The summer of our discontents

Two months ago, if you prompted Version 3 of the AI-art generator MidJourney to generate depictions of an “otter on a plane using wifi,” you were rewarded with the nonsense in the left panel of our lead graphic. A month later, Version 4 could take the same prompt and render, in seconds, multiple detailed drawings that are likely beyond 80% of the population’s imagination and certainly beyond 99.9% of the population’s acumen at illustration (above right panel).

Imagine what our new year will bring.

This matters. And we shall return to our wifi-enabled Mustelidae further down.

This lengthy essay has a lengthy preview essay authored by CSO Casey Flaherty. See Post 347. These two essays reflect nearly everything we are learning through our industry meetings. Although the act of writing is a crucial step in crystalizing our thinking for ourselves and our clients, we’ve done our best to make these essays enjoyable for readers.

2022 was a very good year, and 2023 looks even better (for us)

Canby Wood

LexFusion operates at the intersection of enterprise, legal services, and innovation. Market listening and trust are core to our unique value proposition. See Posts 203267 (exploring the LexFusion model). Business is booming. We recently brought on a superstar Global Director of Litigation Solutions in Canby Wood. And we have now opened the search for her transactional counterpart—if anyone knows an uber-connected innovation obsessive with a transactional background, send them our way.

LexFusion, however, benefits from self-selection bias. Our sample is skewed. Our success is not predicated on unanimity, a majority, or even a plurality. We thrive at the center of the edge. For our small company, with its outsized reach, prosperity depends merely on the existence of innovative outliers—of which there are many, relatively, in raw numbers.

In 2021, we heard the hopes, dreams, and fears of 327 law departments and 240 law firms. We analyzed market sentiment in Post 280, where we celebrated the excitement in the ecosystem while lamenting the cultural conditions that often frustrate the attendant ambition (also ice zombies).

As much as we hate to admit it, our conclusions have only hardened. In 2022, we conferred with 435 law departments and 250 law firms. We did not merely have more meetings. We had deeper conversations. We executed NDAs with multiple law departments and law firms so we could dig beneath surface-level discussions of practical innovation into the painful realities of budgets and politics. We uncovered far more chronic pain than even we anticipated.

What follows is a rather literal year in review. We frequently write follow-on summaries of meetings for our customers’ own reference—contributing to the decks/memos they are preparing as part of internal pitches and value storytelling. We’ve remixed that content here, excising all identifying information. We’re sharing some of what we have been telling law departments and law firms. Importantly, our insights mirror what we have been hearing from law departments and law firms.

What may come off as criticism is, in fact, a reflection of our customers’ lived experience. While we may help them refine their vocabulary and advance their thinking, the discontent being conveyed is theirs, not ours. Broadly, they know what needs to be done. Depressingly, they recognize most of it will not happen, regardless of how righteous their cause and how superhuman their effort. The inertia of immediacy has a preternatural win rate. See Casey Flaherty, “Maybe, Don’t Be MacGyver,” 3 Geeks, Sept 12, 2021; Casey Flaherty, “Scary Stories About Our Wicked Problems,” 3 Geeks, Oct 31, 2022.

If you prefer to skip ahead to the otters, we offer a TLDR version of the long middle section, which very much builds on Casey preview essay (347):

Short-term easy is long-term hard. Long-term easy is short-term hard.

Legal organizations (departments and firms) should calculate what percentage of their total spend is directed to projects that will progress their ability to deliver at scale—i.e., the leveraging of expertise through process and technology such that an increase in work does not require a proportionate increase in human labor.

In most legal orgs, this percentage is near negligible, especially if the org is being honest with itself about (i) how many personnel in putative innovation roles (legal operations, knowledge management, project management) are consumed by active matters, existing programs, and administration, (ii) how much technology spend is maintenance, and (iii) how many projects are purely aspirational with no real resources save the illusory spare hours of already busy people.

Anyone interested in shifting this equilibrium should contemplate the politics required to boost the resources allocated to real innovation to 1% of total budget. Which stakeholders would need to assent? Which stakeholders would need to affirmatively contribute, including expending political capital to achieve, and maintain, consensus?

In most legal orgs, these politics present as impossible without an external, existential threat. In the case of law departments, active and sustained C-Suite intervention. For law firms, a clear and sustained change in buying behavior by a critical mass of clients.

How much would this dynamic recalibrate if a truly transformational technology emerged? See Casey Flaherty, “My legal tech innovation: The Magic Money Machine.,” 3 Geeks, Feb 9, 2018. How far and how fast could your legal org move without an external, existential threat?

The bitter truth is most legal orgs would move neither fast nor far. The (i) requirement for consensus combined with (ii) key stakeholders being too busy to reach, let alone act on, consensus is more than sufficient to delay the supposedly inevitable. Add in the (iii) practicalities and politics of fiscal friction, and there are more than enough structural barriers to change without ever citing bad decisions or bad faith.

Which is not to say, nothing changes. Rather, it should be unsurprising that change is episodic and insufficient. Our organizational orientation services the short term. We should therefore recognize that the long term will become increasingly challenging.


A quick aside on law firms

The composite that follows is written with law departments as the primary audience despite incorporating much of what we’ve expressed to law firms. The single POV is aimed at coherence and ease of consumption. But it is also an acknowledgment of positional authority.

