On 25^th of June 2025, LatentBridge hosted an exclusive Legal AI Roundtable in London, bringing together leaders from across the legal and financial ecosystem. The room was packed with senior professionals from firms and institutions like Clifford Chance, Paul Hastings, Charles Russell Speechlys LLP, Simmons & Simmons, MONZO, Investec, Mizuho, Santander Corporate & Investment Banking, Addleshaw Goddard, Withers LLP, and others shaping the legal AI agenda from the frontlines.

The conversation was led by Hema Gandhi (CEO, LatentBridge), Anthony Vigneron (Director of Legal Technology, Clifford Chance), and Tara Waters (Founder, TLW Consulting)—each bringing deep experience in driving legal transformation at scale. 

And this wasn’t a showcase. It was a candid, behind-the-scenes dialogue on the realities of AI adoption in legal: what’s working, what’s not, and what’s next. 

Here’s what the most forward-thinking legal teams had to say. 

AI Has Moved In—But Adoption Is Still Uneven

Nearly all UK law firms report some level of AI integration. But inside the room, leaders were honest: real adoption is messy. 

One participant shared how their early AI lab failed—three times—before finally taking root with the right governance model. Today, their AI governance council involves 350+ people across business functions. That’s what it took to stabilise AI adoption at scale. 

The lesson? AI tools aren’t the problem—alignment, change management, and trust are. 

There was also a refreshing consensus that the legal sector is now moving beyond the hype and into practical, grounded use cases. It's no longer about experimenting with AI for the sake of it—firms are now rigorously testing real implementations across legal workflows.

The Back Office Is the Big Opportunity 

There was a strong consensus: the real AI value lies in the legal back office.
Tasks such as: 

• Business operations like billing, payables etc.
• Litigation prep 
• Document summarisation and clause analysis
• Compliance monitoring
• Contract deviation checks

...are still largely manual and time-consuming. That’s where GenAI, when paired with domain-specific agents, can slash hours off workloads. 

But progress remains slow. Data remains fragmented, buried in contracts, siloed across systems, and often restricted by privacy regulations. As one participant noted, “There is still a lot of untapped knowledge buried in contracts, and without data availability and governance, AI just can’t operate efficiently.” 

There were also cautionary notes around misplacing trust in GenAI for document-heavy workflows—some innovation leaders urged firms not to abandon best-practice automation tools in favour of flashy GenAI replacements.

Agentic AI Is Here—But It’s Not Plug and Play 

While some firms have built experimental agents—especially in banking and onboarding workflows—Agentic AI is still largely in discussion, not deployment, and certainly not a scaled deployment. 

Everyone agreed: 

• Agents work best when tied to workflow orchestration, not as standalone tools 
• Human-in-the-loop (HITL) remains essential for legal quality, trust, and compliance
• And while agents can deliver speed, most firms prioritise quality and reliability

At LatentBridge, this is exactly the focus:

• IntellixCore helps legal ops leaders build and orchestrate explainable, traceable AI agents—with built-in human validation and compliance alignment.
• Epic AI is built for knowledge-heavy teams, surfacing context-rich insights across jurisdictions without the tab-hopping nightmare.

Governance Is the Non-Negotiable Layer

Everyone in the room agreed: if it doesn’t have governance built-in, it doesn’t belong in legal. 

AI systems must show: 

• How decisions were made
• What data sources were used
• Where exceptions occurred
• Why an outcome was reached

Some attendees voiced concern that most off-the-shelf solutions still lack meaningful governance. It’s why the room leaned heavily toward hybrid strategies—especially when client data is involved. 

There was also renewed interest in SMEs playing a critical role (champions) in these innovation programs—not just for technical input, but to educate peers, tackle adoption bottlenecks, and validate AI-prepared outputs before they reach clients. “Your SMEs,” one guest shared, “are your real innovation champions.”  

Buy, Build, or Both? The Verdict Is Clear: It Depends 

When it came to the age-old “buy vs build” debate, the group landed here:

• Off-the-shelf solutions are good for speed—but often lack governance, customisation, and depth.
• In-house builds provide control—but demand resources, budget, and internal capability as well as planning for the long-term maintenance.
• The hybrid model won the vote. It's pragmatic. It lets firms start fast where possible, and build where it truly matters, based on their custom workflows. Here, external vendors can help augment internal teams to expedite build. 

