Matters: The 3 Challenges Facing CIOs Today

CIOs sit at the fault line of competing demands. Clients want the gains of generative AI (faster drafts, tighter diligence, sharper risk detection) without sacrificing privilege, confidentiality, or control. Some clients push for broad adoption; others prohibit it entirely. Many require matter-by-matter constraints on where data is processed, how long it is retained, and which models may touch it. Internally, partners and associates bring preferences shaped by practice, training, and specialization. What sounds like a single question, “Can we use AI safely?” fractures into thousands of policy variants that must be enforced with precision across thousands of clients, thousands of lawyers, and hundreds of tools.

The practical consequence is a widening governance surface. Every new tool, integration, or model adds another place where policies must be expressed, enforced, and audited. The CIO’s job is no longer only to enable access; it is to encode obligations like ethical walls, client-specific restrictions, and regulatory constraints into systems that were never designed to carry them natively. The stakes are high: A single misconfiguration can compromise confidentiality, derail a deal timeline, or erode client trust accumulated over decades.

Beneath this complexity sit three enduring problems the firm must solve well and repeatedly: (I) protect client data, (II) organize data so the right context is always at hand, and (III) leverage that data to best compete. Each is hard on its own. Together they define whether AI compounds value or compounds risk. Crucially, the point where all three converge has always been the matter. Matters are the unit that keeps data correctly siloed, the scaffold for organizing the firm’s context, and the arena where performance is measured and relationships are won. When work, data, and controls align to the matter, confidentiality holds, context is complete, and gains from AI can actually be demonstrated.

This series explores the three core problems that define these challenges, and why they are exacerbated with generative AI: (I) Ethical Walls (protect client data), (II) Firm Differentiation (leverage data to best compete), and (III) Fragmented Context and Tools (organize data for a context-complete workspace). We will also preview a path forward: a matter-centric approach that respects ethical walls, unifies context for GenAI, and creates more avenues for firm-specific differentiation.

Permissions and access controls sit at the heart of a law firm’s professional obligations. Each time a new product is introduced — legal research, a document-management add-on, a workflow tracker, or a generative AI assistant — an entirely new set of access rules must be created and maintained. These rules rarely map cleanly onto the firm’s existing structures; they must be configured manually, reconciled with multiple legacy systems, and monitored for consistency over time. The burden falls on small teams already stretched thin, pulling them into a reactive cycle of patching rather than building sustainable solutions.

Ethical walls add another layer of complexity. Each new environment is yet another place where client confidences must be walled off, conflicts prevented, and access tightly controlled — yet no two systems enforce these rules in precisely the same way. A wall configured in the DMS may not translate directly to the AI research tool or the diligence tracker, leaving gaps that must be managed through duplicative processes. Over time this creates a labyrinth of overlapping controls that are difficult to audit, harder to explain, and nearly impossible to guarantee with absolute confidence. GenAI can make this worse by multiplying surfaces and enabling cross-source queries that, if misconfigured, traverse boundaries that were never meant to intersect.

The operational risks are immediate and material. A junior associate might mistakenly be granted access to a restricted workspace; a partner could be locked out of critical diligence materials delaying a deal closing; a misaligned setting in an AI platform could allow queries across matters, raising questions about confidentiality and exposing the firm to liability. As matters evolve — teams expand, co-counsel join, side letters impose bespoke permissions, lateral hires arrive — rights must be updated everywhere, across systems that don’t talk to each other. What should be a core competency (strict control over who can see what) has become a fragile operational weak point, sustained by manual vigilance and after-the-fact audits.

The practice of law has always been defined by context. To draft a contract, negotiate a deal, or argue a case, lawyers need the right mix of documents, prior work product, case law, discovery material, business context, and client correspondence. Today, that essential information lives across a patchwork of disconnected systems: DMS platforms, email inboxes, legal research databases, diligence trackers, knowledge repositories, and collaboration tools. Without that mix in one place, work product risks being incomplete or incorrect.

