidence a partner and a client can rely on.

Deployment

makes it work inside the firm, on the firm's own systems.

Each stage earns the right to the next
Interactive · Embedded With The Work

The stated workflow is never the real workflow

Ask a partner how NDA intake works and you will get five clean steps. The real workflow lives in the exceptions, the workarounds, and the routing logic that exists only in one senior lawyer's head.

The forward deployed legal engineer traces every step

Click any step. The clean version hides the real work.

Half of what you just read is written down nowhere. It lives in people's heads, and if you build on the stated workflow instead of the real one, you are automating a fiction. Decoding this gap is a trained skill. We call it workflow intelligence, and it is where every serious deployment begins.
The Loop · Part One

Embed: find the workflow worth rebuilding

Sitting with the work determines what should be built before anyone builds anything. It ends in one artifact the whole firm can argue with.

1  Collect the context
interviews emails precedents playbooks DMS exports billing entries approvals exception paths
2  Trace the findings
bottlenecks
repetitive work
judgment points
required systems
failure modes
privilege boundaries
what is worth building
3  Produce the operating map
What embedding produces · one operating map
Current statehow the work actually happens today
Future statethe same work, rebuilt around intelligence
Selected use caseone workflow, chosen for value
Boundarieswhat the system may and may not do
Escalation ruleswhat must always reach a lawyer
Expected valuehours, risk, and cost, quantified
The Loop · The Judgment Call

Decide what should be automated, and what should not

a step in the mapped workflow DETERMINISTIC SOFTWARE When the rules and inputs are predictable. AN AGENT When the objective is clear but the inputs, path, or required actions vary. A LAWYER IN CONTROL When the decision carries ambiguity, accountability, privilege, or irreversible consequences.

swipe the diagram sideways if it is cut off

Prioritize high volume workflows where the improvement is large enough to matter. Then design the split.

The Design

Then decide how the work should operate with AI built in

Most of the workflow should not be AI at all. Deterministic software does what it is told. The model gets exactly the steps that need judgment. A lawyer holds the decision. And every step, human or machine, lands in the audit trail.

an exception escalates to a lawyer

Deterministic software

intake · logging · routing

The agent acts

extracts · compares · drafts

A lawyer approves

one clear decision

The record updates

every system · obligations tracked
Audit trail · every step recorded

The forward deployed legal engineer understands the current workflow, designs the intelligent one, and builds the system that connects the two.

The Loop · Part Two

Evals: turn nondeterminism into evidence

A golden dataset of real cases, labeled by lawyers. Run the system against it, keep the evidence, and set the rules for what must always escalate. Law firms need defensibility and predictability in their agentic workflows. This is how you deliver both.

The evaluation cases
Right dataRequired stepsMatches the expertSafe to act
The standard NDA
The unusual counterparty form
A missing exhibit · · → lawyer
An ambiguous instruction · → lawyer
A material indemnity change → lawyer
✓ expected behavior  ·  ✗ failure, with the evidence kept  ·  → lawyer, escalated by rule

swipe the table sideways to see all columns

The evaluation report
Pass rate
0 / 50 graded runs passed · 82%
Failure categories · 9 runs
missing exhibit
5
wrong precedent pulled
4
Operating rules
Confidencebelow 0.80, the system must escalate
Escalationprivilege and high risk terms always reach a lawyer
Readinesspilot, with lawyer review on every action
The Loop · Part Three

Deployment: make it work inside the firm

S A N D B O X THE SYSTEM IN PRODUCTION matter data and precedents firm systems and APIs identity and permissions lawyer review logs alerts rollback the practice group

swipe the diagram sideways if it is cut off

1Integrate with what already exists

Build on the DMS, the matter system, and the research stack. Never force a migration.

2Test inside the firm's walls

A sandbox on the firm's own licenses, to run, inspect, and debug safely. No client data leaves.

3Increase autonomy gradually

Shadow mode first. The smallest useful action. More authority only as the evals prove it.

Deployment is where software begins carrying professional responsibility.

Map your contribution to the capability. Six components. About nine minutes. You receive an individual evidence and confidence profile immediately.

Take the assessment
Who Is Behind This

We have been rebuilding legal work since before the tools could keep up

Between us, more than twenty years across every era that led here. Each one taught the same lesson from a different angle: the technology is never the hard part. The work, the incentives, and the people are.

lean and process engineering legal process outsourcing ediscovery at scale the ALSP build out firm strategy and growth forward deployment

Josh Kubicki

Strategy, and the buyer's side of the table

Strategy counsel to global law firm leadership on practice, pricing, and growth. Professor of legal innovation who built the University of Richmond's Legal Business Design Hub, winner of Fast Company's 2022 Innovation by Design Award and a finalist in Bloomberg Law's inaugural Law School Innovation Program. Currently teaches this work at Indiana University Maurer School of Law. Publisher of The Brainyacts, read daily by thousands of senior legal professionals. Has built books of business past $20 million by redesigning how firms sell and deliver work. Six Sigma Green Belt: workflow auditing and future state design.

