Tim Golden

Direct client consulting

In my final year at a global Am Law 100 firm, I worked directly with clients as a consultant through the firm's legal service-delivery innovation program, applying data and technology to legal-service delivery and business problems.

That work included developing a matter-pricing engine that produced data-driven menus of legal services, reconciling hundreds of client-required and firm-created task lists into a standardized set, and building text-mining that mapped historical time entries into those categories so pricing could draw on actual delivery history. I was the only non-attorney on the seven-person team.

It was not the beginning of my client-facing work — I did billable technology consulting for firm clients between 1997 and 2000 — but it was where I learned how a consulting relationship differs from an internal one.

Enterprise governance and data leadership

Before AI, I spent a decade on the unglamorous half of this work: IT governance, enterprise architecture, executive reporting, and security responsibilities including vulnerability management and third-party audits. Master data management and independent data validation came later, alongside the AI work.

I created a global Am Law 100 firm's first formal IT governance team and a national one's first Project Management Office. Both were cases of building an organizational capability that did not previously exist — which is considerably harder than advising one that does, and is most of what AI governance actually asks of a company right now.

At a global Am Law 50 firm I held responsibilities associated with a $44 million global IT budget; at a global Am Law 100 firm I managed a worldwide IT budget of nearly $40 million. I mention it because governance decisions that ignore cost and capacity do not survive contact with a budget cycle.

Hands-on AI delivery

I have been running enterprise LLM infrastructure since late 2022, starting when GPT-3.5 was released. That has meant launching and leading a firm's generative AI program from inception, architecting centralized model access, building custom applications and knowledge workflows, leading enterprise platform rollouts with identity and access governance across multiple regions, integrating AI with the systems people already use, and building the adoption reporting that told leadership what was actually happening.

I care a great deal about whether what I build can be handed over. I have written close to a hundred internal software packages so that work could be reused rather than rebuilt, and produced technical specifications complete enough for a separate team to reimplement a system in a different language. That instinct — that delivery is not finished until someone else can run it — is the same instinct behind good governance.