Fragmentation carries real consequences. When context is spread across multiple systems, the risk of error rises, compliance becomes more difficult, and costs grow. For every matter, lawyers reconstruct context by pulling emails, uploading documents, running searches across different platforms, and trying to remember which system is appropriate for which task. Associates can spend hours assembling context before substantive work begins — inefficiencies that slow progress and increase the likelihood that critical details are missed, undermining both quality and client trust.

GenAI raises the stakes. Missing context doesn’t just delay work; it produces confidently wrong answers that can mislead the lawyer. The very promise of AI — to accelerate drafting and analysis — collapses when the system is starved of what it needs to reason correctly.The target state is a context-complete workspace where matter-specific documents, knowledge sources, communications, and workflows are discoverable and queryable together, without compromising the ethical walls that govern access.

The goal of every firm is simple to state and hard to achieve: get better at completing matters. In a GenAI world, differentiation depends on three capabilities that reinforce one another: measuring performance, organizing and leveraging data, and training talent to work effectively with new tools. Firms that master these will demonstrate value more clearly to clients while improving speed and quality internally.

First, measurement. Can you measure the performance of attorneys and AI on client matters to prove ROI, benchmark matter performance holistically, and show clients the benefits of AI deployments? Without instrumentation across workflows and outcomes, firms can neither improve nor credibly claim the ROI from adopting AI tools broadly. In-house customers have rapidly gone from "Please don't use AI on our work" to "Please demonstrate the explicit value you're creating with AI and how it's saving me time and money on outside counsel.” That pressure is unlikely to wane anytime soon, so it's on firms to prepare accordingly.

Second, data leverage. Can you organize historical precedent, workflow exhaust, and market data so that GenAI can actually use it at the point of work? Fragmented data remains inert; structured, accessible data becomes an accelerant. The feedback we hear from customers is not that they need more data — it’s that they need it delivered inside their existing workflows, through connected applications, where it drives decisions and automates routine judgment. But data leverage alone isn’t enough; the next frontier is collaboration. In a GenAI world, the firms that win will be those that collaborate most effectively with their clients. Today’s legal tools are largely single-player — built for internal teams rather than the multi-institutional reality of modern legal work, with no true “multiplayer mode” that connects in-house teams and outside counsel in a shared workspace. The opportunity ahead is to create that environment, where GenAI, data, and client collaboration compound one another’s impact to turn legal work into a truly networked enterprise.

Third, talent and training. Upskilling attorneys — and especially developing new partners who are fluent in AI-assisted practice — turns tooling into capability. GenAI compresses commodity work and raises expectations; the firms that differentiate are those that align people, process, and data so lawyers can operate at the top of their license. Without this alignment, GenAI can make things worse by amplifying noise, fragmenting workflows further, and obscuring accountability.

Solving these problems requires a matter-centric GenAI platform that respects ethical walls, unifies fragmented context, and leaves room for firm-specific “secret sauce.” Put differently: bring together all the relevant context needed to answer a question or produce work — without exposing it to the wrong user. The challenge is to create a context-complete workspace for each user’s needs at scale while guaranteeing strict data isolation.

Today, the burden often falls on users and admins. Lawyers must assemble context themselves — crafting prompts, finding and uploading the right documents, selecting knowledge sources, choosing workflows, and deciding which product surface to use for a particular matter type — while admins spend time managing permissions, tracking usage across tools, and troubleshooting how lawyers should gather context. The result is a fragmented experience where errors arise not from bad lawyering, but from missing context and uneven access controls — exactly the patterns that GenAI can exacerbate if left unaddressed.

Matter OS is our path to change that dynamic. It turns Harvey into a single, matter-centric, context-complete workspace: Lawyers can enter a “Matters” view, select a case, and work in a pre-assembled environment where vaults, precedents, workflows, correspondence, and research are already connected — and queryable — within the right ethical walls. For lawyers, that means no more hunting across legacy systems; for admins, permissions and walls are built in from the start; for CIOs, it offers a safer, simpler, more customizable way to manage how matters, data, and workflows are shared across the firm. In the blog posts that follow, we’ll unpack each problem — and how a Matter OS approach addresses it — in depth.