Industries / Finance and business services / Legal and professional services / LegalTech
Running a practice
Running the business of a law firm or legal department: tracking time and bills, keeping documents, checking conflicts and protecting client data under the profession's rules.
7
Milestones achieved, from 1998 to 2024.
3
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1998
A shared e-billing format
In the mid-1990s US law departments and insurers found paper bills from outside lawyers almost impossible to compare. A joint group of the American Bar Association, the Association of Corporate Counsel and PricewaterhouseCoopers created UTBMS, the Uniform Task-Based Management System, a set of codes for the phase and kind of work, and a consortium led by PricewaterhouseCoopers issued LEDES 1998, a file format for electronic bills, soon replaced by LEDES 1998B.
LessonShared codes help only when people apply them the same way.
Credit
ABA, Association of Corporate Counsel, PricewaterhouseCoopers; LEDES Oversight Committee
2010
A bar allows cloud storage
New York's bar ethics committee said in Opinion 842 that a lawyer "may use an online data storage system to store and back up client confidential information", provided the lawyer takes reasonable care to keep it confidential. It noted that no earlier New York opinion had addressed the question.
LessonThe duty of care stays with the lawyer, who has to check the provider.
Credit
New York State Bar Association Committee on Professional Ethics, 10 September 2010
2012
A duty of technology competence
In August 2012 the ABA added to the comment on its competence rule (Comment 8 to Model Rule 1.1) a duty to keep up with the benefits and risks of relevant technology. By LawNext's tracker, 40 states, the District of Columbia and Puerto Rico have adopted it. Canada's Federation of Law Societies added similar wording in October 2019: a lawyer "should understand the benefits and risks associated with relevant technology."
Credit
ABA House of Delegates; state supreme courts; Federation of Law Societies of Canada
2016
Billable hours measured at scale
Clio, a maker of cloud practice-management software, published its first Legal Trends Report, drawn from its users' records. It found lawyers billed 28% of their available hours, about 2.24 hours of an 8-hour day, a share called the utilization rate. The data is Clio's own; the ABA Journal reported it.
LessonOnce time is recorded in software, a firm can see how much of the day never reaches a bill.
Credit
Clio
2017
Malware shuts a global firm
In June 2017 the NotPetya malware hit DLA Piper's data centres and Windows servers. Staff worked about 15,000 hours of paid overtime in three weeks to recover, and the firm wiped its Windows systems to start again. Its general counsel said no data was lost because the backups were untouched.
LessonBackups kept apart from the main systems decide whether a firm can recover.
Credit
none, a failure; reported by iTnews
2022
A fine after ransomware
The UK Information Commissioner's Office (ICO) fined Tuckers Solicitors £98,000 after a 2020 ransomware attack encrypted 972,191 of the firm's files and the attackers posted 60 court bundles online. The ICO said the firm lacked multi-factor authentication, a second login check such as a code sent to a phone, for remote access, a "comparably low-cost preventative measure."
LessonBasic controls such as multi-factor login are now the standard a regulator expects for client data.
Credit
UK Information Commissioner's Office; reported by the Law Society Gazette
2024
Ethics rules applied to AI
ABA Formal Opinion 512, issued in July 2024, applied lawyers' existing duties to generative AI. Lawyers need "a reasonable understanding of the capabilities and limitations" of the tools they use; they need a client's informed consent before entering client information into tools that may learn from it; and lawyers who bill by the hour "must only bill for their actual time."
Credit
ABA Standing Committee on Ethics and Professional Responsibility
Watch for
An AI-run law firm
In May 2025 the Solicitors Regulation Authority (SRA), which regulates solicitors in England and Wales, authorized Garfield.Law, "the first purely AI-based firm we have authorised." Its AI guides businesses through small debt claims and cannot propose case law. Garfield says it has handled more than 600 claims and recovered about £500,000, and in June 2026 a client it helped won £7,000 at trial in Wandsworth County Court, with barristers arguing for both sides.
