You’re a lawyer: will AI eat your lunch?

They say lawyers seldom give you a straight answer. So, I’m going to start by answering the question:

AI is not going to eat your lunch.

However (with lawyers, there’s always a “however”), lunch will be shared. I’ll explain. But first some historical context.

1.?? Lawyers and IT—a personal history

I’ve always been a business lawyer, always on the transactional side and always in private practice. I’m writing only about what I know (or think I know).

In my final year at boarding school, there was only one guy in my dorm who could type. He knew he wanted to be a journalist and his mom had taught him shorthand and typing. In 1970, boys didn’t type and the smartasses in the dorm might have been tempted to laugh at him, but his nickname was Animal—no one ever laughed at him.

At university I read Law. We wrote our essays and research papers by hand. It was torture. Yet somehow I avoided learning to type until 1992.

I first set foot in a law firm as an Articled Clerk in 1975. My Principal had three secretaries, two of whom used manual typewriters. Soon after I joined the firm, the lead secretary got a word processor, which used 8-inch magnetic discs, each of which held about one page of memory. IBM had introduced the ‘Mag Card’ system back in 1969.

Outside my office there was a telex machine. Fax machines would not arrive until the 1980s even though the technology had been evolving since 1843.[1] But it wasn’t useful until most clients and firms had one.

Lawyers had desktop dictating machines, but Articled Clerks had to write letters and memos by hand, have them typed by a secretary and settled by a partner, usually in a face-to-face session that involved varying degrees of humiliation. In time, I learned the art of composing sentence after sentence in my head and dictating them fluently into a microphone. This is a skill I no longer have.

Before the advent of word processors, contracts were semi-standardised. If you wanted to change a clause, you wrote a proviso and (literally) pasted it at the end of the clause in the template. So, if you’ve ever wondered why some lawyers add proviso after proviso to the clauses they draft (instead of using the more logical “if, then” style of drafting), it’s a hangover from that ‘cut, copy and paste’ era.

Yet, since Law School, I had been using the first of the AI-powered legal research tools, and so were a lot of others. This was LEXIS (now Lexis-Nexis). Introduced in 1973, it began to be profitable in 1977, which means that a lot of lawyers were using it by then.[2] Along with Westlaw, it improved steadily over the years.

When I arrived in the City of London in 1983, most law firms had a Post Room, often staffed by wisecracking retirees from the Smithfield meat market. There were always lots of leather-clad motorcycle couriers milling around the Post Room ready to ferry documents across the City. There are no Post Rooms now, though there are even more bike couriers. They don’t deliver documents—they mostly deliver food.

The whole top floor of the building we occupied was open-plan and populated by women. This was ‘Central Typing,’ and it was open 24 hours each weekday and 8 hours on Saturdays, Sundays and Bank Holidays—and it was always busy. It paid to deliver your documents personally and spend a few minutes chatting with ‘the girls’ (as they called themselves—though most were middle-aged).

Each junior lawyer shared a room with a partner or senior associate, and every lawyer had a secretary or shared one with another lawyer. The secretaries typed letters, memos and shorter contracts.

In the 1990s, computers appeared on desks. Then email, a technology first developed in the 1960s, began to displace letters and faxes, at first slowly, then rapidly as a critical mass of clients began using it.

Mobile phones were not common until the Blackberry burst onto the scene in 1999. When they improved it so that you could attach documents to an email and send them from your Blackberry, we thought it was magical. By 2007, the iPhone had killed off the Blackberry.[3] By then, we all had lightweight laptops and used Wi-Fi and Bluetooth.

In the early 2000s, document automation became fashionable. I was an enthusiast, but it didn’t make the inroads I’d expected. Automated document comparison, electronic signing and document management systems were widely used and other AI products aimed at lawyers began to proliferate around this time, many using ‘logic gate’[4] and ‘decision tree’[5] software.

Around 2020, the adoption of AI in law firms surged. A September 2023 LexisNexis survey revealed that 41% of lawyers were using generative AI at work, up from just 11% in July 2022.[6]

Overall, law firms have been comparatively slow to integrate innovative technologies. They spent (usually overspent) on technology, then failed to maximise the value it could add. Over the past fifty years, productivity growth has mainly been a function of more lawyers billing more hours at higher rates, with some offshoring of routine work to low cost places like India. AI has not yet changed the game—but it soon will.

