The Builder Shift: Turning Individual Productivity Into Institutional Capability

Almost every lawyer now has access to AI. What they do with it varies enormously, between organisations and within them. Some lawyers never even log in. Others use it every day, on almost everything. I’ve spoken to many who even said they would pay for AI tools out of their own pocket if access was taken away.

Of particular interest is a small but growing group that has moved beyond just using the tools and now builds with them. They’re creating agents, playbooks and reusable assets for themselves, and for others to rely on.

This transition from consuming to creating is where individual productivity becomes an institutional capability. In our new research for Harvey, we label this transition the “Builder Shift”.

Where This Shift Comes From

TITANS conducted independent research into how legal organisations define, measure and build AI fluency. Commissioned by Harvey, the research draws on in-depth interviews with legal technology and enablement leaders at Harvey customers: seven law firms and four in-house teams, spanning the UK, US and continental Europe.

In addition to the interviews, several of those organisations provided product usage data for analysis or reviewed our research outputs against their product analytics directly. The full findings are published in The AI Fluency Framework report.

The research is also rooted in our own first-hand experience. Over seven years, TITANS has worked with a fifth of the UK’s top 50 law firms, several AmLaw 50 firms, the Big Four, private equity-backed ALSPs and global corporates. That work includes some of the largest Harvey rollouts globally, and award-winning AI fluency programmes that have trained thousands of lawyers. The patterns and trends we observed informed our research questions. This provided the foundation for the leaders we interviewed to validate, challenge and sharpen the frameworks that appear in the report.

Five Levels of Fluency

At the centre of the report sits a five-level framework describing an individual lawyer’s fluency in applying AI to legal work.

The five levels sit within two broader categories. Levels 1–3 reflect individual productivity. This is the journey from the Dormant lawyer who rarely logs in, through Functional use for summarising and simple drafting, to the Proficient lawyer who works with AI daily on substantive legal matters, iterating and verifying for reliable output.

Levels 4 and 5 are where individual fluency becomes an institutional capability. Here Builders and Catalysts create reusable AI assets like agents and playbooks that can carry their expertise to others. As Pierre Zickert of Hengeler Mueller put it, “at some point their activity becomes an institutional asset.”

Between those two categories sits the framework’s key transition: the passage from Level 3 to Level 4, from consuming AI to creating with it.

What Changes at Level 4

The Builder Shift reflects a new mindset as much as a new skill set. A Proficient lawyer already has most of the capability a Builder needs: they use AI confidently on real legal work, choose the right feature for the task, and often chain features together into flows.

But builders also have a process mindset. They see their work differently, looking at a matter and seeing inputs, steps and outputs. They think about processes, not documents. Where a Proficient lawyer may ask “how do I get the best output for this task?”, a Builder asks “how do I make this whole task repeatable, for me and for the people around me?” Andrei Salajan of Schoenherr describes the change: “…You are thinking about how you can reshape end-to-end processes. So it’s less the technology per se… It’s more of a design thinking, process thinking, creative thinking.”

Once a lawyer crosses that line, their impact can increase dramatically. RSGI’s research on power users, commissioned by Harvey, describes users who would sit at the top of Level 3 and beyond. They found that these users save about twice the hours of the typical user. A Builder’s impact can spread even further. Because their knowledge is codified, the assets they create carry their judgement to others, multiplying the reach of their expertise.

Why the Crossing is Hard

We noticed a pattern across the organisations we interviewed: the Builder Shift is often the largest step to make and the least consistently achieved transition on the fluency curve. I see this play out regularly across my work, where there is a strong desire to build but the foundational requirements are not yet in place. Every organisation is inherently different, and the approaches taken to achieve this transition must reflect the realities of the organisation’s environment.

Building typically requires specific user permissions that many organisations are careful in granting, often managed on a user-by-user basis. Groundwork in data and governance must be laid first. “It’s a bit of a catch-22,” as Anne Wong of SUSE told us: “to build an effective agent, you need to have the foundations in order.”

Incentives also play a key role. In law firms, billable hour pressure can place a hard ceiling on the Proficient users who may otherwise experiment their way into building. The organisations producing Builders at scale have tackled this head-on, with some firms allocating creditable time that can be used for learning and practising with tools.

Building involves experimentation, some of which may not work. Making space for that experimentation is a strong indicator that an organisation is serious about enabling Builders.

Governance That Scales With Reach

Once building is unlocked, new problems rapidly emerge. The challenge flips from encouraging people to build, to managing an influx of built assets. Organisations quickly discover crowded libraries, duplicate agents and unmaintained builds, often with unclear long-term ownership.

The organisations handling this well use tiered governance that scales rigour with the reach of a built asset. Alex Bazin of Lewis Silkin captured the model:

“If you’re doing stuff just for yourself, that’s fine. If you’re doing it for your team, you can build it, but there’s a check on that. And if you’re building it for the whole firm… now you do need to come to us and we’ll validate it.”

This tiered approach keeps the level of oversight proportionate to the risk involved, making it easier to open building up. And when the environment is right, the shift itself can happen fast. Pierre Zickert states: “Between level three and four, that [shift] goes quite quick. However, it needs permissioning on our side… that’s a user-by-user decision.”

Having suitable structures in place enables engaged Builders to flourish. Sarah Chambers of Ashurst Perkins Coie sees it every week: “There are the people that come to us every Monday morning and say, ‘oh, I built this thing at the weekend, can we productionise it?’.”

Not Everyone Needs to Build

The goal is not to turn every lawyer into a Builder. Not everyone needs to build, and not everyone wants to. The universal target is for all lawyers in an organisation to reach Proficient (L3). Alongside that, the organisation needs a sufficient cohort of Builders to support the wider organisational objectives.

One law firm leader we interviewed put the trade-off like this:

“If we feel it’s more beneficial to have 20 lawyers spend half their time on AI workflows, that’s equally acceptable as getting 100 users to spend a small portion of their time building workflows. The goals are the same.”

Advanced organisations are creating the conditions to enable their Builder cohorts. With the permissions, incentives and governance covered above in place, those Builders allow the firm’s collective expertise to accumulate and scale.

Where to Start

If you take one action from this research, start by measuring fluency in your organisation. Place your people on the five levels by reviewing product usage data and the qualitative signals associated with each level as outlined in the report. For Harvey customers, the easiest way to do this is in Command Center. A prompt describing the five levels and their product signals will show you how your users are distributed across each level.

Then focus on your Level 3 population to locate your enthused future Builders. In my experience, the biggest barrier is often the environmental conditions around the lawyer rather than their capability. Determine whether they have the product permissions, incentives, governance foundations, visible backing and suitable support to grow a process mindset and make the Builder Shift.

The full report sets out the complete framework, including detection signals for each level, a Harvey feature map, a learning curriculum, a training matrix and the team rituals that sustain fluency well beyond training sessions.