Expand Matter Capacity with Governed AI

The pyramid under privilege: how a law firm actually earns

The principle

Strip the mystique and a law firm is a pyramid that converts judgment into billed time. Partners win matters and take positions, associates produce the hours, and the economics turn on four dials: the rate the firm commands, the hours it bills, the leverage it runs between partner and associate time, and the realization that survives write-downs and collection. Alternative fee arrangements sit at the edges of that model, and wherever they appear they quietly reprice everything, because under a fixed or capped fee, efficiency stops threatening revenue and becomes margin. Over all of it sits privilege: everything the firm touches is someone's confidence, which makes every operational question in this playbook a professional one as well. And the binding constraint is the familiar one: partners who can win the work, take the positions, and sign the filings do not scale by hiring.

The trap

The trap is buying throughput while ignoring the fee structure it lands in. In an hourly practice, an hour saved by a tool is not automatically margin; unexamined, it is simply an hour that was not billed. The same hour inside a fixed-fee matter is margin directly. Firms that skip this decision get the worst of both: associates quietly worried that efficiency is disloyalty, partners quietly suspicious that the tools are eating the inventory, and a program that stalls because nobody said out loud where saved time is supposed to go.

The checklist

• Name the constraint in each practice group: winning matters, staffing them, producing the work, or supervising it. They are different problems with different fixes.
• Follow one week of partner hours and mark which of them only that partner could have spent: positions, negotiations, client counsel, sign-off.
• Map the book by fee structure, hourly against alternative arrangements, because that map decides what saved time is worth and where.
• Count how often the firm redrafts what it has already drafted: agreements, briefs, memos rebuilt from memory at associate rates.
• Choose workflows to improve, not tools to buy. A tool dropped on an unchanged workflow returns almost nothing.

Where judgment beats the tool

An hours analysis shows where partner time goes. It cannot say which of those hours are the counsel the client is paying for. Some partner time spent slowly, on a position, on a negotiation, on a hard conversation, is exactly the product, and automating around it would cheapen the firm. Deciding which senior hours are the value and which are habit is a call only the partnership can make.

Four outcomes that matter, and one that does not

The principle

Four outcomes justify this whole program. Take on more matters without proportional hiring, so the book can grow faster than headcount. Increase delivery capacity without sacrificing quality, so the growth does not quietly spend the firm's name. Turn the firm's accumulated work product into controlled, reusable leverage, so its best drafting compounds instead of retiring with its authors. And adopt AI while protecting privilege, professional judgment, and client trust, because in this profession trust is the license to operate. Notice what is not on the list: adopt AI. Adoption is a means. The moment it becomes the goal, the program starts optimizing for usage instead of for the firm, and usage is the one thing no client ever asked for.

The trap

The trap is measuring the means. Licenses issued, seats active, prompts per lawyer per week: activity metrics reward the appearance of change while the operating outcomes sit unmeasured. A firm can hit every adoption number it sets and end the year with the same pitch turnaround, the same redrafted agreements, the same write-downs confirmed at the bill, and the same partners underwater.

The checklist

• Write each of the four outcomes as an operating sentence with a named owner, not a slogan on a retreat slide.
• Tie every initiative to exactly one outcome. An initiative that maps to none is a hobby.
• Baseline the outcome metrics before the first pilot, or the after will have no before.
• Retire activity metrics from management reporting. Keep them for operations, where they belong.

Where judgment beats the tool

Outcomes conflict at the margin: pushed far enough, capacity presses on quality, and reuse presses on the bespoke attention each matter deserves. A dashboard will not arbitrate that tension. Where the firm sets each trade, matter by matter and client by client, is a leadership decision, and it is the one clients experience directly.

Who has to say yes: the six chairs in the room

The principle

Nothing durable happens in a partnership without consensus, and this program touches every chair at the table. The managing partner or chair asks whether it grows the firm without diluting what the name stands for. The executive director asks whether operations get more reliable or just busier. Practice group leaders ask what happens to the quality of work that carries their group's name. The CIO or director of innovation asks whether this becomes one governed platform or a different tool in every practice. The general counsel asks what it does to privilege, confidentiality, and the firm's professional obligations. The CFO asks what it costs, what it returns, and how the answer survives the billable hour. Six different questions, and the program has to hold a real answer to all of them.