Rob Saccone

Engineering, and the operator's side

Technology founder who built a legal software company and sold it to Thomson Reuters. Two decades of systems inside and alongside the Am Law 100: portals, platforms, knowledge systems, and now agentic AI. Six Sigma Black Belt. Fluent in both firm economics and software architecture, which is rarer than it should be.

Every era taught us what survives contact with a law firm. This is the first one where the tools can keep up.

The initial benchmark and instrument study are being conducted with Indiana University Maurer School of Law and Ice Miller LLP.

Why Now

What is known, and what nobody knows yet

Known

Current postings put a price on the work

An OpenAI forward deployed engineering posting lists $162K to $280K plus equity. Current Harvey legal engineering postings list total compensation from $220K to $320K plus equity. Titles and compensation structures vary. The market signal does not.

Sources: OpenAI · Harvey

Most pilots show no measured financial return

A widely cited 2025 MIT NANDA preprint estimated that 95 percent of the enterprise generative AI initiatives it examined showed no measurable profit and loss impact. That is one early study, not a universal failure rate. Its useful finding is narrower: access to a capable model does not create deployment value on its own.

Source: The GenAI Divide · State of AI in Business 2025

Capability compounds

Model capability improves every quarter. Firm workflows do not. The gap between what is possible and what firms capture widens on its own.

Not yet known

The billable hour

What agentic delivery does to hourly billing, rates, and realization.

Nobody knows yet

Client stickiness

How clients reprice loyalty when the work gets faster and partly machine made.

Nobody knows yet

The leverage model

Which firm structures survive when execution stops being scarce.

Nobody knows yet
These questions will not be answered from the sidelines, and a license is not a strategy: the same subscriptions sit on every competitor's desks. The answers will belong to the firms that go learn them on their own work, with the right people at the table. The sooner you start, the sooner you stop guessing. Move with intention.
Common Questions

What people ask about the Hexagon

The framework is yours to use. These are the limits worth knowing before you do.

Can I get certified in this?

No. This is version one of a self assessment with no norms behind it yet. A credential would be a lie. What you can produce is artifacts, and those are checkable by anyone.

Can I use the framework in my own job descriptions and interviews?

Yes, freely. That is what it is for. The six components and the interview questions in this guide are yours. No permission needed and no attribution required.

Can I use the assessment to screen candidates?

No. It is self reported and unvalidated, and people who want a job will report generously. Use it to write a better role and ask better questions. Do not use it to rank people.

Can my whole team take it and see the results together?

The assessment is free for individuals and always will be. A composite read across a team, showing where you have coverage and which component nobody owns, is something we do with firms directly. Write and ask.

We are trying to hire a legal engineer. Can you help?

We do not recruit. We will help you define the role and build an evidence based interview for it. Map your existing people first: most firms hunting a unicorn are missing two specific components and already hold the other four.

Is this only for large firms?

No. The six components describe the capability at any size. A four person legal department composes a pod the same way an Am Law 50 practice group does, with fewer people wearing more of the hexagon.

What Comes Next

This guide gives you the language. The next move is yours.

Use the framework. Produce evidence. Follow the work as the field develops.

Understand

Map the capability

See where you contribute.

Complete the Hexagon assessment. Use the result to develop your skills, understand team coverage, shape a role, or prepare evidence for an interview.

Build

Produce the next artifact

Move from confidence to evidence.

Your assessment result opens an artifact brief for the thinnest reported component, with a concrete output, quality bar, and reviewer.

Follow

The Brainyacts

Pragmatic and responsible AI in legal.

Follow the field notes, operating ideas, and evidence as forward deployed legal capability develops in practice.

Working Together

There is nothing to buy on this page.

We do the work described above. Embedding with a practice group, mapping how it really operates, building the evals, getting a system into production on a firm's own systems. It is done in person, which is why we take on very few firms.

We do not recruit or place candidates. We do not sell software and we take no referral fees from anyone who does. We do not certify anyone. We are not going to sell you a transformation program.

If you want help with a team, a practice group, or a firm, write to me.
jkubicki@transformlegal.com — tell me the problem, not the solution.

They build shallow for show.
We build deep for keeps.

Legal Transformation Institute · transformlegal.com