Clients have always been the channel captains and urgency drivers. But, increasingly, the locus of corporate legal activity has shifted in-house, both in terms of bodies (see Post 262) and money (54% of spend is now in-house per the ACC 2022 Law Department Management Benchmarking Survey).

It is a buyers’ market. Clients get the law firms they want—even if these are not the law firms law departments say they want.

A founding mantra of the in-house revolution is “we hire the lawyer, not the law firm.” David B. Wilkins, “The In-House Counsel Movement, Metrics of Change,” The Practice, May/June 2016. This is the mantra of individualism, both internally and externally. It elevates individual in-house counsel as the arbiters of the worthiness of individual external lawyers while also exempting all parties from scrutinizing the accompanying infrastructure—i.e., how individual and collective expertise are leveraged through process and technology.

In an environment where clients hire lawyers, not law firms, the rational response from law firms is to allocate their marginal dollars towards attracting and retaining the lawyers clients hire. Lateral frothiness. The abandonment of lockstep. The growing compensation spread within the shrinking equity partner ranks. Boom-and-bust associate hiring/salary frenzies. All explicable reactions to law department buying behavior.

In an environment where clients hire lawyers, not law firms, the rational response from the lawyers clients hire is to maximize their annual take-home pay while keeping their book as portable as possible. This includes resisting any investment of (their) money into infrastructure that makes the firm stickier for themselves or their clients. They need not be self-aware Machiavellians in this regard. Rather, they merely need to be laudably client centric—always consumed by client work or client development—and therefore too busy to be consulted. When major investment decisions arise, they can genuinely object that they were not consulted and are not comfortable approving expenditures they do not understand.

Jae Um

We can’t recommend enough Bill’s recent series, and the included commentary by our beloved advisor Jae Um, on law-firm dynamics, see Posts 330, 331, & 335). One inescapable conclusion is that Everyone Else (the non-premier firms) should pursue service-model innovation. But service model innovation is necessary and hard across the board, including for law departments. Service model innovation is particularly challenging for law firms when their clients will not permit them to change.

This dynamic will persevere as long as current client buying behaviors persist. Thus, it is the clients to whom we now turn.


Fine, you must save money, immediately. Cost discipline is essential business hygiene. But it becomes more acutely essential during periods of economic turmoil.

No matter what we say, you will likely feel compelled to do the whole performative discount thing with your law firms. We implore you to try to avoid governance by fiat and, at least, engage in dialogue with your primary firms around time-boxed, mutually beneficial commercial arrangements that satisfy whatever mandates you face.

More productively, we urge you to seize the opportunity to explore more impactful and enduring changes to your buying behavior that can also deliver immediate fiscal results:

  • Package work. Identify opportunities to enter portfolio arrangements, including integrated law relationships with New Law offerings.
  • Move work. Right source, including greater use of legal marketplaces to find the right talent at the right price.
  • Re-examine costs on autopilot. Major advances in e-discovery, ADR, staffing, etc. present substantial, immediate spend-optimization opportunities.

While we’re happy to help with the above, we recognize it likely seems daunting given your timelines, bandwidth constraints, and the stifling politics of doing anything differently–which is a great segue into our broader point that change management is a lie.

The lie of change management. Change management is a fine discipline. Applying change-management best practices is beneficial. But there is a lie at the heart of the common discourse around change management. It is a disservice to perpetuate the myth that, with good change management, all change is possible.

Absent the proper environmental and cultural preconditions, many changes will simply never happen. These preconditions often involve structures, systems, politics, and resources beyond the purview of the would-be change agent.

In particular, leadership buy-in is a pre-requisite, not an outcome—despite what one may read about managing up, managing your manager, influencing up, leading without authority, etc. And it is buy-in, not “lip service.” There is a price to be paid.

Leadership is not a person in a position of authority remarking in the abstract about the desirability of a particular outcome. Expressing the expectation that subordinates find a way to forge and maintain consensus—so no political capital need be spent—is a leader’s subtle way of informing everyone, including skeptics, they have not bought in.

Leadership entails exercising authority to the extent required to achieve the outcome. The exercise of authority is required for real change. Not everyone will come along willingly even with excellence in empathy, communication, the art of the business case, value storytelling, planning, stakeholder engagement, listening, pivoting, execution, measurement, transparency, etc.

Not all change is possible. Thus, choosing which changes not to pursue is mission-critical. This demands reasonable clarity as to critical paths (which stakeholders must be on board, and to what degree) and constraints (authority, attention, time horizons, money, immovable objects).

A bad outcome is energy wasted accomplishing nothing. A worse outcome is energy consumed by innovation theater that feeds the illusion that something is being done. The worst outcome is path-of-least-resistance comprises that result in energy expended moving in the wrong direction—sprinting the wrong way is regression, not progress.

The lies we tell ourselves about our commitment to change. One-million dollars is life changing for most people. It is 2,261% of the median American income. One million dollars does not move the needle at most large corporations. It is 0.002% of the median Fortune Global 500 company’s revenue. One million dollars is to the median Fortune Global 500 company as one dollar is to the median American. Our minds are rather bad at appreciating such orders of magnitude–indeed, Casey’s preview essay (347) suggests that the entire legal ops movement is at risk of getting this wrong.