2.?? The rise of the suits

When I qualified in 1976, the ‘Golden Age of Capitalism’[7] was already over. From 1945 to the mid-1970s, the global economy experienced rapid and sustained growth, with an average annual growth rate of 4.8%.[8] The United States saw unemployment drop to 4.8%, while in countries like Britain and France it virtually disappeared. Income inequality decreased during this period.

From the mid-1970s, global economic growth has trended downwards. In the UK, the average annual rate of GDP growth for the period since 1975 has been 1.45%.[9] Yet, since the 1970s, the number of solicitors in the UK has increased by approximately 50% every decade.[10]

Although the revenue of many law firms has grown strongly during the last 50 years, profit margins are tightening. With the end of the recent ‘free money’ era, this margin compression looks likely to continue.

But the numbers only tell us part of the story.

3.?? From the Three Musketeers to Taylor Swift

Although ‘Big Law’ has done far better than the overall economy since the 1970s, this has been at the expense of the “all for one, one for all"[11] ethos that prevailed when I entered the profession.

Law firms now reflect what Sherwin Rosen, an economist at the University of Chicago, calls “The Economics of Superstars”—where financial rewards and market share tend to concentrate among a small number of highly successful firms or highly talented individuals.[12] In the entertainment industry, Taylor Swift epitomises this phenomenon.

For the last 30 years or so, the idea that there are partners who ‘own’ particular client relationships has gained traction. The remuneration of these ‘rainmakers’ has pulled away from that of their peers. In part, this is a by-product of the Information Age—better communications leading to higher visibility for the best performers. But is the Superstar model sustainable?

I think the ‘Moneyball’ model is a better fit for law firms, not only because practising law is a ‘team sport’, but also because AI is likely to de-throne many of the legal superstars. The Moneyball approach (the name comes from the 2003 book of the same name by Michael Lewis)[13] is known as sabermetrics.[14] As used by the Oakland As, it relied on statistical analysis to identify undervalued players and create a competitive team despite having one of the lowest payrolls in Major League Baseball. What Moneyball revealed was that, in cost-benefit terms, superstars are less valuable to a team than a cohesive collection of unflashy, proficient, positional players.

In less than a century, legal practice has evolved from a cottage industry to a multinational service sector, but the metrics it uses have not evolved as fast. Properly used, AI will do at least three things for law firms:

·?????? First, it will lift the performance of non-superstar lawyers by improving their productivity, range of competence and written communication skills.

·?????? Second, it will enable law firms to measure the performance of their partners more comprehensively, which (I suspect) will show that superstar partners are much less valuable absent the ecosystem the firm provides for them. There will always be stars whose personal magnetism brings in clients, but the quiet achievers are likely to become more visible to law firm management and more valued.

·?????? Third, AI is likely to change the revenue model of law firms.

4.?? Time is not money

When I started in the law, we didn’t charge by the hour. We charged by the task. The Law Society in my home State had a scale of charges for defined tasks. It arose in the context of litigation so that costs could be assessed (‘taxed’) by a quasi-judicial officer called a ‘Master’ and allocated according to the outcome of a case. Transactional lawyers used the scale too, and any client could ask a Master to ‘tax’ a lawyer’s ‘costs’ (adjudicate fees and expenses) whether the matter related to litigation or a transaction.

Task-based billing for transactional work went out with purple shirts, long collars and paisley ties. In its place came the ‘chargeable hour’. From the law firm’s point of view, its main drawback was the paltry number of hours in each day. From the client’s viewpoint, the downside was Parkinson’s Law—work expands to fill the time available to perform it.[15]

Unloved since birth, the chargeable hour refuses to die. The Clio 2023 Legal Trends Report discloses that: “Since 2016, lawyer utilization rates have significantly increased from 28% to 37%—an improvement of 32%. This increase represents nearly three-quarters of an hour of billable time added for every day worked.”[16]

The attention span of knowledge workers is surprisingly short. And, after an interruption, it takes more than 20 minutes to regain full focus. Many experts suggest that there is a natural upper limit to deep, focused work of about 3 to 4 hours per day.[17] Yet many firms require their lawyers to bill 1,800 to 2,000 hours a year. Even if we were to assume a utilisation rate of 50% and 300 working days a year (which means working 6 days a week, most public holidays and some Sundays, and taking only 2 weeks annual leave), this translates to a working day of 12 to 13 hours, with 6 to 7 hours devoted to billable work and the rest spent on other things, mainly administrative and business development tasks.

