Sullivan & Cromwell AI Playbook: How a 145-Year-Old Firm Is Deploying Generative AI in 2026

TL;DR#

Sullivan & Cromwell ai adoption is the most-watched bellwether in BigLaw because S&C is the prototypical white-shoe firm — founded in 1879, profit-per-equity-partner above $6.7M, advisor of record to OpenAI on the Microsoft partnership, and historically described by legal-tech analysts as "conservative at best" in software adoption. In 2026 the firm runs a real AI program: mandatory training tied to tool access, an Applied Intelligence practice that pairs lawyers with data scientists on fixed-fee engagements, and active GenAI use across M&A diligence, drafting, and closing-checklist work. The firm also became the most cited cautionary tale of the year after an April 2026 emergency filing in the Prince Group bankruptcy contained fabricated citations and forced a public apology to a federal bankruptcy judge. The story matters beyond Sullivan & Cromwell: when even a 145-year-old institution moves generative AI from pilot to production, the rest of BigLaw has effectively already shifted. What hasn't shifted yet — for S&C or anyone else — is the front of the funnel, where conflicts checks and new-matter intake still run on PDFs and questionnaires. That is the next domino, and it is where conversational AI changes the institutional math.

What Sullivan & Cromwell's AI Footprint Actually Looks Like in 2026#

Sullivan & Cromwell's AI stack in 2026 is organized around three layers: client-facing advisory, internal practice tooling, and risk governance. The firm publicly advises clients on AI governance, IP, privacy, and regulatory exposure through a named Artificial Intelligence Practice, and represented OpenAI in the Microsoft partnership — work that gives the firm an unusual front-row view of how frontier-model contracts and governance actually get written.

Inside the firm, S&C published a January 2026 memo on the Use of AI Tools in M&A Transactions that effectively functions as a market signal. The memo names six concrete deal-cycle uses: summarizing diligence deliverables, drafting agreements from precedent, building issues lists, benchmarking market-standard terms during negotiation, generating signing and closing checklists, and managing signature pages. None of those uses are theoretical at S&C; the memo reads like documentation of an active workflow.

The third layer is the Applied Intelligence practice, which pairs associates and partners with data scientists to build bespoke models and workflows for individual mandates — frequently on fixed-fee structures rather than hourly billing. That structural choice matters more than the technology choice. Fixed-fee AI engagements are an admission that the billable hour does not survive contact with token-priced automation, and that the firm's profitability has to come from leverage and judgment rather than time.

Why an S&C Adoption Curve Is a Bellwether, Not a Headline#

When Sullivan & Cromwell goes conversational, the directional question for the market is settled. The reason is institutional rather than technological. S&C has roughly 1,000 attorneys, a 145-year operating history, a partnership culture that treats hallucinated citations as career-ending failures, and a client list — sovereigns, central banks, the largest US issuers — that punishes reputational mistakes more severely than any other vertical of the economy. Firms with that risk profile do not adopt new infrastructure to look modern; they adopt it because the cost of not adopting has become higher than the cost of governance.

Compare the curve to what is happening elsewhere in BigLaw:

• Paul, Weiss, Rifkind, Wharton & Garrison began piloting Harvey in 2023 and embedded AI tools into daily practice years before S&C published its M&A memo.
• Quinn Emanuel built an internal litigation platform on Anthropic's Claude — a deeper engineering commitment than most firms attempt.
• Clifford Chance announced headcount reductions tied to AI productivity gains in 2025, the most explicit acknowledgment to date that AI is reshaping leverage models.

S&C is not the first mover. It is the lagging indicator that the move has happened. Our Latham & Watkins AI adoption analysis and the Kirkland & Ellis AI strategy breakdown describe scrappier rivals that are several quarters ahead operationally. The S&C signal is about the floor of the curve rising, not the ceiling.

What 145 Years of Institutional Culture Means for AI Rollout#

A century-plus of partnership history changes how an AI program gets built. Three constraints shape every rollout decision at a firm like Sullivan & Cromwell.

Governance gates everything#

Tool access at S&C is conditioned on completing mandatory training that, in the firm's own language after the April 2026 hallucination incident, instructs lawyers to "trust nothing and verify everything." That is not marketing copy — it is the policy that the partner co-leading Global Finance & Restructuring cited in his apology letter to the bankruptcy court. White-shoe firms run AI like they run a new associate: the technology can do the work, but a partner signs the brief.