There is a natural inclination to speak rather breathlessly about how much an enterprise is spending on this or that. When put in raw terms, the numbers are substantial. When translated into percentage terms, they often become microscopic. Percentages place numbers in context.

Often, context is absent when discussing organizational investment in change. Sometimes, this entails using raw numbers instead of percentages. But, more frequently, it involves avoiding numbers entirely. Instead, we are subject to the recitation of long lists of in-flight projects, technologies under consideration, and aspirational goals. These are smashed together, conflated, and presented devoid of context to support the general proposition that a team is “already doing a lot” on the innovation front.

They probably are doing a lot, relative to available bandwidth and fiscal resources. But if we broaden the context so the investment is situated relative to other expenditures, the size of the problems being addressed, and the expected return on investment for the business, we find most change efforts border on trivial.

Speaking in, and comparing, percentages surfaces actual priorities. For law departments and law firms, the priority is legal labor.

Spend optimization and scale—two paradigm shifts that probably will not happen but should. Generally, the only reward for coming in under budget is a lower budget in the future. Excepting Veblen goods, everyone would prefer to spend less money on everything while getting more for each dollar spent. Corporations would like to spend less on legal. Law departments would like to spend less on law firms. We all want things. But life, unfortunately, demands tradeoffs.

What law departments and law firms are not currently positioned to do well is meet the legal needs of business at scale and pace. Legal services remain labor centric. Labor is not only expensive, it’s linear. The relationship between corporate legal needs and the attendant demand for legal labor is largely static. When legal needs increase (which they almost always do), then so, too, does the attendant demand for legal labor. Thus, starving a law department of resources generally results in corporate needs going unmet, in the long term.

In the near term, fiscal shortfalls can be addressed through savings. But most instances of “savings” are one-time lifts because they are different flavors of labor arbitrage. Discounts are about paying less per hour. Insourcing is mostly pre-purchasing hours in bulk at a low(er), fixed price point. Same with most ALSP and shared-services plays, etc.

There is nothing wrong with labor arbitrage per se. Right sourcing is imperative. But labor arbitrage does nothing to fundamentally bend the cost curve. While savings efforts may result in the mixed cost of legal labor being temporarily reset to a lower baseline, the linear relationship to business needs remains.

Centering “savings” as the law department’s mission results in short-term glory and long-term pain. It amplifies the hyper-palatable partial truth that the enterprise can, and should, spend less on legal. Long term, the pure-savings narrative only leads to chronic underfunding of the law department and, most importantly, business needs going unmet. See Post 347 (preview essay).

Thus, the first easier-said-than-done hurdle for law departments is to graduate from cost center to value center, shifting the narrative from savings to spend optimization. Spend optimization is still concerned with maximizing the yield from every dollar. But spend optimization is not only focused on spending less, it also accounts for where and how the resulting savings are invested.

Savings should be invested in scale, the second easier-said-than-done hurdle. Scale is about decoupling business needs from legal labor, such that increased business needs can be satisfied without a proportionate increase in legal labor. Very few law departments or law firms are making meaningful investments in scale.


Segregate extraordinary spend. You need a clear view of where you are currently spending money. Eventually, you will also need to figure out why.

Extraordinary spend encompasses massive, discrete, outlier disputes, investigations, transactions, etc. These are the kind of big-ticket items on which there is constant communication with the C-suite. The budget is often a separate line item (bc extraordinary). Prominent 10-K, annual-report-level type one-offs. Such anomalous matters are not particularly informative for long-term planning purposes and can be so material they skew any attempt at analysis. Extraordinary spend can, and should, be addressed directly—just separately.

Consolidate ordinary spend (if you can). When law departments fail to meet the needs of the business (and they frequently do), the corporate instinct is to create rival, quasi-legal silos (e.g., compliance, privacy, government affairs, regulatory affairs, tax) or even spilt the law department in two outright. Some adjacencies are so significant from a business perspective they warrant their own function (often larger than Legal itself). But, consistent with the scholarship on complicatedness, see Post 347, these new functions are usually layered on (additional stakeholders, procedures, chokepoints) rather than integrated/aligned. This layering increases friction and reduces business velocity. It also obscures true “legal” spend and frustrates attempts to bring rigor thereto.

Divide ordinary legal spend between “plug the dikes” and “build the dams.” Headcount should be translated into dollars (fully loaded). And dollars should be translated into percentages of total budget. At massive organizations, large raw dollar figures can represent minuscule percentages and trick the mind into believing the org is making a sizeable investment when it is not, on a relative basis.

  • Plug the dikes is work that increases in a linear fashion with the needs of the business. You, of course, still need to understand the nature of the work itself (work sorting), including the composition of the work (work decomposition), the business drivers of work volumes (work drivers), and the value of the work to the business (work segmentation).
  • Build the dams is work that enables scale—i.e., reduces the demand for legal labor relative to business needs. This is project work, a key theme of last year’s Review, see Post 280, that should be as upstream as feasible (i.e., compliance by design). But leveraging legal labor through process and tech remains necessary; underestimating the return on reducing low-end friction is a common mistake.