This is good news for AI vendors and it ought to be good news for over-worked lawyers. But is it good news for the way law firms ‘utilize’ their lawyers and bill their clients?

There are tasks that would take a lawyer (or even a team of lawyers) days or weeks to perform that AI could take care of in minutes. For example, Epiq AI Labs recently announced the launch of eDiscovery software that can process 500,000 documents an hour.[18] AI takes time out of the value equation and clients will be among the first to recognise this. Legal services, being intangible, have always been hard to measure and value. Time spent has always been a lazy proxy for value. AI exposes the need for a better one.

If time is not the measure, where does value lie?

5.?? The sorcerer’s apprentice

When I left Law School, I couldn’t write a decent letter, far less a contract. I had no idea how to conduct myself in meetings and I had never negotiated with ‘grown ups’ other than my parents. I learned these things from experienced lawyers in the firm where I was articled. Most of the lawyers who taught me the ropes were nicer to me than I had any right to expect. They patiently corrected my work, took me to client meetings, shared war stories and, occasionally, dropped me in at the deep end, but didn’t let me drown. This master-apprentice relationship is no longer what it was.

Over the years, leverage (the ratio of associates and trainees to partners) has crept up and partners have been required to run ever faster on a financial treadmill that slopes ever upwards. Big Law pays its trainees and newly qualified lawyers very well despite their lack of experience. But, below partner level, remuneration does not keep pace with experience. So, who trains junior lawyers these days? Is it the associates, who are underpaid relative to their experience, or is it the partners, who are run off their feet?

What clients value is judgement—and judgement requires, among other things, experience.

Malcolm Gladwell’s 2008 book, Outliers,[19] popularised the idea that 10,000 hours of practice is a magic bullet to success. It became a “rule”, even though it arose from Gladwell’s controversial misreading of a 1993 study by Anders Ericsson and colleagues.[20] Ericsson made no claim about the number of hours required for professional mastery, emphasizing that quality is far more important than quantity. He said that “…not every type of practice leads to improved ability. You don’t get benefits from mechanical repetition, but by adjusting your execution over and over to get closer to your goal”—the goal being excellence.? He calls this “deliberate practice” and it is usually guided by an expert, a coach or a mentor. In the practise of law, the first 5 to 7 years are critical. During this period the need for guidance is at its highest and it should involve not only legal practice knowhow but also guidance on developing ‘personal’ skills.

6.?? Trust me; I’m a lawyer

People trust lawyers more than they trust accountants, bankers, advertising executives and real estate agents, but less than they trust doctors, nurses, teachers and engineers.[21] Trust is delicate. It’s hard to win and easy to lose. Above all, it is personal. Even when law firms have institutional connections with their clients, those doing the hiring usually seek out the lawyers they trust in the firms they are directed to use. Many clients say: “I don’t hire the firm; I hire the lawyer”.

David Maister, one of the world's leading authorities on the management of professional service firms, describes the ideal lawyer as a ‘trusted advisor’, someone who uses expertise and judgement to persuade and guide the client. Maister argues that understanding human behaviour and relationships is crucial in professional services. Expertise is largely assumed, he says; judgement and persuasion are not. Judgement is born of experience, which comes from ‘deliberate practice’.[22] It requires trust, which grows out of personal relationships. Persuasion also requires trust. Additionally, it requires the lawyer to identify with the client. As Benjamin Franklin said: “If you would persuade, you must appeal to interest rather than intellect”. This means putting the client’s interests before our own (but not that of the law). Clients need to trust that we are not self-interested, which they are more likely to do if they see value for money in our advice. It is hard to trust someone you think is over-charging you.

In summary, lawyers dispense professional advice, which requires trust, experience, judgement and persuasion. There is a learned alchemy to this—one that AI doesn’t have and may never have.

7.?? So, who gets to eat lunch?

The key question in our relationship with AI is not whether we can tell a machine from a person in a blind test, but whether a machine can tell a person from another machine. In other words, whether AI will ever understand human nature—understand why we enjoy eating lunch or what indigestion feels like.

Alan Turing called his test “the imitation game”.[23] AI already goes beyond imitation. It can do things we consider creative. And, depending on your definition of thinking, it can do that too. But it can’t think about thinking. It doesn’t know what it’s like to look forward to lunch because it is not sentient. If it were, it could exercise judgement the way we do, involving that strange fusion of reason, emotion, contextual sensitivity, sociability and morality.

