The last 20 per cent

IT’S not a hypothetical. AI is already handling the work that was previously the remit of junior lawyers. First-cut legal documentation, due diligence, compliance checks and background searches. It does all of this faster, more thoroughly and without billing by the hour.

Clients are now better informed even before they meet the lawyer. A quick AI search brings up the relevant laws and regulations before they consult you. Virtual meetings are mainstream. Attendance notes are transcribed and translated contemporaneously. The entire dynamic of the lawyer-client relationship has shifted.

Robson Lee has been a corporate and capital markets lawyer for 32 years. He started in 1994, when the fax machine was the fastest thing in the office and faxing a 200-page document took him past midnight standing next to the machine and hoping the pages do not get jammed. Technology has made things a lot more efficient and expedient. He witnessed the rapid changes through every wave of technological developments in the last three decades. Lawyers who adapt and embrace technological developments will thrive in the new ecosystem. Those who don’t will be relegated to the “Jurassic age”, and marginalised or made redundant.

You have practised law for 32 years and lived through several waves of technological change. Looking back, which shift came closest to what AI is doing to the profession today, and why is this time different?

When I first started, we had the 286-series computers. No laptops. You drink a cup of coffee, come back and the computer was still booting itself. For prospectus drafting, we had to mark up in manuscript, legibly enough for your secretary to read and make amendments.

I still remember when I was part of the China practice involved in the Suzhou township project. Whenever a minister came to the township for a visit, clients and counter-parties would rush to have the legal documentation signed at signing ceremonies witnessed by the minister. We had to work until 2am in our hotel rooms, printing and faxing the amended documents to all parties at exorbitant costs charged by the hotel’s business centre. You couldn’t seamlessly send a 150-page document. It would jam by the 30th or 40th page. Lucky, if you got more than half across.

AI is different because it doesn’t just speed up the process. It compiles and segments documents seamlessly with a high degree of accuracy and completeness. That’s new territory.

AI can now do in hours what once took legal teams weeks. At what point does that stop being an efficiency gain and start becoming an existential question for the profession?

Clients will not accept 50 hours of billing for work that was previously undertaken by junior lawyers. They will say: what did you do in those 50 hours? AI can do this super-fast. You can’t charge hourly rates for something the machine can undertake in less than 10 per cent of the time undertaken manually.

A clause that once took 30 minutes to draft now takes five. You read through it, refine it, customise it for the client. It cuts at least 50 to 60 per cent of the original time. Background checks that used to require ploughing through court case reports and old newspaper records can now be condensed and compiled in a fraction of the time.

Clients know what is available and expect you to deploy available software as an integral part of your service delivery.

There is a school of thought that says AI will eventually replace lawyers entirely. You’ve been in enough boardrooms, courtrooms and deal rooms to have a view on that. Where do you stand?

The machine can generate something that is probably 60, 70, even 80 per cent of the getting-up. But the remaining 20 per cent is what’s most critical. That’s where experience and human judgment is required, to deliver bespoke advice tailored to the client’s specific circumstances and commercial needs.

You still need to sit down, interview the client, have meetings, understand their needs and unique position, read the body language, the nuances, and know where and how to apply the law to best address their specific needs. AI cannot do that.

Let me give you an example. I had a client from China, a very old-school gentleman, very hospitable who had business clients visiting Singapore. After dinner, he drove them to the airport to see them off. At the airport, his client wanted to buy a specific Chinese liquor. But the shops were closed and the only option was the duty free liquor shop inside the airport’s departure hall.

So he bought an air ticket, went through immigration and helped his client buy the liquor, sent him off at the departure gate, and came out without flying. That’s when he was arrested, handcuffed and taken to the police station where he was detained until 3 or 4 in the morning.

The airport is a gazetted security area. People cannot enter the departure hall and exit without departing. Being a foreigner, he was unaware that he had broken the law. He was under medical treatment and had taken strong medication that day. There were language problems as well. The investigation officer had problems understanding his regional Chinese dialect which he spoke with a heavy accent. None of that context went into his statement. It came out because I sat with him and heard the full story. That’s what I was able to put before the authorities. The outcome was different because of it. AI isn’t going to do that. Such mitigating factors could not be pre-programmed into any system. AI would not be able to generate a representation letter to explain the extenuating medical circumstances.

AI generates content quickly, but courts have seen lawyers submit cases citing precedents that never existed. How serious is that risk?