In most organizations, a candid assessment is likely to reveal that almost all current spend is dedicated to plugging the dikes. Thus, if business needs increase, legal resources must increase proportionately or business needs will go unmet in some form or fashion. That’s the math. And while modest decreases in labor costs can offer temporary relief, labor arbitrage cannot fundamentally bend the cost curve long term.

Skeptical stakeholders are not wrong when they suspect that taking attention away from plugging the dikes in order to build the dams will result in drowning long before the dams are complete. Our dour perspective is that some drowning is inevitable, one way or the other, and building the dams is, on net, more critical to the long-term health of the business.

Further categorize ordinary spend. Again, headcount translated into dollar figures, and then dollar figures translated into percentages. While there may be some baby splitting (multiple hats), the key is to allocate individuals according to how they actually spend their time—aspirations are non-pertinent.

Here are seven categories of ordinary spend. It’s a mistake to skip even one.

  1. High-end, embedded, internal advisory. These are lawyers with valuable expertise deeply embedded within the business. They are in the room where it happens—where the most impactful business decisions are made. The business considers them part of the leadership team and integral to making better decisions, faster. This is the category where it will be most tempting to fudge (many in-house lawyers would place themselves here; most, though not all, would be wrong). Do not succumb to temptation.
  2. Core internal personnel. This is the high-volume legal work, as well as departmental operations. Contracting. Marketing reviews. IP. Standard litigation. Routine advice. External resource management. Program administration. Et cetera.
  3. Traditional matter-level external spend. Matter-by-matter assignments to law firms with fees calculated at the matter level. Includes panel firms, discounts of any flavor, matter-specific fee arrangements (no matter how alternative), etc.
  4. Structured external arrangements for substantive legal work. Portfolio partnerships and managed-service relationships with law firms and New Law providers, as well as the use of legal marketplaces. A law firm exclusively handling an entire tranche of work (i.e., the entire portfolio) is not automatically a portfolio partnership unless the pricing has been negotiated at the portfolio level (i.e., the price is for the portfolio, as opposed to rate concessions predicated on being awarded the portfolio).
  5. Core infrastructure. Tech, support, maintenance, etc. Already committed and already implemented. Includes SaaS.
  6. Other external. This is a catch-all category, which does not make its content unimportant. These expenditures (sometimes routed through outside counsel) can represent significant outlays. Staffing. Electronic discovery. Mediation & arbitration fees. Court reporting. Subscriptions. Non-substantive managed services (e.g., first-pass review of outside counsel bills). Often, these costs have been on autopilot for years and can therefore offer some relatively quick, sure-footed first steps towards spend optimization, presuming the savings are put towards meaningful investments in scale.
  7. Project resources. These are resources dedicated to building the dams (i.e., sustainable scale). Projects involve a series of planned activities designed to generate a deliverable (a product, a service, an event). These activities—which can be anything from a grand strategic initiative to a small program of change—are limited in time. They have a clear start and end; they require an investment, in the form of capital and human resources; and they are designed to create predetermined forms of value, impact, and benefits. Every project has elements that are unique. That’s key: Each contains something that has not been done before. Here, too, the temptation to fudge will be strong. A headcount with “legal ops” in their title is core personnel if most of their time is spent administering e-billing or some other standard system/program. Same for a “project manager” whose primary activities involve maintaining systems/programs or supporting core personnel performing plug-the-dikes work. This is particularly true of in-house lawyers who may be listed as stakeholders on 74 project plans but never have time to contribute to any of them because they are so crushed with “real work.”

In most departments, most resources will be dedicated to core personnel and traditional external spend. Thus, an upsurge in demand will result in overflow captured by law firms until the law department “saves” money by insourcing what it can. This savings noise distracts from the fact that the response remains fundamentally linear in nature.

Press most law departments on their spend-optimization efforts, and they will point to (i) purportedly aggressive external cost containment programs, (ii) insourcing, and (iii) a long list of projects, including a grab-bag of tech under consideration. Depressingly, the department may have more projects than people. Yet these people have no time for projects, most of which are somewhere between aspirational and unserious. This is identical to what we discussed in last year’s Review. See Post 280.

Spend percentages are a more reliable signal of seriousness than lists, decks, plans, and target operating models. If only a minute percentage of spend is dedicated to projects, then few projects will be completed, and little will change.

Despite that dismissive comment about target operating models, the department should have one, as well as the roadmap to get there. Most models, however, suffer from a decided lack of specificity (more of the same, vaguely better, #innovation). Most roadmaps, therefore, present a similar absence of focus and patience.

The result is disparate projects that throw insufficient resources at unachievable goals on unrealistic timelines. While underpinned by the best of intentions, unserious projects are distractions that consume finite resources (there are still meetings and, more often, endless email exchanges regularly rescheduling meetings). Unfortunately, some projects are in fact semi-serious and achieve the status of folly—consuming considerable resources but delivering negligible-to-negative ROI because of poor planning, execution, or, most frequently, follow through.


It depends on context. Organizational context is key. It is your job to master organizational context. This demands more than being able to explain to outsiders why “that won’t work here.” Real change requires figuring out what will work and then making it work, including proper sequencing. What follows is general advice that will only be useful if meticulously tailored to the specific context. We know you want a fish, but all we can offer is a worm.