The downstream consequence is that procurement at S&C does not look like procurement at a Silicon Valley startup. Every generative-AI tool — internal model, vendor product, or third-party integration — has to pass through information governance, conflicts, professional responsibility, and client confidentiality reviews before it sees a billable matter. That review pipeline is the single biggest reason elite firms appear to move slowly on AI even when they are spending seven figures internally.

Conflicts are a data problem, not a workflow problem#

S&C's conflicts database is one of the firm's most sensitive assets. It encodes 145 years of representations across hundreds of thousands of entities, deal counterparties, sovereigns, and family offices. AI tools that touch that data — for matter intake, due diligence, or counterparty screening — have to be deployed inside the firm's perimeter, not on a vendor's cloud. This is one of the structural reasons elite firms have moved toward in-house Applied Intelligence teams rather than relying solely on third-party SaaS.

Knowledge management is the moat#

S&C's competitive advantage is institutional memory: which clauses survived which negotiations, what every regulator across 40 years has signed off on, which precedents the firm itself wrote. A well-trained model that surfaces that memory to a fourth-year associate compresses ten years of pattern recognition into a query. A poorly trained model leaks it. Every senior partner at the firm understands both halves of that trade, which is why knowledge management — not chat — is the first-class AI application inside white-shoe firms.

The Front Door Is Still Broken#

For all the work happening inside the partnership, the front of the funnel at S&C — and at every comparable firm — still runs on the same intake architecture it used in 2005. New-matter forms, conflicts questionnaires, engagement-letter templates, and pre-onboarding KYC packets are PDFs and web forms. That is also where the institutional asymmetry shows up: an associate spends an hour reformatting client-provided diligence into the firm's templates; a partner spends fifteen minutes asking the same five clarifying questions on a kickoff call that the form did not capture.

The argument we have made consistently — see why law firms are replacing forms with conversations in 2026 and the full intake playbook for AI client intake — is that intake is the highest-leverage AI surface in a firm because it touches every matter, every client, and every partner before billable work begins. Conversational intake at scale changes three things at once:

1. Conflicts checks complete faster because the AI captures entity disclosure as the client speaks, in their own language, rather than waiting for a form to be returned.
2. Engagement scoping becomes accurate because the system probes "it depends" answers — the messy uncertainty that PDFs flatten — and surfaces scope risks before fees are quoted.
3. Knowledge management improves because every intake transcript becomes a structured asset for the firm's matter graph, not a buried email thread.

For Sullivan & Cromwell specifically, intake automation is also a conflicts-engineering problem. A conversational interview can be tuned to capture beneficial-owner trees, family-office relationships, and corporate-affiliate webs that a checkbox form will never elicit. That is a white-shoe-grade problem, not a SaaS-grade problem, and it is the natural next AI surface after M&A diligence and drafting.

Contrast With Scrappier Rivals#

S&C's institutional caution is the right strategy for S&C. It is not the right strategy for every firm. Compare three adjacent posts in our cluster:

• The Cooley AI strategy for a startup-focused firm replacing intake forms describes a firm whose clients expect product-grade onboarding from their lawyers — and who will switch firms if a competitor delivers it first.
• The Wilson Sonsini AI strategy for Silicon Valley founder intake plays the same game on the West Coast, where the cost of friction is higher than the cost of automation by an order of magnitude.
• The Skadden Arps AI adoption analysis profiles a Wall Street rival whose AI roadmap looks more like S&C's than like Cooley's, but with a sharper focus on client-discovery workflows.

The white-shoe firms — S&C, Cravath, Davis Polk, Skadden — are converging on a similar pattern: conservative production deployment inside the partnership, fixed-fee AI engagements at the client-facing edge, and a long pole on intake. The startup-focused firms are inverting that order: aggressive intake automation first, internal tooling second. Both pictures are correct; both reflect what their client bases actually want.

For partnerships shopping vendors, our law firm intake software comparison for 2026 maps the eight options most often shortlisted across both camps, and the Cravath, Swaine & Moore AI roadmap analysis is the closest sibling read to this post — both firms are M&A powerhouses, both move on AI only after the governance question is answered first.

For broader BigLaw context, the Davis Polk AI strategy for modernizing corporate workflows and the Mayer Brown global firm AI playbook across 27 offices round out the picture.

