AI adoption within legal functions is rapidly moving beyond proof-of-concept projects and isolated use cases. As organisations begin embedding AI into day-to-day legal work, the conversation is shifting from possibility to practicality.
The challenge is no longer simply understanding what AI can do. It is mabout making informed decisions about how to use it effectively, sustainably and at scale.
Questions increasingly facing legal leaders include:
These were some of the key themes explored in Radiant Law's recent webinar with Thomas Sittek, AI Lead at TT International Asset Management, hosted by Johanei Venter, Head of Delivery Operations at Radiant Law.
Drawing on experience implementing AI in highly regulated environments, Thomas shared practical insights on model selection, cost management, workflow design, governance and the organisational foundations required for successful AI adoption.
Here are the key takeaways.
According to Thomas, one of the most significant developments in recent years has been the emergence of reasoning-capable AI models.
Earlier generations of large language models were heavily dependent on detailed instructions and carefully engineered prompts. Success often depended on how well users could explain exactly what they wanted.
Today's models can reason through problems more holistically. For example, instead of simply updating a clause in a contract, modern reasoning models can:
This shift fundamentally changes how legal teams should think about AI implementation.
The challenge is becoming less about writing perfect prompts and more about ensuring AI has access to the right information, systems and context at the right point in a workflow.
As Thomas explained, organisations should increasingly think of themselves as a landscape of interconnected systems rather than isolated tasks.
One of the strongest themes throughout the discussion was that implementation is primarily an organisational challenge rather than a technical one.
Thomas described what he calls an AI Agility Framework, beginning with AI literacy.
This does not mean lawyers need to become engineers. Instead, teams need a practical understanding of:
The objective is less about technical expertise, and more about, confidence. Teams that understand the technology are more likely to use it responsibly, identify useful opportunities and engage constructively with change.
Importantly, AI literacy also reduces some of the resistance that often accompanies adoption initiatives.
When lawyers understand how the technology supports rather than replaces their work, engagement typically improves.
Many organisations rush toward AI solutions before fully understanding how their current processes work. Thomas cautioned against this approach.
Before introducing AI into a workflow, organisations need a clear understanding of:
This becomes especially important in large organisations operating across multiple jurisdictions. The same legal output may be produced differently by teams in different regions.
A workflow designed around one team's process may unintentionally create friction elsewhere.
Process mapping helps organisations identify these differences before implementation begins. It also delivers an additional benefit - the exercise often reveals process issues that have nothing to do with AI at all.
As organisations examine workflows more closely, they frequently discover opportunities for simplification, standardisation and efficiency improvements before introducing any technology.
One of the most practical sections of the discussion focused on a question many legal teams are currently wrestling with: Which AI model should we use for which task?
Thomas suggested evaluating use cases through three lenses:
Not every task requires the most sophisticated model available. Administrative activities can often be nandled by smaller, lower-cost models such as:
More complex activities may justify the use of more advanced reasoning models, such as:
The key is aligning capability with business need, rather than defaulting to the most powerful option for every use case.
Many organisations focus initially on software licence fees. According to Thomas, this is often the smallest component of the overall investment, and as adoption scales, additional costs emerge. These include:
Model Consumption
Many providers charge based on usage rather than fixed subscription costs. Higher volumes of prompts, larger datasets and more sophisticated models increase spend.
Data Integrations
AI systems frequently need access to:
Building and maintaining these connections has its own cost profile.
External Content Sources
Legal teams often rely on external providers such as legal research databases and specialist information services. Connecting these sources may introduce additional licensing requirements.
Time and Change Management
Perhaps most importantly, legal professionals require time to learn. Training, experimentation and adoption all carry operational costs. Successful implementation requires dedicated capacity alongside day-to-day legal responsibilities.
A recurring theme throughout the webinar was flexibility.
The pace of AI development remains exceptionally fast and the model that performs best today may not be the market leader six months from now. For this reason, Thomas encouraged organisations to avoid tightly coupling workflows to individual models wherever possible. Instead,they should aim for architectures that allow providers and models to be swapped relatively easily.
The objective is to maintain optionality. Future flexibility allows legal teams to:
In a rapidly changing market, flexibility may become just as important as capability.
One of the clearest messages from the session was that AI should not become a separate destination but a part of existing ways of working. Lawyers are unlikely to embrace tools that require them to leave their core applications and repeatedly copy information between systems.
The most effective implementations reduce friction - and this often means integrating AI directly into environments such as:
The easier the technology is to access, the more likely adoption becomes.
Perhaps the most important lesson from Thomas Sittek's session is that successful AI implementation is rarely about the model itself. The organisations getting real value from AI aren't the ones with the biggest tech budgets or shiniest tools.
They're the ones focused on fundamentals:
For legal teams, the next phase of AI adoption is less about experimentation and more about operationalising technology in a sustainable way. That means building legal functions that can confidently evaluate new tools, integrate them into existing processes, manage risks appropriately and continuously improve how legal work is delivered.
This is where many will begin to encounter a broader challenge: not whether AI works, but how to embed it into legal operations, workflows, governance frameworks and service delivery models in a way that creates lasting value.
The organisations seeing the strongest results are typically those that approach AI as part of a wider legal transformation effort, combining process improvement, operational design and technology implementation to create more scalable, efficient and resilient legal functions.
You might be wondering what your next steps should be. Let us guide you with three easy options:
We use cookies to monitor the performance of our website and to improve user experience. By continuing to browse our site, you agree to our use of cookies.