Make extraordinary spend less extraordinary. Most law departments need more standard mechanisms for formally segregating extraordinary spend. Otherwise, they are reliant on the fidelity of short-term institutional memory when periodic reviews surface instances of blown budgets. Many well-run departments appear profligate due to some massive matter(s) they could not have possibly anticipated or controlled. They also see their priorities derailed by the unexpected and then lack the audit trail necessary to validate why they are under-resourced.

This is not to suggest prevention is impossible, let alone that matter management is useless. The former is a major topic below (i.e., embedded advisory and compliance by design). With respect to the latter, we know of seven- and eight-figure cost reductions on matters just because someone asked for a budget (not even with an eye towards savings, simply for the purpose of understanding the matter plan). Extraordinary ≠ blank check. There should be a programmatic approach to external matter management that becomes more rigorous and bespoke as size, scope, and impact increase—emphasizing total cost of ownership, outcomes, and ROI.

Cream work will remain mission-critical and expensive (external). Cream work is one step down from extraordinary. More common. Slightly less material. Still periodic and bespoke (i.e., mostly not systematizable from an in-house perspective) with substantial business impact (high ROI). For too long, too much work has been considered cream in order to sidestep scrutiny (and, as a result, too many law firms have been treated as sacred cows). But accurate as this observation may be, it does not obviate the fact that premium work still exists.

Source: Jae Um, “Kings of the Hill,” Am Law (May 2022) at 48.

Parroting the sharpest mind in legal, Jae Um, Premier League law firms remain excellent at fulfilling their original purpose. They have deep benches of pedigreed lawyers across specialties who can be assembled into ad hoc tactical teams to address intermittent business needs that involve extremely complex questions in high-consequence matters.

Oversight should not be absent. But the focus should be on consistent quality of outcomes (malpractice claims skyrocketed during the recent M&A boom) at market prices (much harder to determine than one would think) rather than achieving microscopic, mostly performative savings (relative to business impact).

Top-end advisory is the crème de le crème (external). Top-end advisory is a small subset of cream. Invaluable, niche expertise. Required irregularly. Universally recognized (by the business) as having extraordinary yield. Largely price insensitive. Though we concede cost discipline is essential, no one really cares whether tax advice that saves $5 billion (with a “b”) per annum comes in at $1,800/hr or $2,200/hr—because the ultimate aim is to price the work, not the lawyer, based on business value. Indeed, group this under “extraordinary spend” if feasible, except it will rarely meet the materiality threshold.

Embedded advisory should become a focus (internal). When top-end advisory starts becoming regular, it should become embedded, if possible. These roles will likely never be numerous (on a relative basis), even at the largest orgs. But the business impact is considerable, and the attendant relationship with leadership is instrumental in changing the narrative around the department. Consistent with the understanding that headcount will be capped, the bias should be towards filling strategically vital roles rather than insourcing routine work, despite the latter being easier to effect and the savings easier to measure. Almost all legal work benefits from proximity to the business (less friction). But the real question, given headcount constraints, is “where does the business benefit most from proximity to legal?” Embedded advisory is the answer.

Solve for scale complexity, orienting towards compliance by design. This is the core struggle, literally. For completely comprehensible reasons, short-term demands get in the way of the investments necessary to sustainably drive superior business outcomes at scale and pace—better decisions, faster and more consistently, in an increasingly complex operating environment.

While the industry constantly invokes “people, then process, then technology,” there is scant evidence that people are being deprioritized. Much the opposite. Ultimately, leverage through process and tech is the whole ballgame when it comes to tackling scale complexity—i.e., the high-volume core work.

Compliance by design is a system-level solution that involves moving upstream and embedding legal knowledge into business processes (i.e., de-lawyer without de-legaling) to enhance outcomes and velocity. When business process throughput increases, the quantity of interactions with legal professionals should not (or, at least, in a far less linear fashion than today).

Compliance by design may not always be feasible, especially in the near term. There will often be intermediate steps, like process redesign and tech enablement, that deliver incremental improvements on the path to compliance by design.

Compliance by design and the intermediate steps are project work. There is nothing unique to law departments in their struggle to allocate sufficient resources to projects. Projects are disruptive. Projects not only consume the same bandwidth needed for operations; projects interfere directly with operations because their express purpose is to change operations—transform how work is done.

Almost no one is good at solving for scale. If you dig beneath the surface at the most lauded law departments, you will find the resources devoted to solving for scale are negligible. So, too, is the impact thereof. The success stories are quite real; they’re just small relative to the size of the problem (again, no reason to believe this is peculiar to law departments).

Be willing to automate so many tasks your efforts eliminate some roles in their current form. This is where most people get off the bus. While the framework is to automate tasks, not jobs, eventually this results in some jobs being so materially transformed that they become different in kind, not just degree. It is all well and good when we are talking about making everyone more efficient through tech. Few, however, are prepared to follow this logic to its natural conclusion.

Source: Peter Ungphakorn, The 101

The savings trap is not merely that the Red Queen’s Race offers no respite, let alone any offramps. It is that, in the quest to realize near-term savings, law departments insource routine work. Easiest to insource. Simplest savings calculation. But also the work most amenable to process improvement and automation. Thus, it is law department personnel who are now subject to the greatest threat from solving for scale.