The trap

The trap is the champion-led initiative that answers one chair. An innovation push moves fast on borrowed enthusiasm, then dies in a partner meeting the day the general counsel asks the privilege question nobody prepared for, and the silence costs the champion a year of credibility. Partnerships rarely vote a program down; they decline to adopt it, which looks gentler and ends it just as completely.

The checklist

• Map the six chairs to named people, including the ones who hold the role without the title.
• Write down each chair's question and the evidence that would satisfy it, before the program is proposed.
• Brief the skeptics privately before the partner meeting, not at it.
• Give the general counsel a genuine design seat. Controls added at the end read as concessions; controls designed in read as competence.

Where judgment beats the tool

An org chart names the titles. It does not reveal whose quiet no actually ends a program in your partnership, whose support moves the undecided, or which practice group's blessing the others wait for. Reading the real decision structure of your own firm is judgment, and no tool has ever held it.

Precedent-to-Draft: the firm has written this before

The principle

Precedent-to-Draft runs from the moment a matter needs a document to the moment an attorney signs it out: precedent retrieval, first drafts built from the firm's own work product, research support, and attorney review before anything leaves the firm. It is where the pyramid spends most of its hours, and where an enormous share of them go to reinvention, because the firm's best drafting lives in matter files and partner memory, findable only by whoever happens to remember it. The redesign moves retrieval and first assembly onto governed rails, grounded in what the firm has already written and vetted, and returns attorney attention to the two things that were always the job: the position the document takes, and the judgment that it is right.

The workflow, stage by stage

• Precedent retrieval: the firm's own agreements, briefs, and memos surface with provenance and matter context, walled by client and matter, instead of living in whoever drafted them.
• First drafting: documents start from approved firm work product and playbook language, so associates edit upward from the firm's best instead of assembling from a blank page.
• Research support: background arrives synthesized with the source trail attached, so the attorney's verification is a check against sources, not an act of faith.
• Attorney review: the supervising attorney reviews position, authority, and risk with provenance visible, and verifies every authority cited before it is relied on.
• Sign-out: nothing leaves the firm, for a court, a counterparty, or a client, without an attorney's sign-off, exactly as before.

The trap

The trap is fluency. An AI-assisted draft reads finished: confident, well structured, correctly formatted, and possibly wrong. The risk is documented rather than imagined: the same field experiment that reported AI gains on tasks inside its capability frontier has also been reported to show quality declining when professionals relied on it for tasks outside that frontier (Dell'Acqua et al., 2023; consulting tasks at one elite firm, not legal work, and where the frontier sits shifts with the model and the domain). In a law firm the outside-frontier failure has a specific shape: the authority that was never verified, the citation that does not exist or does not say what the draft claims, the legal conclusion asserted without an authority check. Unverified citations have consequences in front of courts that no operations program can absorb. A draft that reads finished invites a tired reviewer to skim, and skimming is how a machine's error becomes an attorney's error.

The checklist

• Curate the precedent bank: approved exemplars with owners, matter context, and review dates, not the whole document management system.
• Wall retrieval by client and matter, mirroring the firm's ethical screens, so one client's confidences never surface in another client's draft.
• Have the attorney verify every cited authority in the source before the draft is filed or sent. The signature admits no exceptions, so the process cannot either.
• Keep supervising review pointed at position and risk by making provenance visible at the point of review.
• Track first-draft turnaround and review cycles together. Either one alone will lie to you.

Where judgment beats the tool

Retrieval can find the closest precedent the firm has. It cannot decide what this client's position should be, which arguments to make and which to hold, or what risk this client can carry. The draft is assembly. The position is counsel, and counsel is what the client is paying for.

Pitch & Matter Intake: win the right matters, opened cleanly

The principle

Pitch & Matter Intake runs from the first signal of a new matter to a cleanly opened one: pitches and RFP responses, experience retrieval, conflicts-aware intake support, engagement terms, and the handoff to the matter team. It decides two things at once: whether the firm wins the work, and whether the work should be won. Most of the hours in it are assembly, gathering what the firm knows about its own experience, its people, and its terms, and most of that assembly is senior time spent below its value. The redesign puts the assembly on governed rails and returns partner attention to the calls that matter: whether the matter fits the firm, on what terms, and at what fee.

