A different kind of issue this week. A founder letter, marking how Advanta's work has changed shape over the last several years and what the firm is now. It is longer than a regular issue and lighter on the usual signal-grade analysis. The full intelligence cadence resumes next Tuesday, with a clearer sense of what the cadence is for.
Over the last three years, I have spent a great deal of time inside Legal AI portfolios. Procurement, programme governance, post-deployment review.
One pattern has appeared with unusual consistency.
Capability is arriving on schedule. Models improve quarterly. Vendors ship faster than procurement can evaluate them. The architecture to operate any of it — the governance, the evidence trail a board or regulator would accept — is not arriving on the same schedule. In most institutions it is not arriving at all.
That asymmetry is the thing I have been watching widen. The pattern is not a failure of any specific firm. It is a structural feature of how the category is forming.
The realisation, for me, has been gradual. Early on, I assumed the binding constraint was tooling. Pick a vendor, run a pilot, and the function would begin to operate AI capably.
I no longer believe that. Tooling has become the easy part.
I now see Legal AI transformation as, at its core, an operating systems problem. The constraint is not which model the firm chooses. The question is whether the firm has the architecture to govern what is deployed, measure what is produced, and defend the system under scrutiny.
Capable AI without governance produces shadow systems and regulatory exposure. Capable governance without measurement produces compliance theatre. Strategy, governance, execution, and measurement operate together, or they do not operate at all.
That realisation has changed Advanta.
When I started the firm, the work was straightforwardly advisory. Procurement reviews, pilot design, vendor diligence. Useful work, but bounded.
Around eighteen months in, I began to notice that the engagements producing durable change were not the ones with the cleanest tooling decisions. They were the ones where the operating spine got built alongside the deployment — a defensibility position, a governance cadence, a way of measuring the function a year after the contract was signed. The advisory work that did not produce that spine produced pilots that quietly ran out of energy by the second year.
The firm reorganised around that observation. Advisory remained, but it began to operate against something larger underneath it: a framework, a maturity instrument, an intelligence base, and a library of operating artefacts that legal functions could deploy inside their own environments.
Advanta has been building, over the last several years, the operating layer that sits between vendor capability and institutional readiness. The work has always pulled toward this. I was slow to see it.
Today's surface area is the result of that reorientation. Some of it has been visible for a while. Much of it has only recently moved from internal work into the public site.
The framework has a name now. Legal AI OS: a canonical operating standard for legal organisations running AI, built around eight pillars and six layers. The methodology is published and versioned, and the advisory work runs against it as the substrate.
Around it sits a set of supporting instruments. A free baseline diagnostic that locates the organisation against a five-band maturity stack. The Intelligence Hub, where the anchor essays and the briefs live. A library of operational modules. Governance instruments — a defensibility posture, a risk taxonomy, an AI lifecycle, an AI bill of materials — that the function can adopt and operate.
A book is coming. Legal AI OS, first edition 2026. The operating manual, written for general counsel and heads of legal operations who need to make decisions that hold up at board and regulator level. The waitlist is open.
None of this was the plan three years ago. The work asked the firm to build it.
Over the coming months, this newsletter will look different.
There will be more original research. Less commentary on the news cycle. The cadence is weekly — a flagship Tuesday issue, with shorter signal-grade notes between.
I will share previews from the manuscript as it moves toward publication, governance work the advisory practice produces, and maturity intelligence drawn from the diagnostic cohort. Some of this will eventually sit behind a subscription tier. Most will remain open.
Underneath the cadence is a simpler intention. To make Advanta the institutional intelligence base for Legal AI — the place general counsel and heads of legal operations come when they need a defensible answer to a question about AI, before they ask their vendors.
That is the work. It will continue.
If you have been subscribed for a while, thank you for staying. The shape of the work has changed, and I wanted to mark that change directly rather than let it land sideways through a website rebuild.
If you are reading for the first time, welcome. There is more here than there used to be, and the next several months will show you what the firm is now.
Either way, I am glad the conversation continues.
Nishant Bhaskar · Founder, Advanta
Continue from here
01 · Understand the framework
Legal AI OS — the operating standard for legal organisations running AI. Eight pillars and six layers across the operational surface of the function. The maturity stack and the defensibility architecture the framework is built around.
02 · Assess readiness
Free Baseline Diagnostic — a 5-minute strategic assessment. Twenty questions. Locates the organisation on the five-band Maturity Stack and produces a personalised gap analysis across all eight pillars.
03 · Explore the intelligence layer
The Intelligence Hub — anchor essays, executive briefs, and operating-model analysis. No vendor advocacy. No surface-level coverage. The institutional intelligence the advisory work produces.
Share this issue