Achieved whenthe SRA or an outside reviewer publishes the service's success rates, errors and complaints.
Claimed by
Garfield AI, June 2026
Automatic time capture
Laurel, which sells AI that turns lawyers' computer activity into time entries, says it recovers "Up to 30 minutes of additional billable time" per lawyer each day, and that "10–30% of billable work never makes it to an invoice."
Achieved whenan outside party compares billed hours before and after rollout at several firms and publishes the change.
Claimed by
Laurel and Thomson Reuters, August 2026, after a Laurel survey of 2,000 professionals in November 2025
Practice data meets research
Clio completed its US$1 billion purchase of vLex, a legal research company whose AI assistant is called Vincent, on 10 November 2025, and says Vincent offers "the strongest foundation for legal reasoning." Clio reports 400,000 legal professionals as customers.
Achieved whenan independent test scores Clio's combined tools on practice tasks and publishes their accuracy.
Claimed by
Clio, November 2025
Open
01
A scored UTBMS coding test
A public set of invented time entries, the short notes lawyers write such as "Draft motion to compel; review objections", each given its correct UTBMS task and activity code by two trained coders, and a script that scores AI models against the people.
The briefCredit · unclaimed02
A conflict-check test
A conflict check is the search a firm runs before taking a client, to make sure it will not act against a current or former client. Build an invented firm database of clients, opposing parties and past matters, full of name variants (abbreviations, typos, former names) and corporate families (parents and subsidiaries), with new-client queries whose true conflicts are known. Score the conflicts found and the false alarms.
The briefCredit · unclaimed
Open 01 · the brief
A scored UTBMS coding test
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of invented time entries, the short notes lawyers write such as "Draft motion to compel; review objections", each given its correct UTBMS task and activity code by two trained coders, and a script that scores AI models against the people.
Why it is still open
Brightflag, a company that reviews legal bills, writes that "creating an accurate, standardized dataset is impossible because lawyers interpret and use the codes differently", and Litera says the codes are "applied with frequent inaccuracies and generalizations." Neither publishes a test set or an accuracy rate, and I found no public labeled set.
Built with what exists
the UTBMS code lists; Claude, Gemini or GPT models; a classifier that returns a probability for each code on a fixed list, such as Gemini Decision 1 (see the Note "A thousand questions, one pass"); a scikit-learn baseline, a simple statistical classifier.
First step
Write 200 entries for one mock lawsuit, have two people code them separately, and measure how often they agree.
Done looks like
At least 2,000 entries across at least 10 mock matters; the two coders' agreement reported as Cohen's kappa, a standard score for agreement beyond chance; and accuracy on each code for at least three models beside the coders' own agreement, under an open license.
Open 02 · the brief
A conflict-check test
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A conflict check is the search a firm runs before taking a client, to make sure it will not act against a current or former client. Build an invented firm database of clients, opposing parties and past matters, full of name variants (abbreviations, typos, former names) and corporate families (parents and subsidiaries), with new-client queries whose true conflicts are known. Score the conflicts found and the false alarms.
Why it is still open
CorpFam, a public benchmark released in September 2026, found that the best of five matching methods caught 4.2% of same-family company pairs whose names share nothing; it uses supplier records, not law firm data. I found no public conflict-check test.
Built with what exists
CorpFam's data, drawn from US federal award records in which suppliers report their ultimate parent; the subsidiary lists companies file with the SEC (Exhibit 21 of the annual report); Faker, a library that makes up names; RapidFuzz for fuzzy name matching as a baseline; Claude, Gemini or GPT models.
First step
Take 200 families from CorpFam, build a mock firm with 1,000 past matters, and write 50 queries with answers.
Done looks like
At least 10,000 invented parties and 500 queries with labeled conflicts, with the share of conflicts found and the false-alarm rate for each method, reported separately for name-variant and hidden-affiliate conflicts.