The workflow, stage by stage

• Experience retrieval: the firm's relevant experience and teams surface with provenance, cleared for confidentiality before any of it appears in front of a client.
• Pitch and RFP drafting: first drafts start from approved capability language and experience records, not from the last pitch someone can find.
• Conflicts-aware intake: intake support assembles what the conflicts process needs, and waits for it. The clearance decision belongs to the people and procedures that own it, never to a tool.
• Engagement terms: engagement letters draft from approved templates, with scope, staffing, and fee structure as partner decisions rather than defaults.
• Handoff: what was pitched and promised arrives with the matter team as a structured record, including the budget the pitch implied, so the matter starts the way it was sold.

The trap

The trap is a pitch machine that outruns the firm's discipline. When drafting gets cheap, the temptation is to answer everything, and pursuit judgment quietly dies. Worse is intake racing ahead of conflicts: a matter opened on an incomplete clearance is not a growth story, it is a problem the firm volunteered for. Faster pitches only pay when the gates they pass through hold.

The checklist

• Keep conflicts clearance ahead of everything. No pitch ships and no matter opens while the check is incomplete.
• Clear every experience reference for confidentiality before it reaches a pitch, because what the firm knows and what it may say are different things.
• Draft engagement letters only from approved templates, and treat scope and fee structure as decisions a partner makes on each matter.
• Hold partner review of promise, team, and fee on every pitch. Speed is not a reason to skip the one review that prices the firm's word.
• Hand off every win as a structured record of scope, staffing, and budget, so the matter team builds what was promised instead of what was remembered.

Where judgment beats the tool

Retrieval can assemble everything the firm knows about a pursuit. It cannot decide whether the client is worth the rate pressure, whether a waivable conflict should actually be waived, or whether the matter belongs in the firm at all. Acceptance is a judgment about risk, relationships, and the kind of firm the partners are building, and it never belongs to a tool.

Matter-to-Management Insight: reporting partners can rely on

The principle

Partners run matters and firms on synthesized truth: matter status, budget against actual, staffing load, and the health of the portfolio. In most firms that truth is assembled by hand when someone asks, or not at all, and the first reliable signal of a budget problem is the pre-bill nobody wants to send. Matter-to-Management Insight rebuilds the picture from live sources instead: status synthesized from time entries, dockets, and matter plans; budget-to-actual by matter phase while the phase is still open; portfolio reporting on one spine; and every number traceable to the artifact behind it. Alternative fee arrangements raise the stakes, because on a fixed or capped fee a budget overrun is not a realization problem at the bill. It is the firm's own money, leaving quietly, in real time.

The trap

The trap is synthesis that launders problems. A model summarizing matter reports will smooth a slipping matter into confident prose, and a management committee can sit on top of a red matter reading amber for a month. The failure is not the summary. It is a reporting chain where no number can be challenged, because no number can be traced.

The checklist

• Give every number one source of record: time entries for hours, the matter plan for phase and budget, and synthesize from sources, not from prior summaries.
• Put budget-to-actual by phase in front of the responsible partner while the phase is open, not at pre-bill.
• Escalate variance to a named partner with a date, never to a distribution list.
• Keep client reporting human: synthesis prepares the status, and the relationship partner owns what the client hears.
• Read the portfolio on one spine, so practice leaders and firm management argue about decisions instead of about whose numbers are right.

Where judgment beats the tool

Synthesis can surface the variance. It cannot decide whether the overrun is scope the client should hear about, an investment in a relationship worth keeping, or a staffing problem to fix quietly. That conversation, with the client or with the team, is the partner's, and how it goes is worth more than the number that triggered it.