Some think that sentience will ‘emerge’ from processing troves of data in artificial neural networks. I’d like to think that sentience is organic, not electronic. But I could be wrong. After all, intelligent life probably emerged on this ‘goldilocks’ planet from a mysterious combination of lifeless matter and energy. Perhaps quantum computing combined with vast inputs of data and electricity could somehow produce a good facsimile of human consciousness, complete with deep understanding of human feelings, imagination and sociability. If it could, then combining this with AI’s computational and pattern analysis capabilities, would take away everyone’s lunch.

There is an adage in engineering that the first 90% of a project takes 10% of the time and the last 10% takes 90% of the time. This is hyperbole, but something like it may apply to AI. Some of the current hype about AI should be discounted, though by no means all of it. Someday, Artificial General Intelligence, the holy grail of AI, will come to a device near you. But for now, you can enjoy your lunch while AI works through your lunch hour—provided that—the firm you work for uses AI to help you learn your craft in a human and humane environment that rewards skill, ethics and judgement and not just hours billed.

?

[1] https://www.thoughtco.com/history-of-the-fax-machine-1991379.

[2] https://en.wikipedia.org/wiki/LexisNexis.

[3] https://www.macworld.com/article/562752/blackberry-devices-software-obsolete-iphone.html#:~:text=The%20end%20comes%20almost%20exactly%2015%20years%20after,believed%20that%20Blackberry%20would%20suffer%20such%20a%20fate.

[4] Logic gates such as AND, OR, and NOT are used to create complex decision-making structures. These gates allow AI systems to evaluate multiple conditions simultaneously and determine appropriate actions based on the inputs.

[5] Decision trees are often used in conjunction with logic gates to create hierarchical decision-making processes. Each node in the tree represents a decision point, where logic gates can be applied to determine the next step or action.

[6] https://www.lexisnexis.co.uk/blog/future-of-law/ai-adoption-soars-across-uk-legal-sector.

[7] https://en.wikipedia.org/wiki/Post%E2%80%93World_War_II_economic_expansion.

[8] Calculated using data from https://www.macrotrends.net/global-metrics/countries/wld/world/gdp-growth-rate.

[9] Calculated from data in https://www.macrotrends.net/global-metrics/countries/gbr/united-kingdom/gdp-growth-rate.

[10] https://www.legalfutures.co.uk/regulation/solicitors/inexorable-growth-of-solicitors-profession-as-number-on-roll-tops-150000.

[11] This famous phrase is traditionally associated with the titular heroes of Alexandre Dumas' novel "The Three Musketeers," first published in 1844.

[12] Rosen, S. (1981). The Economics of Superstars. American Economic Review, 71(5), 845-858.

[13] Michael Lewis, Moneyball: The Art of Winning an Unfair Game. W.W. Norton & Company. 2003.

[14] Sabermetrics is the empirical analysis of baseball through advanced statistics, aimed at objectively evaluating player performance and team strategies. This analytical approach was coined by Bill James in 1980 and is derived from the acronym SABR (Society for American Baseball Research).

[15] Parkinson, C. N. (1958). Parkinson's Law: or The Pursuit of Progress. London: John Murray.

[16] https://www.clio.com/resources/legal-trends/2023-report/read-online/.

[17] https://www.kcl.ac.uk/news/are-attention-spans-really-collapsing-data-shows-uk-public-are-worried-but-also-see-benefits-from-technology https://brainmindsociety.org/posts/are-attention-spans-actually-decreasing. https://www.pesto.tech/resources/can-we-really-do-deep-work-more-than-6-hours-a-day

[18] https://www.epiqglobal.com/en-gb/resource-center/in-the-news/epiq-global-launches-ai-powered-e-discovery-platform-and-ai-res.

[19] Gladwell, Malcolm. Outliers: The Story of Success. Little, Brown and Company, 2008.

[20] Ericsson, K. A., Krampe, R. T., & Tesch-R?mer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review, 100(3), 363-406.

[21] https://www.lawgazette.co.uk/news/public-less-likely-to-trust-lawyers/71519.article; https://www.statista.com/chart/19397/trusted-professions-in-britain-ipsos/; https://news.gallup.com/poll/608903/ethics-ratings-nearly-professions-down.aspx.

[22] See generally Maister, D. H., Green, C. H., & Galford, R. M. (2000). The Trusted Advisor. Free Press. A 20th anniversary edition published in 2021. Still one of the best books on the subject.

[23] Turing, Alan. Computing Machinery and Intelligence. Mind. Oxford University Press. 1950.