Very serious. AI can generate something grammatically perfect, well-structured, but completely false. There have been embarrassing situations where courts were misled and judgments were made on the basis of fake precedents cited by lawyers using AI. It turned out those cases never happened. The precedents cited as persuasive or binding authority never existed. The lawyer had used AI and assumed the cited court judgments were genuine. They were fake precedents.

Verify what the system churned out. You cannot press a button, get something that looks authoritative, and submit it. Every cited precedent needs to be verified. Lawyers will be held accountable for false submissions. AI gives you the first cut. You have the duty to check and verify.

You’ve talked about the 80 per cent that AI can do. But the 20 per cent it can’t, that’s where outcomes actually change. What does that look like beyond the document?

Although I am not a litigation lawyer, I have attended court proceedings on watching brief. In court proceedings you need an experienced advocate with good advocacy skills. To argue a case, to cross-examine witnesses, you can’t put a robot there. You can’t have a pre-programmed robot asking razor-sharp questions, and expect the witness to crumble. When you are personally in court, you can read the body language, is the witness flinching and cringing? You have to press the person until you break the witness.

For corporate transactions, before you onboard the client, as part of the know-your-client due diligence, you must meet the client, ask questions and listen to his verbal responses and detect the non-verbal signs by observations. This will help you make a better judgement of the prospective client. While AI can pull the records fast, it can’t raise the follow-up questions to uncover underlying important facts and circumstances that will only come out when you continuously probe and listen.

If clients can already see what AI does and what it costs, how does that change the conversation about what lawyers charge and what they’re actually worth?

The billing model has to change.

Clients want solutions. They aren’t paying lawyers to recite which law is applicable. They want to know: I have a problem, how do you solve it? They are paying for experienced judgement and solutions to their specific problems.

That part has not changed. If anything, it’s now the only part that matters.

You work primarily in capital markets. How is AI changing the way companies raise funds and go public?

The heavy lifting of initial research and prospectus drafting that used to take days can now be done in hours. I have a client I’m currently advising on an IPO. He spoke to us in Mandarin throughout our meeting, but his phone was running an AI feature that transcribed and translated everything into English in real time. By the time our meeting ended, he had already sent us the transcribed notes. My junior lawyer hadn’t even finished taking notes by hand. That is how fast things are moving.

That changes the billing model. You cannot charge for hours of drafting when AI produces a draft contemporaneously at the meeting.

Capital markets is also being reshaped by how investors think about valuations. Companies now need a credible AI component to command a market premium. Investors buy on hope and sell on fear. What drives investors’ interests is something that promises to be transformative. Whether the business model commands a significant valuation multiple is, in the end, a judgment call. Accurate earnings forecast needs market experience and human judgment.

The Business Times marks its 50th anniversary this year, a milestone that invites reflection on how much Singapore’s business landscape has changed. As someone who has practised through much of that period, what do senior lawyers owe the next generation navigating this moment, and what does the profession need to protect as it transforms?

We have to be in sync with technological developments as this comes with changes in market expectations, changes in what clients want from their lawyers, and changes in the industry itself. The legal industry has evolved, requiring lawyers to exercise good professional judgment and to deliver legal solutions efficiently, deploying technology as a core tool of trade.

The back-breaking tasks of research, initial drafting and verification of records will increasingly be handled by AI. That’s not a science fiction. It’s now part of the legal landscape and ecosystem. It frees lawyers to focus on what the machine cannot do. The judgment calls, the face-to-face, reading people and situations, and dextrous application of the law in the best interests of the client.

Certain things still have to remain as core aspects of the practice of law. You have to be responsible for what is communicated. You have to check the output AI generates to ensure that what comes out from the computer is not fake, false or a product corrupted by hallucinations and prejudices as a result of the AI’s intrinsic defects or shortcomings. You still have to sit down with the client and understand their situation and know where to put the emphasis.

You have to stand before a judge and address his concerns. You need greater situational awareness and respond intuitively while exercising professional wisdom and judgment to sieve the gold from the dross, to achieve the best outcome for your clients.

Technology will keep changing. From the telegraph machine, the fax machine, the computer, email, the smartphone to virtual meetings, and now AI. The profession has to continually keep pace and evolve. That is what the next generation of lawyers has to do. Adapt, apply the latest available tools of the trade, and exercise cognitive professional judgment which no machine can substitute or replicate.

Robson Lee is a senior corporate, finance and capital markets lawyer with more than 30 years of transactional experience.