Attorney review and privilege boundaries: where the signature lives

The principle

The profession solved supervision long before AI arrived. Attorneys answer for work performed under their direction, whoever or whatever performed it, and assistance has never diluted accountability. The profession has now said so about AI in terms: ABA Formal Opinion 512 (July 2024) walks through the duties a lawyer using generative AI must fully consider, competence, confidentiality, communication, supervision, meritorious claims, candor toward the tribunal, and reasonable fees, and it sets the competence bar exactly where an operations program can reach it, a reasonable understanding of the specific tool's capabilities and limits, not expertise in AI (the opinion interprets the ABA Model Rules; state adoption and state opinions vary, and it is not jurisdiction-specific legal advice). That existing architecture is the control plane for everything in this playbook, applied deliberately: a named supervising attorney on every AI-assisted workflow; verification of every authority, every quotation, and every factual assertion an assisted draft relies on, done by the attorney in the source; privilege and confidentiality boundaries mirrored into the systems themselves, so the walls that bind people bind the tools; and sign-out points where a person, not a process, decides that work leaves the firm. None of this is new duty. It is the old duty, written down where the new leverage operates.

The trap

The trap is review that decays into skimming. Fluent drafts are the cause: work that reads finished invites trust it has not earned, and a saturated reviewer approves what a skeptical one would have caught. The signature does not distinguish between what the attorney wrote and what the attorney adopted, and neither does the court or the client. Review capacity is therefore the real ceiling on everything else in this playbook. Throughput that outruns it is not capacity. It is exposure, accumulating quietly.

The checklist

• Name a supervising attorney on every AI-assisted workflow before it runs, not after it produces something.
• Verify every cited authority in the source before filing or sending. A citation nobody checked is a risk nobody priced.
• Mirror ethical screens and matter walls in system access, provisioned and revoked with the matter team.
• Keep privileged and confidential material inside approved systems only, and keep the list of approved systems short enough to know.
• Budget review capacity like the scarce resource it is. When throughput rises, review time is the first thing to protect, not the first thing to trim.

Where judgment beats the tool

A procedure can force verification. It cannot make the supervising attorney's call: whether the argument serves this client, whether the position is one the firm will stand behind, whether the draft that is technically defensible is actually wise. Supervision is not a checkpoint. It is the practice of law, and it does not delegate.

Where the risk lives: privilege, confidentiality, and client trust

The principle

A law firm is trusted with other people's confidences under privilege, and AI touches that trust in four places. Confidentiality and privilege: client material entering prompts, corpora, and tools no engagement ever approved, and privilege questions about third-party systems that the firm's own general counsel should be answering before the tools run, not after. The warning is now written at bar level: California's generative AI guidance holds that a lawyer's duties apply in full and cautions that these tools may use prompts and uploaded documents for training or share queries with third parties (State Bar of California; one state's guidance, persuasive elsewhere but not binding, and vendor data practices change, so current terms govern). Verification: authority that reaches a court unverified, where the consequences land on the attorney and the client, never on the tool. Judgment: positions and advice quietly shaped by a model's framing instead of an attorney's conviction. And client guidelines: some clients' outside counsel guidelines now address AI use directly, tool by tool and disclosure by disclosure, and they do not all say the same thing (a practice observation; the research library holds no survey of guideline language, so treat how common this has become as unmeasured), which means the firm's posture is no longer only its own decision. None of these risks argues against the program. All of them argue for running it inside boundaries, under the rule this playbook repeats on purpose: AI assists operations and documentation, and attorneys retain every legal judgment, every client position, and every signature.

The trap

The trap is treating confidentiality as an IT setting instead of a professional boundary. A tenant toggle does not know which matter a document belongs to, what this client's guidelines prohibit, or that two of the firm's clients sit on opposite sides of a market. General policy plus default settings is how a firm ends up technically compliant and actually exposed, and in this profession the exposure is not an operations incident. It is an ethics problem with a client's name on it.

The checklist

• Set boundaries at the client and matter level, mirroring ethical screens. Tenant-level settings protect the firm's software, not its clients.
• Check each engagement's AI posture against that client's outside counsel guidelines, because some now speak to AI directly and they do not all say the same thing.
• Keep an approved-tool list with a sanctioned intake path for new needs, so the shadow alternative never becomes the easier one.
• Route privilege questions about new tools to the firm's general counsel before client material touches them.
• Stand up an incident path that assumes a miss will eventually happen, and rehearses who tells the client, how fast, and with what in hand.

Where judgment beats the tool

A control framework sets the floor. It cannot weigh whether this matter, this client, or this dispute justifies more caution than the standard, and it cannot conduct the conversation with a client who asks hard questions about the firm's AI use. Proportion and candor are relationship judgments, and they belong to the partner who holds the relationship.