Most agreed that AI for document automation can often be bought. But for deep domain workflows, like regulatory reviews or litigation support, customisation is critical. 

Legal AI Is as Much a Culture Shift as a Tech Shift

One thing that came up repeatedly: AI adoption is not just a tech story—it’s a people story.

• Adoption happens when lawyers see AI in action—not when it’s forced on them.
• Trust is built when firms involve knowledge teams, innovation squads, and fee earners together.
• Partners must remain the interface with clients—but AI can power the insights behind that conversation. 

Interestingly, some firms had even handed over the reins of AI ownership to their knowledge teams—considered bold at the time, but ultimately seen as effective in driving traction.

Legal AI Is as Much a Culture Shift as a Tech Shift

One thing that came up repeatedly: AI adoption is not just a tech story—it’s a people story.

• Adoption happens when lawyers see AI in action—not when it’s forced on them. 
• Trust is built when firms involve knowledge teams, innovation squads, and fee earners together.
• Partners must remain the interface with clients—but AI can power the insights behind that conversation. 

The Innovation Budget Bottleneck: AI Strategy Under Pressure

Legal teams know the reality all too well: general counsels aren't getting the budget they need to innovate at scale—despite mounting pressure to do more with less.
But creative strategies are emerging: 

• Group-wide platforms are being adopted to spread costs
• Cross-functional collaboration is raising internal visibility
• Pilot programs are being launched quietly before scaling

Paralegal workloads also emerged as a strong candidate for early AI adoption—seen as a proving ground where AI can assist with volume-heavy tasks without compromising legal nuance. 

The Rise of AI + Human Workflows 

Far from replacing junior lawyers, the discussion shifted to: how do we train them in an AI-first world? 

There’s concern that juniors are missing exposure to foundational work. The fix? Pair humans and AI, not pit them against each other. Prompt engineering, workflow design, model oversight, and legal judgment must be learned—not automated away.

Several firms noted that reskilling is no longer optional—lawyers now need to learn how to supervise AI, validate outputs, and adapt to a hybrid legal delivery model. The transformation isn't about replacing talent—it's about raising the bar on how that talent is applied.

Designing better work became a theme—if routine tasks move to AI, the human workload becomes emotionally and cognitively harder. Burnout risk rises. Which is why some leaders advocated for not just better AI tools—but better work design altogether. 

Client Expectations Are Getting Sharper 

Clients are asking hard questions: 

• "Where exactly is AI used in your process?"
• "Why are you using it?"
• "How does it impact our data security?"

Several attendees shared that while GenAI is being requested in pre-RFP conversations, the real scrutiny happens after the deal is signed. Clients want clarity on usage, benefits, risks, and controls. 

That means transparency, not just capability, is becoming the new competitive edge.

Final Take: The AI Future Is Not Agents vs Lawyers—It’s Agents With Lawyers

Legal leaders are no longer debating if they should adopt AI. The question now is how to do it right. 

Over the next 12 months, expect: 

• More firms orchestrating multi-agent workflows across litigation, compliance, onboarding 
• Rising demand for explainable AI from regulators and clients
• A greater emphasis on cross-functional collaboration between legal, IT, and innovation teams

Because the legal AI workforce of the future won’t be all human—or all machine.
It’ll be orchestrated. Intelligently. Transparently. And governed by design. 

LatentBridge: Built for the New Legal Workforce 

Whether you're building your first agent, scaling GenAI, or operationalising AI across legal teams—LatentBridge brings structure, speed, and scale to your journey.

We’re organised around two core engines:

• AI Labs – Home to our flagship platforms Epic AI and IntellixCore, built to simplify legal knowledge discovery and orchestrate complex agentic workflows with full transparency and governance.
• AI Centre of Excellence – This is where we help clients move from ambition to execution. Through AI Pods, we embed dedicated AI teams within enterprises and law firms to accelerate AI adoption on the ground. These Pods work shoulder-to-shoulder with your teams to build and deploy custom AI agents, virtual assistants, and domain-specific automation tailored to your legal and compliance needs.

From low-overhead pilots to modular deployments and trusted governance—we help legal teams not just adopt AI, but truly integrate it into the way they work.