The View from Each Chair

For the associate, AI arrives as a paradox. The same tools that draft a first memo in minutes, summarize a deposition before lunch, and surface the controlling authority in seconds are precisely the tools dismantling the path by which earlier generations became lawyers. Research, first drafts, document review, due-diligence checklists — the unglamorous repetition that quietly trained judgment — is exactly the work most exposed to automation. The leverage model that historically justified large incoming classes is under pressure at the same moment the work those classes did is being absorbed by software.

The naïve response is to lean on the machine and ship faster. The strategic response is to recognize that speed is now table stakes and judgment is the scarce asset. The associate who thrives treats every AI output as a draft to be interrogated rather than an answer to be forwarded — checking citations against primary sources, stress-testing the reasoning, and catching the confident-but-wrong passage before it travels up the chain. The sanctions cases that made headlines are, concretely, failures at the associate’s layer of the work: machine-generated authority that no human verified before it reached a court.

Verification is becoming a craft in its own right, and it is a craft partners will pay for. Knowing how a model fails — where it fabricates, where it flattens nuance, where it confidently misstates a holding — is a form of expertise that did not exist five years ago and is in short supply today. The associate who can both generate work with AI and certify its correctness is more valuable than the one who can only do one or the other.

Fluency itself is, for now, a career accelerant. Associates who can configure tools, build retrieval over a practice group’s precedent, write effective prompts, and design small agentic workflows make themselves disproportionately useful while those skills remain rare. But fluency without substantive grounding is hollow: the goal is to be a lawyer who is excellent with AI, not a prompt operator who happens to sit in a law firm. The durable strategy is to become demonstrably excellent at what the machine cannot sign — synthesis, client-ready judgment, the relationship, the courtroom, the negotiation — and to let AI compress everything beneath it.

The partner reads the AI landscape through the lens of the book of business. The first questions are commercial: will clients demand AI-driven discounts now that they know a brief can be drafted in a fraction of the time? What do the client’s outside-counsel guidelines permit, require, or price? If a rival group is turning work around in forty-eight hours, has the client’s sense of what is reasonable already been reset — and is that group winning beauty contests on speed the partner cannot match without changing how the team works?

Underneath the commercial questions sits a quieter and sharper one: supervisory liability. When an associate’s machine-assisted draft goes out under the partner’s name, the duty of competence and supervision attaches to the partner, not the tool. The partner is the human in the loop whose signature converts a probabilistic draft into professional work product. That makes verification not a junior chore to be delegated and forgotten but a risk the partner personally owns.

Practice-area variance is the partner’s lived reality, and it is why firm-wide pronouncements about AI tend to ring false on the ground. What is true for high-volume, pattern-heavy work — fund formation, routine diligence, contract review — is not true for bet-the-company litigation or a novel cross-border deal. The partner has to calibrate where AI genuinely compresses cost without compromising the work and where it introduces risk that the client is not pricing.

Pricing is where the strategic pressure becomes concrete. As the production cost of certain work falls, the partner must decide whether to pass savings through, hold rates and absorb the margin, or move toward value- and outcome-based arrangements that decouple fees from hours. The defensible position is to price the judgment — the part clients cannot get from a vendor — rather than the keystrokes. Partners who can articulate why their counsel is worth a premium when the drafting is cheap will hold their franchise; those who only sell hours will find the hours repriced for them.

The managing partner owns the execution gap — the distance between a strategy deck about AI and what actually changes in how the firm works on Monday morning. The questions on this desk are unglamorous and decisive: total cost of ownership measured against demonstrable return; pricing strategy when AI compresses the very hours the firm has historically sold; the size and shape of the incoming class when traditional leverage assumptions no longer hold; and which practice groups should adopt first to generate proof points the rest of the firm will believe.

