Welcome to Spaarke

The first wave of legal AI was a wave of attorney-task tools. A contract review assistant. A tool to redline drafts. A summarizer for long-form documents. A chat surface that could answer a question about case law without losing the citation halfway through. Each of these is genuinely useful. None of them changes how a legal department actually runs.

The next wave is something different, and it is not a tool at all. Drafting faster does not change cycle time on an NDA when intake is informal. Reviewing redlines faster does not change spend visibility when invoices arrive in fourteen formats from twelve firms. Summarization does not change matter handoffs when the rationale behind a settlement strategy still lives in the head of the attorney who is leaving. The tool is not the bottleneck. The operating model is.

The operating-model gap shows up as concrete operational reality. Overloaded legal teams with no surface to triage from. Lost institutional memory — decades of negotiation rationale walking out the door when senior staff leave. Outside-counsel coordination chaos: the wrong version, the wrong CC list, the missed approval on the final draft. Matter context fragmented across seven systems that do not share a record. None of this gets fixed by a faster drafting assistant. It gets fixed by a different kind of platform.

What we mean by Legal Operations Intelligence

Spaarke is a Legal Operations Intelligence platform. The definition changes the legal ops perspective: a system of intelligence where legal work, data, and AI run together across every matter, every document, and every side of the engagement.

Operating model is the load-bearing phrase. A matter-management system tracks matters. A document-management system holds documents. An e-billing platform processes invoices. Each is partial. The legal department that runs on seven of them connected by spreadsheets and email has tools but no operating layer above them — nothing that connects what was previously disconnected, remembers what was previously forgotten, and surfaces what was previously invisible. That layer is what we call Legal Operations Intelligence, and the long-form case for it lives in What is Legal Operations Intelligence?. The architecture that runs underneath — Data, Memory, Inference — lives in The Legal IQ stack.

Operational intelligence is not productivity. Productivity is what individual tools sell. Operational intelligence is the layer above — visibility into spend, counsel performance, matter status, and outcomes, with AI directing and humans controlling. It is a different category, and the rest of this piece is about why.

AI-directed, human-controlled

The philosophy of the platform compresses to four words: AI-directed, human-controlled. They are not a hedge, and they are not transitional language. They are the architecture.

AI-directed means agents take bounded actions inside the platform. Foundry IQ — the grounding and memory layer — indexes matters, documents, emails, tasks, and counsel-and-firm history into a structured source so AI features operate on your department's actual data rather than on generic web text, and so what the platform learned yesterday informs what it does today. The Microsoft Agent Framework — the orchestration and execution layer — turns playbooks into running automations: opening matters, routing invoices, drafting via Word Copilot, posting to Teams. Microsoft is consolidating its agentic stack here, which is where Spaarke builds. Microsoft Power Platform and Microsoft 365 — the experience layer — are where users actually meet the platform: Outlook, Teams, Word, SharePoint, M365 Copilot, Power BI, and the model-driven apps Spaarke composes on top. We name what each layer does because "AI-powered" had no useful meaning by 2025 and has even less now.

Human-controlled means the parts of legal work that are not delegable stay with the lawyer. Judgment, privilege, accountability, the call on whether to settle or to push, the question of whether a redline is acceptable — none of those are agent decisions. Agents read grounding, take bounded actions, and surface what they did to a human. The human reviews, corrects, signs.

This is not a transitional architecture. Conservative legal buyers ask, by default, how human judgment is preserved in any AI-enabled workflow, and they are right to. Professional work where the consequences land on a person's name and a firm's reputation is not delegable to a model. The interesting design problem is not "how much can the agent do" — it is "what should the human always own, and how does the platform make that boundary structural rather than aspirational." Spaarke is built around that boundary.

Why your Microsoft tenant matters

The third commitment is structural, and it is where most legal-platform conversations get vague. Spaarke runs inside the customer's own Microsoft tenant. Not adjacent to it. Inside it.

Most legal platforms integrate with Microsoft — an OAuth handshake, a webhook, a connector, a quarterly batch sync. Spaarke runs on Power Platform, Dataverse, and SharePoint Embedded, deployed into the customer's M365 tenant. Identity comes through Microsoft Entra ID. The same conditional access, DLP, sensitivity labels, retention policies, and audit logging that govern the rest of the customer's M365 environment govern Spaarke automatically — because Spaarke is in that environment, not next to it. Documents live in SharePoint Embedded, with native versioning, metadata, and co-authoring. Cross-matter analytics run in Power BI on the Dataverse model. AI is grounded by Foundry IQ, surfaces in M365 Copilot, and never leaves the tenant boundary your security team already validated.

This is not a deployment preference. It is the precondition for the operational intelligence we claim. AI becomes operationally valuable only when it is grounded in the matters, documents, and spend the customer actually has — and that grounding only holds up if the data sits in a place the customer governs. Customers who are tired of multi-tenant SaaS hosting their most sensitive material — privilege, M&A, regulatory — already know why this matters. The full version of the argument lives in Why we built on Microsoft; the short version is that the security perimeter is the one IT already approved, and the legal operating model should not require IT to approve a second one.

From tools to operating models

The first wave of legal AI sold productivity. Drafting faster, reviewing faster, summarizing faster. Useful, and now widely available — among in-house legal departments using AI, generative-AI use roughly doubled in a year, from 23% in 2024 to 52% in 2025. Productivity AI saturates because productivity is a per-task improvement, and per-task improvements are easy to copy.

Operational AI compounds. It compounds because it operates on the organization's own data, inside its security boundary, on its own processes, across its own history of matters. The decisions it informs feed back into the data it operates on. Each new matter sharpens the memory of every similar matter that came before. The platform learns. That is what we mean by an operating model — a system that gets measurably more useful over time because it captures, retains, and reasons about how the department actually works.

A few practical consequences fall out of this. Spend visibility stops being a quarterly spreadsheet and becomes a forecast against the department's own history. Matter handoffs stop being an attorney-to-attorney apprenticeship and become a structured transfer of context. Outside-counsel performance stops being a partner's gut feel and becomes a comparison against two hundred similar matters the department has already handled. None of this is "the future of legal." It is the operating model the rest of the enterprise has been running for years. Legal is not unique in needing it. It is unusual only in not yet having it.

The Intelligence Era of Legal Operations

Today we are announcing Spaarke.

Spaarke is a Legal Operations Intelligence platform — a Microsoft-native system where legal work, operational data, documents, and AI run together across every matter, every document, and every side of the engagement.

But Spaarke is more than software. It is also a delivery model: legal operations, technology, and AI specialists working alongside legal departments and law firms to improve how legal work actually runs. And it is a long-term commitment to the legal community itself — participating in the operational, technical, and architectural discussions shaping the next generation of legal infrastructure.

That matters because the field is changing quickly. AI capabilities are evolving monthly. Enterprise expectations around governance, grounding, tenant architecture, and operational control are changing with them. Legal departments are not simply adopting new tools; they are redefining how legal work is organized, executed, measured, and remembered.

We do not think product development can happen separately from that conversation.

As part of the Spaarke launch, we are also establishing an ongoing research and writing practice focused on Legal Operations Intelligence and the systems emerging around it. The point is not content marketing. The point is to publicly work through the ideas, tradeoffs, and implementation realities behind this transition as the market evolves.

To delve deeper, we recommend reading What is Legal Operations Intelligence? — the category framework this platform is built around. If you want to understand the architectural foundation underneath it, read Why We Built on Microsoft. And if you want to see how these ideas translate operationally, explore the platform modules and deployment models that make up Spaarke itself.

Legal operations is entering its intelligence era. Spaarke is being built for that transition.