The maturity path: baseline, AI-enabled, selectively AI-native

The principle

There are three honest stages, and they belong to workflows, not to firms. Baseline is where most legal work lives today: precedent in partner heads, drafting from the last file someone can find, status by asking around. It is not failure; it is a leverage ceiling. AI-enabled is the working middle: the same workflow, with retrieval, drafting, and synthesis assisting inside governed boundaries, and attorney review exactly where it was. Most law-firm workflows should live here, and the profession's supervision duties are why: enabled keeps the attorney's hands on everything that matters while removing the assembly beneath it. Selectively AI-native is the far stage: a workflow redesigned around governed AI because volume and structure justify it, recurring document types, intake assembly, status synthesis, with attorneys owning judgment and sign-off by design. The operative word is selectively. The path is walked one workflow at a time, on evidence, and the positions a client pays for never walk it at all.

The trap

The trap is the firm-wide transformation that tries to make everything native at once. It burns a year on platform debates, frightens the partners whose consent it needs, and collides with the reality that much of legal work is bespoke by nature. The opposite trap is subtler: staying enabled forever out of comfort, while two or three recurring, high-volume workflows have long since earned the redesign and quietly keep consuming associate years.

The checklist

• Place each significant workflow on the path separately. The firm does not have a maturity level; its workflows do.
• Earn native with evidence from enabled: volumes, error rates, and review outcomes, not enthusiasm.
• Keep judgment, client positions, and sign-off with attorneys at every stage, including native. What moves is assembly. What never moves is accountability.
• Revisit placements quarterly. Workflows earn promotion, and some earn demotion.

Where judgment beats the tool

A maturity model can place a workflow on the path. It cannot decide whether the redesign is worth the disruption this year, in this practice group, with this bench and this client base. Sequencing is strategy, and strategy is what the partnership is for.

Governance that speeds the firm up, and measurement that keeps it honest

The principle

Governance here is a small set of operating patterns, each of which buys speed as well as safety. Client and matter boundaries decide what AI can reach. Access mirrors ethical screens and matter staffing. Approved sources, the precedent bank and the template library, decide what drafts may be built from. Provenance rides along, which is what makes attorney review fast. Verification duties are written where the reviewer works. Named sign-out points decide who says it leaves. Retention rules say what happens to prompts and outputs when the matter closes, in whatever way the client's guidelines require, and reuse rules say what the firm may carry forward as precedent and what stays behind as the client's. Then the second half, because governance without measurement is hope: baseline the six numbers this program answers to. Pitch and RFP turnaround. First-draft turnaround on recurring document types. Associate hours per matter phase against budget. Write-offs and write-downs by matter type. Review cycles per document before partner sign-off. Matter-budget variance, visible during the matter. Count program cost once, expect the value after an adoption lag, and present ranges the partnership can believe.

The trap

The trap comes as a matched pair. Governance as committee: a policy nobody opens and an approval queue that takes a week, which people route around, leaving the firm with the risk and the bureaucracy at once. And measurement as marketing: hours saved that nobody redeployed, benefits counted twice, payback promised for day one. The billable hour makes lazy math worse, because an hour saved is only worth what the firm decides to do with it, and a business case that ignores that will die in front of the CFO, taking the true parts down with the false ones.

The checklist

• Build controls into the path of work as defaults, not into a policy document as clauses.
• Price every control by what it buys: speed gained or risk retired. A control that buys neither is not governance, it is decoration.
• Baseline every KPI before the pilot that is supposed to move it, with one owner and one source of record per number.
• Count shared program cost once, model the adoption lag explicitly, and present ranges with the assumptions attached.
• Name where freed hours land: more matters, alternative-fee margin, client development, associate training. In an hourly firm, unassigned capacity does not become profit. It becomes an argument.

Where judgment beats the tool

Patterns can be adopted; proportion cannot. How much boundary a routine engagement needs against a bet-the-firm dispute, when a control has become theater, and whether freed capacity should become growth, margin, or a saner pace for a tired associate class: those are calls about consequences and about the kind of firm the partners want, and no measurement makes them.