Unfortunately, personnel can be difficult to repurpose. Meanwhile, their survival instincts are well honed—every in-house hire is another potential impediment to real change. And pursuing projects that negatively affect the livelihood of colleagues is painful. Understandably, the most common choice is simply not to do so.

This dynamic, btw, is one of the rationales behind the corporate allergy to headcount. See Post 347 (preview essay).

Stop insourcing routine work even if it will save money short term; outsource routine work to free up headcount for strategic roles. Automating work is far less fraught when the work is being handled externally. This is true in the compliance-by-design sense of eliminating touchpoints with legal professionals as legal knowledge becomes embedded in business processes. But it is also true from a process-improvement/tech-enablement/right-sourcing perspective.

The wallet can be a much more potent change instrument than the retail politics of effecting change inside a law department already drowning in work. But this presumes the wallet is being deployed properly. Sophisticated spend management requires projects that eventually become programs.

Where possible, package the work, and price the package. Portfolios are the proper level of resolution for many large tranches of legal work. The transition to portfolio partnerships can offer both near-term and long-term benefits. Portfolios can be structured to continuously improve outcomes, speed, predictability, consistency, and data quality while reducing unit cost from a total-cost-of-ownership perspective.

On the litigation side, law firms are the primary portfolio partners, augmented by a programmatic approach to managed services (like e-discovery) and other associated costs (like ADR). On the transaction side, there is likely to be more of a mix (and even combination of) law firms and New Law (i.e., ALSPs) with the objective in many instances being integrated law (a topic that requires its own post).

Paying bottom dollar is not the goal. But, consistent with spend optimization, cost-effectiveness is. We must avoid the savings-filter where an option is considered “better because it is cheaper” in order to achieve a deeper understanding of how and where an arrangement can be “cheaper because it is better.” Portfolio arrangements can shift us in this direction, in part, because they are material enough to merit the appropriate level of sustained attention, from both sides.

Yet, packaging is not always possible, especially with the onslaught of novel issues introduced by net new regulatory complexity. In re-thinking non-cream work, law departments should not only reconsider their law-firm mix (are you deluding yourself with discounts? are you paying premium rates for non-premium work?), they should also look to mechanisms like legal marketplaces to expand their options, reduce administrative burden, and enhance pricing rigor (another topic that demands its own post).


Admittedly, all easier said than done. It is quite understandable if much of the above proves infeasible. Accomplish what you can. But, more importantly, avoid wasting precious energy pursuing the unobtainable.

The current orientation of most law departments is that of cost centers trying to manage traditional matter-by-matter law firm relationships through discounts (and their variants) while insourcing as much work as permitted to fulfill their more-with-less savings mandate.

The proposed orientation is that of value centers seeking to optimize spend in the service of business value with an emphasis on (i) embedded advisory, (ii) compliance-by-design and other scale-enhancing projects, and (iii) a programmatic approach to the external value chain, moving more work into portfolio partnerships and diversifying sourcing (law firm mix, New Law, legal marketplaces).

The proposed orientation tends to elicit plenty of nodding agreement. But theoretical support for change is meaningless without congruent actions, including a material shift in resource allocation. The latter is rare, for comprehensible reasons. Indeed, endless discussion and highly abstract agreement around change are among the status quo’s greatest allies.


For those who took that almost 5,000-word tour summarizing what LexFusion heard this last year, thank you. We hope it merited your time and attention. We assure you we will bring it back around.

For this coming year, we have a bold prediction:

In 2023, AI will be capable of generating, near instantly, a legal opinion or contract superior to the work product of 90% of junior lawyers.

We are convinced. We’ve had sufficient exposure to transformer-based neural nets through our business relationship with Casetext’s AllSearch to form strong opinions (lightly held) on what might become possible. But we recognize—and understand why—various contingents vehemently disagree. While who is right or wrong has profound implications, we have a more modest prediction that will produce some of the same outcomes and is likely to garner sign-off from even the most well-informed skeptics:

In 2023, 90% of the population will believe AI is capable of generating, near instantly, a legal opinion or contract superior to the work product of 90% of junior lawyers.

That is, we can disagree on the trajectory of the tech while coalescing around likely perceptions thereof, especially with the avalanche of buzz being generated by ChatGPT.

ChatGPT is a driver and a distraction

Almost all the current chatter centers on what ChatGPT is and is not capable of, today.

We, however, opened the piece not with ChatGPT but with a little-known AI-art generator, MidJourney, and its one-month progress from useless to impressive when prompted to illustrate “otter on a plane using wifi.” Midjourney is representative of an entire generative AI landscape where rapid progress is being made—OpenAi’s ChatGPT’s is but one player. Sam Altman, OpenAI’s CEO, has himself referred to ChatGPT as merely a “preview of progress.”

The important question is not so much “What can ChatGPT do today?” A better question is “What will emerge from this constellation of companies, and new entrants, in the relatively near term?”

Source: Base10

Moreover, even the focus on the “generative” aspect of AI distorts the productizable potential. Generative capacity is cumulative— the output of collation, categorization, connection, analysis, synthesis, and production of content based on enormous data sets. Each of these foundational steps is a capacity in and of itself. Anyone who has spent time in legal tech, for example, is painfully aware of the challenges in adding structure to unstructured data—any advancements on that front extend the shadow of the possible, whether or not AI ultimately uses the more structured data to generate first drafts.