Benchmark pressure lands on this desk first. When a peer firm announces a nine-figure technology commitment or a flashy build, the managing partner is the one who must answer the partnership’s question — are we behind? — without mistaking a press release for a strategy. The discipline is to separate genuine capability from positioning, and to invest against the firm’s actual client demand rather than against a competitor’s marketing.

The most acute liability the managing partner manages is the training-and-policy deficit. Adoption in most firms is running ahead of governance: lawyers are already using AI, often through tools the firm never sanctioned, with no clear policy on confidentiality, verification, or client disclosure. Closing that gap — clear usage policies, mandatory verification standards, sanctioned tools, real training rather than a memo — is the single highest-leverage risk-reduction move available, and it is squarely this desk’s responsibility.

Talent strategy is shifting under the managing partner’s feet. The firm now competes not only for excellent lawyers but for legal-operations leaders, knowledge engineers, and technologists who can turn the firm’s institutional knowledge into systems. Lateral strategy, compensation structures, and career paths all have to flex to accommodate roles that did not exist in the partnership’s mental model a few years ago. And none of it works without change management: the hardest part of an AI program is not procurement but driving adoption past the early enthusiasts into the skeptical majority.

The chairman’s task is one of framing, and the framing determines everything downstream. Is this a tool cycle — a new generation of software to be evaluated, procured, and absorbed the way firms once absorbed email, electronic research, and e-discovery — or is it a restructuring of the industry’s underlying economics, in which the relationship between headcount, leverage, and profit is permanently altered? Leaders who misclassify the event will either overspend defending against a threat that was manageable or underspend against one that was existential.

Scale shapes the available moves. A firm with deep capital reserves can contemplate a large, integrated build — committing serious money to proprietary systems and absorbing the risk that some of it is wasted. A firm without that balance sheet may be better served by disciplined fast-following: letting others fund the expensive experiments and adopting what proves out, accepting a lag in exchange for avoiding stranded investment. The chairman has to be honest about which firm theirs actually is rather than which it wishes to be.

The decade-scale risks are genuinely symmetrical, which is what makes the seat uncomfortable. There is the risk of being the leadership that missed it — that watched a competitor reset client expectations and rebuild the cost base while the firm protected a model that was quietly eroding. And there is the risk of being the leadership that overspent on it — that poured capital and partner goodwill into systems the market commoditized eighteen months later. Neither error is recoverable on a short timeline, and both are visible only in hindsight.

Finally, the chairman is the custodian of the firm’s identity. AI strategy is not only an operations question; it is a statement about what the firm believes it sells. A firm that defines itself by judgment, relationships, and institutional knowledge will adopt AI differently from one that competes on throughput and price. The chairman’s job is to ensure the technology strategy expresses the firm’s identity rather than quietly replacing it.

The executive committee confronts a structural mismatch that no amount of enthusiasm dissolves. Serious AI investment is multi-year and, at the top of the market, can run to nine figures — but the partnership is a vehicle that distributes its profits annually and decides by consensus. Asking a roomful of partners to forgo current distributions to fund a system whose payoff is years away, and to do so unanimously, is a governance problem before it is a technology problem. It is one of the clearest structural reasons the largest, best-capitalized firms enjoy an advantage: they can absorb a long investment horizon that a smaller partnership cannot put to a vote.

The committee’s distinctive duties are capital allocation and enterprise risk. Where individual partners optimize for their own books and the managing partner runs execution, the committee must weigh a long-dated technology bet against the partnership’s appetite for retained earnings, debt, and risk. That requires treating AI investment with the same rigor applied to office expansion or a major lateral acquisition — staged commitments, clear milestones, and the discipline to stop a program that is not delivering.

Ownership terms have quietly become board-level issues. Who owns the intellectual property created when the firm’s lawyers train a system on the firm’s work? Who owns the data, and where does it live? What happens to the firm’s knowledge if a vendor is acquired, changes terms, or fails — is there an exit path, or is the firm locked in? These questions, once buried in procurement, now determine whether the firm is building a durable asset or renting a capability it can lose.

