The Legal AI Build Squad

The build vs buy question is divisive.

Folks on either side have their opinions. There are advantages and disadvantages to both approaches. The options are not necessarily mutually exclusive and there is a nuanced approach. In certain scenarios, buying makes more sense and in other scenarios, building makes more sense.

I wanted to go a bit deeper to clarify what building AI in legal actually looks like by unpacking the skills needed to effectively build AI in legal.

The insights come from my own experience of building AI in legal for close to a decade now at leading legal tech startups as well as at the innovation lab at Thomson Reuters, but also the work we have been doing at simplexico over the past few years designing use cases and building bespoke AI solutions for legal.

Part of the challenge of doing AI in legal is that decision-makers in legal who are making purchasing decisions on the buy side or funding decisions on the build side don’t have a good grasp of technology, how it can be delivered and the skillsets needed.

Most folks in legal paint a broad stroke with the term ‘developer’ or ‘engineer’ or ‘data scientist’ and assume there is a singular unified skillset. It’s becoming clearer to folks in legal that IT teams are not going to the AI teams accelerating a firm into the future. There is an gap in understanding the range of skills and abilities that exist within technology and what is required to effectively build AI in legal.

Case in point - hiring a bunch of data scientists into a firm and expecting them to deliver scalable and secure production-ready AI solutions is a gross mismatch of expectations and skills. While it might sound sexy and make for a great press release, firms are going to need more than just a bunch of data scientists to effectively deliver AI into a firm.

So then, what are the skills necessary to build AI in legal? What composition of skills are needed and how do they complement each other?

First, we need to introduce the idea of a Squad from Spotify.

The Squads model was introduced at Spotify and allowed them to scale their teams while staying agile. The model become popular and is well-known as an engineering best-practice model of how to structure nimble modular Squads across an organisation. In the engineering world, Spotify Squads are a model for organising product development teams to promote autonomy and alignment.

A Squad has end-to-end ownership of what they build across the software development lifecycle through design, development, testing, launch and maintenance. A Squad is a cross-functional team made up of 6-12 people with the core skillsets needed to autonomously deliver and actually build and deploy things without relying on anyone outside the Squad.

Back to us thinking about building AI in legal.

Is there a similar Squad model that we can use in the context of building AI in legal? What are the core skills needed to enable a team to autonomously deliver use cases in an agile way that can still be scaled effectively across an organisation? What does that cross-functional team look like?

At simplexico, we’ve put together a model of the Legal AI Build Squad - an autonomous, modular, cross-functional team that can deliver across the AI lifecycle from use case discovery, design and prototyping through to production, release and maintenance.

A Legal AI Build Squad consists of the following skills makeup and responsibilities:

1. Legal AI Architect
2. AI Engineer
3. Fullstack Engineer
4. Cloud Engineer
5. UX Designer
6. Subject Matter Experts (SMEs)

The Legal AI Architect role sits in the Project Manager seat and is responsible for managing stakeholders, ensuring delivery and more importantly providing a high-level picture of how everything fits together across the technology through to the domain in a build solution.

The AI Engineer role is responsible for everything on the AI side including deploying LLMs, orchestrating prompt workflows, finetuning models, creating data pipelines and running evaluation workflows to ensure performance.

The Fullstack Engineer role is responsible for providing the software integration layer taking care of the front-end user interface, developing backend APIs, integration with databases and connecting to third-party APIs and services.

The Cloud Engineer role is responsible for managing testing and deployment CI/CD pipelines, provisioning cloud resources and ensuring best practices for security and data processing on the cloud with private endpoints, encryptions and data residency as well as ensuring deployment scalability and security.

The UX Designer role is responsible for translating user requirements from Subject Matter Experts (SMEs) through user research interviews and workflow mapping into UX designs, mockups and design patterns that the rest of the technical team can implement and action. They are also responsible for conducting user testing, capturing user feedback and iterating on designs to ensure user-centric design of AI solutions.

The Domain Experts or Subject Matter Experts (SMEs) are domain expert practising lawyers who are the target end-users of the solution. They are responsible for providing input, sharing context and giving feedback to the rest of the team to guide them in an iterative manner to a useful solution. With their legal expertise, they are responsible for either creating or curating data that can be used as well as conducting evaluation reviews of the performance of AI tools so that there are minimum acceptable performance requirements.

Building AI in legal is complex but a pragmatic approach can make build efforts so much more effective.

A single Legal AI Build Squad can turn around use cases into prototypes across the firm and where it makes sense evolve those prototypes into production-ready solutions in reasonable timelines. A Legal AI Build Squad brings together the necessary skillsets into a single unit that can effectively and rapidly build AI solutions in legal to keep stakeholders engaged and create momentum around actually doing things AI in practice.

At simplexico, we have built out a Legal AI Build Squad team and rely on clients to provide SMEs for projects in a particular practice area or legal domain. We are sharing these thoughts to help others who are building AI in legal to learn from our mistakes and to create best practices, models and frameworks to help everyone get more out of AI in legal.

Need help building a bespoke AI solution for your law firm or legal team and want to work with a Legal AI Build Squad?

Book a Discovery Call here to learn more.

Thanks,

Uwais

Founder simplexico