Again, our deepest exposure is to Casetext’s AllSearch. As the unfamiliar can infer, it is a search tool. Casetext trained transformer-based neural nets on legal language to develop the most advanced case law search tool ever (and Parallel Search is next-gen, the first true concept-based search). But Casetext also quickly realized neural nets trained on legal language enable applications far beyond searching case law—e.g., searching contracts, brief banks, deposition transcripts, knowledge management, prior art, etc.

Casetext is not on the above map of the generative AI ecosystem—and we doubt Base10 will ever include a “legal” box. Rather, Casetext is a node, working on domain-specific applications. They bring technical acumen, a massive corpus of legal language to enrich the general large language models, and expertise to train the models—reinforced learning from human feedback is critical.

From the perspective of integrating new tech into the corporate legal market, the answer to the question “What can ChatGPT do today?” is interesting but not that informative. A better question would be, “What will emerge from the constellation of generative AI companies and other new entrants?” But even that answer would lack sufficient specificity.

A potential formulation of a more salient inquiry:

In the relatively near term, what is likely to emerge from the domain-specific application of these multi-modal advances in tech as they are enriched by legal language and reinforced learning from expert feedback by various market participants seeking to address specific points of friction?

We’ve seen this movie before (Part 1)

Larry Summers, former Secretary of the Treasury and president of Harvard, went on BloombergTV to tell the world ChatGPT is a development on par with the wheel and fire. And then he signal boosted himself via tweet.

Such hype is not new. In 1970—fifteen years before OpenAI CEO Sam Altman was born—Life magazine proclaimed:

In from three to eight years we will have a machine with the general intelligence of an average human being. I mean a machine that will be able to read Shakespeare, grease a car, play office politics, tell a joke, have a fight. At that point the machine will be able to educate itself with fantastic speed. In a few months it will be at genius level and a few months after that its powers will be incalculable.

In 2016, we chronicled the history of these wild pronouncements in parallel with the career trajectory of a successful lawyer who had finished law school in 1977—at that time, 1977 was the median and modal graduation year of the chairs of the Am Law 10. See Casey Flaherty, “Real Lawyers v. Cyborgs,” 3 Geeks, Feb 19, 2016. We recounted how, every few years, the lawyer could read something about being replaced by AI. Or they could ignore it. Because this never happened. Instead, the failure to live up to the hype resulted in repeat AI Winters—before the cycle would begin anew.

Throughout our history, we compared the hyperbolic statements re AI overlords to the far more mundane reality: the rise of PCs, the internet, email, and smartphones. We concluded by observing that the mundane had slowly transformed reality—the story reads very differently if you excise the hype of everything everywhere changing, in unimaginable ways, all at once. The introduction and maturation of technology has truly changed how lawyers work.

Source: A webcomic of romance, sarcasm, math, and language.

Going forward, we expect a deluge of nonsense hype counterbalanced by ample skepticism, some well-informed and much of it wildly uninformed. There will be impressive progress (like the otters) to bolster the bulls and spectacular failures (like Meta’s Galactica) for the bears to mock.

Indeed, to return to our original markers, there is quite a distance to be traversed from (i) what the tech is capable of and (ii) actually productizing the tech in a manner that can be integrated into a coherent workflow. See also annotated version of Susskind’s five stages. There will be a rush of activity, including many missteps. But the felt impact will not be so immediate nor universal.

The skeptics are correct. We are headed for peak hype. But we submit the advancements also represent crossing an inflection point consistent with Amara’s law: “we tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.” What is emerging today is result of many decades of work (including myriad disappointments) as several technology trends converge—we’ve long had the math and the models behind neural nets, but we lacked the computing power and the vast reservoir of digitized language.

Source:I.A. vs. A.I. – what’s the difference and why I.A. comes before A.I. (Part 1),” Lawtomated, Apr 10, 2019.

We’ve seen this movie before (Part 2)

Our backgrounds all include a fair amount of e-discovery, and that means far too much exposure to painful discussions around technology-assisted review (TAR). For those of you who had the good fortune to skip that interlude, a brief summary:

  1. The Cambrian explosion in data volumes turned document discovery into a complete nightmare, especially the crazy costs of eyes-on document review
  2. A bunch of tech emerged to try to cut down on the volume of documents requiring eyes-on review
  3. Many lawyers objected that they could not trust a machine to do as good a job as a lawyer in reviewing documents
  4. Subsequent empirical studies, now reflected in the caselaw, established that the machines were not perfect but still outperformed lawyers in terms of both accuracy and speed—debunking human review as the so-called “gold standard.” See Maura R. Grossman & Gordon V. Cormack, “Technology-Assisted Review in E-Discovery Can Be More Effective and More Efficient Than Exhaustive Manual Review,” 17 Rich. J.L. & Tech 11 (2011).
  5. More than a decade later, this remains a topic of debate. But a much larger share of documents are categorized by machines—better, faster, and cheaper.

Importantly, it is technology-assisted review. Experts remain at the helm.

To be simplistic, the machine models the decisions that humans make. Predicting each human decision in the background, the humans review, and the models iterate, until a certain confidence threshold is hit—i.e., that the machine will come to the same conclusion as the human experts, whose judgment the machine then applies across the remaining corpus. Humans still look at relevant documents—because they are relevant. But they waste far less time looking at irrelevant documents.