Above all, the committee governs the firm’s relationship with its vendors and its own data. The strategic prize in this market is the firm’s institutional knowledge; the strategic risk is handing that knowledge to a third party on terms the firm has not fully thought through. The committee’s job is to ensure that every major AI commitment protects the firm’s ownership of its most valuable asset.

The chief innovation officer lives inside the practical questions that the rest of the firm debates in the abstract. Vendor evaluation is a daily reality, and the gap between vendor claims and independent benchmarks is wide enough to be dangerous: the CIO has to cut through demos and marketing to understand what a tool actually does on the firm’s work, with the firm’s data, under the firm’s constraints. The pace of the market makes this harder — capabilities and pricing shift on a quarterly cycle, so an evaluation is never finished.

The central architectural choice falls to this seat: one integrated platform or an orchestrated best-of-breed stack. A single platform is simpler to govern and support but concentrates dependency and pricing power in one vendor; an orchestrated approach preserves flexibility and bargaining position but demands real integration capability and middleware the firm must build or buy. Either way, the CIO must preserve an exit path from every contract, because lock-in is the quiet way firms lose control of their own knowledge.

Getting firm knowledge into systems without breaching client commitments is the work that turns strategy into advantage — and it is delicate. Precedent banks, playbooks, and matter history are the firm’s moat, but much of that material is encumbered by confidentiality obligations, ethical walls, and client-specific restrictions. The CIO has to build retrieval and knowledge systems that respect those boundaries by design, not by after-the-fact policy. This is why knowledge management has been promoted from a back-office function to a strategic core: in a world where the model layer is commoditizing, structured institutional knowledge is the durable differentiator.

Finally, the CIO owns the unglamorous discipline of turning pilots into production. The graveyard of legal innovation is full of successful pilots that never scaled, because adoption, training, support, and measurable ROI were treated as afterthoughts. The CIO’s mandate is to measure what the firm gets for what it spends, to retire what does not work, and to embed what does into daily practice rather than leaving it as a demo that impressed a committee.

The general counsel is the demand side in person, and that position confers unusual leverage. As budgets shift, the GC can reallocate spend from outside hourly work to internal AI capacity — bringing more work in-house because the marginal cost of routine drafting and review has collapsed, and reserving outside counsel for genuine judgment and risk. That reallocation is one of the most consequential forces reshaping the market, because it changes what firms are actually paid to do.

The GC also writes the rules that bind everyone else. AI clauses are migrating into engagement letters and outside-counsel guidelines: disclosure of AI use, verification standards, restrictions on training third-party models with the client’s confidential information, and allocation of liability when machine-assisted work goes wrong. In effect, the GC is regulating by procurement — imposing, through contract, the standards that no statute or bar rule yet requires. Where regulators are slow, the buyer of legal services is fast.

Inside the organization, the GC increasingly owns enterprise AI governance writ large — not only for the legal function but as the executive most fluent in the risks AI poses to privilege, privacy, data protection, and regulatory exposure across the business. That makes the GC both a consumer of legal AI and the steward of how the entire enterprise deploys AI responsibly.

Selection of outside counsel is shifting accordingly. Early evidence suggests clients are beginning to choose firms partly on AI capability — on whether a firm can deliver the speed, consistency, and cost structure that AI makes possible while preserving the judgment the client is actually paying for. The GC who understands the landscape can demand more, pay differently, and hold counsel to standards the profession has not yet imposed on itself.

In a single recent week, four legal-technology companies named new chief executives and a fifth installed a new president. Read individually, these are ordinary corporate announcements. Read together, they are a signal: the legal-technology industry is reorganizing its own leadership around AI — repositioning for a market in which artificial intelligence is the product, not a feature.

That signal raises a sharper question for law firms than “who now runs those companies?” It is this: if the vendors are restructuring their leadership for an AI-first market, who is building the equivalent capability inside the firm? Most firms answer by default rather than by design, reaching for one of three imperfect options.