Human expertise remains central. The objective is to properly leverage the expertise through process and technology to reach better outcomes, faster and more cost-effectively. It is not perfect. But it is superior to the untenable status quo ante. And it continues to improve—with the emerging tech likely to represent another quality-improving, labor-sparing refinement.

We are all scarred from straining to explain to people who did not want to hear it that reducing the number of irrelevant documents humans had to review was a positive. In that same vein, we have road-tested our prediction, “In 2023, AI will be capable of generating, near instantly, a legal opinion or contract superior to the work product of 90% of junior lawyers.

Unsurprisingly, what many people hear is, “2023 will be the AI apocalypse for lawyers.” They respond accordingly.

First, 90% is not 100%. More importantly, “junior” is deliberately responsible for an enormous amount of heavy lifting. While garbage in, garbage out merits attention with regard to the reliability of unrefined large language models, we should not forget that so much of junior lawyering is already mindless copy and paste, just slower or more error-prone. See Casey Flaherty,How Much of Lawyering is Being a Copy-and-Paste Monkey?,” 3 Geeks, Jan 28, 2018. Not only do junior lawyers not graduate practice ready, but they also lack structure in their subsequent professional learning environment. See Casey Flaherty, “CLE is Broken (as is our approach to learning/innovation),” 3 Geeks, Oct 31, 2021. The thing about being chained to a desk to review an endless stream of irrelevant documents is that it is dreadfully boring and not the least bit educational.

It is a straw man to characterize the new AI frontier as some sort of galaxy brain that performs like a limitless agglomeration of the smartest humans. Rather, the better framing is an extensible corps of trainees who can perform lower-level work at warp speed.

Consider that until the 1950’s, “computer” was a human occupation. For two centuries, computers were people, who performed long, laborious calculations by hand, making invaluable contributions to the advancement of science and technology. When machines suddenly surpassed humans in accuracy, speed, and cost-effectiveness, facility with math became more, not less, valuable. Advancements in science and technology accelerated while finance and business shifted increasingly towards being data-driven.

We appear to be crossing a similar threshold with respect to language. It is not an event horizon. The new applications are likely to surpass humans at lower value, labor-intensive language-based tasks that insert so much invisible friction into how we currently work. Integration will take time. And not all of it will be good. But, on net, it will be better. On net.

The Luddites were right

Today, Luddite is a pejorative term applied to those who oppose new technology. Indeed, the “Luddite fallacy” is used to dismiss concerns around long-term technological unemployment—i.e., structural unemployment because the machines permanently took our jobs.

In early 19th-century Britain, the Luddites were an organized faction that waged a five-year rebellion that needed to be suppressed by military force. The Luddites were textile craftspeople being displaced by machinery. Many Luddites were owners of workshops closed because they could not compete with the machine-based factories. And when they tried to get jobs at a factory, many could not—because the factories required less labor. This left many people unemployed and angry. The unemployed and angry people turned to violence.

On the one hand, the historical record so far suggests that automation anxiety is likely misplaced and technological unemployment is not long-term because of compensation effects—i.e., the increased productivity creates more jobs than it destroys.

On the other hand, technology does destroy specific jobs. Individuals do not care about long-term structural employment, they care about their own near-term employment. The Luddites were right. The machines messed with their personal livelihoods.

We can only imagine the battles that are coming as (i) GPT-powered variants of LegalZoom 2.0 and access-to-justice advocates with a DoNotPay bent collide with (ii) protectionists who run most state bars and the antiquated rules around the unauthorized practice of law. But that’s not our fight.

We are preparing to help our corporate and law firm customers navigate what will become an even nosier and busier legal innovation landscape. As we help alleviate choice overload, we expect we will also have to work overtime to talk them through the structural implications of the new tech offerings. Our points of contact—those innovative outliers—will welcome such conversations. But their legal organizations might be another story (see, we told you we would bring the narrative threads back together).

Change is still a choice. Choices have consequences.

As former practitioners, we have considerable empathy for the lawyers with whom we work.

These lawyers expertly perform mission-critical work under immense time and resource constraints as regulatory complexity explodes and the attendant impact on the business intensifies. Meanwhile, the gap between the work that needs to be done and the resources available continues to grow. This crush of work and paucity of resources also means there is minimal capacity, and patience, to invest in the kind of scale-enhancing innovation that could start to close the gap. See Post 347 (preview essay discussing the bleak odds of success).

The new AI frontier presents a dual challenge. First, as lawyers, they will be presented with all manner of net new questions as businesses try to leverage the new technology in various ways. This will, almost certainly, be followed by waves of new regulations. Second, as operators, they themselves will be under all manner of pressure to modernize—but often without adequate time and resources.

We started with a review. We then made a prediction. Let us end with a question:

What happens when the CFO hires the reinvigorated [Big Name Consultancy] Digital Transformation Team for a top-to-bottom efficiency review and, among many other recommendations with profound implications for the business, the resulting report plays to the CFO’s confirmation bias, finding that legal is one of many areas where low-level work can be expeditiously automated—to the point where, in many instances, legal can be bypassed entirely?

We may not have enough data points to answer this question next year. But it is a question many of our customers will face soon enough. Buckle up!