How Perspective AI Fits the White-Shoe Stack#

Perspective AI is an AI interviewer that runs conversations at scale — the surface S&C's intake architecture is missing. The product replaces the new-matter form, the conflicts questionnaire, and the discovery call with a conversational AI that follows up, probes "it depends" answers, and produces a structured transcript a partner can review in three minutes instead of thirty.

For a firm with S&C's risk posture, three product properties matter most:

• Conversations are governed, not generated — every prompt, follow-up, and clarifying question is configured by the firm; the AI does not improvise outside that scope, which keeps the firm's voice and the partner's judgment in control.
• Transcripts are first-class data — every intake produces a structured record that flows into the firm's matter graph, conflicts system, and knowledge base, not just an email thread.
• It runs in front of forms, not in place of judgment — the AI handles the parts of intake that are pattern-matching (entity disclosure, jurisdictional facts, scope) and hands the parts that require judgment to the partner with full context.

The conservative-adoption pattern at white-shoe firms predicts where intake automation lands next. The firms that ship it first will close conflicts faster, scope mandates more accurately, and capture institutional memory their competitors have to rebuild every quarter.

Frequently Asked Questions#

Has Sullivan & Cromwell actually deployed AI in production?#

Yes — Sullivan & Cromwell has deployed generative AI across M&A diligence, drafting, issues-list generation, and closing-checklist work, and runs an internal Applied Intelligence practice that pairs lawyers with data scientists on bespoke client matters. The firm published a January 2026 memo detailing six concrete use cases, and access to AI tools inside the firm is conditioned on completion of mandatory hallucination-risk training.

What was the Sullivan & Cromwell AI hallucination incident?#

In April 2026, Andrew Dietderich, co-head of S&C's Global Finance & Restructuring practice, sent an apology letter to the chief judge of the US Bankruptcy Court for the Southern District of New York after an emergency motion in the Prince Group bankruptcy was found to contain fabricated and misquoted legal authorities generated by AI. The firm acknowledged that its mandatory training and AI-use policies had not been followed on the filing. The incident has become the most cited cautionary tale in BigLaw AI deployment.

Is Sullivan & Cromwell a "white shoe" firm?#

Yes — Sullivan & Cromwell, founded in 1879, is one of the original "white shoe" firms, a label that historically described elite, Wasp-establishment Wall Street firms and now refers to the small cohort of prestige New York firms with the highest profits-per-partner, the most senior corporate client rosters, and the most conservative cultures. S&C, Cravath, Davis Polk, and Skadden are the most commonly cited white-shoe peers; S&C's 2024 profit per equity partner was approximately $6.74 million.

Why does S&C's AI adoption matter to other law firms?#

S&C's adoption matters because it sets the floor, not the ceiling, of BigLaw AI deployment. When a 145-year-old firm with white-shoe risk tolerance moves generative AI into production — including in highest-stakes M&A and restructuring work — the directional question for the industry is settled. Firms with less conservative cultures or less reputational exposure can no longer credibly delay on governance grounds alone; the question has shifted from "should we deploy" to "where on the funnel do we deploy first."

Where will AI hit law firm workflows next?#

AI will hit client intake and new-matter conflicts next, because intake is the highest-leverage surface in a firm — it touches every client and every matter before billable work begins, and it is still run on PDFs and questionnaires at firms that have already deployed AI to M&A and litigation. Conversational intake compresses the time to first conflict check, improves scoping accuracy, and converts every kickoff conversation into a structured asset for the firm's knowledge graph. Our analysis of why law firms are replacing forms with conversations walks through the operational math.

Conclusion#

Sullivan & Cromwell AI adoption is the bellwether moment for elite BigLaw. A 145-year-old firm with $6.7M profit-per-partner, a sovereign-and-issuer client list, and a culture that has historically been described as conservative on legal tech has moved generative AI into production across M&A diligence, drafting, and closing work — and built an in-house Applied Intelligence practice to run it. The April 2026 hallucination incident did not slow the firm down; it sharpened the governance regime around the technology. What S&C and its white-shoe peers have not yet automated is the front door — new-matter intake, conflicts capture, and scope discovery — and that is the next domino. Perspective AI runs conversational intake at scale: governed conversations, structured transcripts, and partner-grade outputs that fit the institutional standards a firm like Sullivan & Cromwell actually enforces. If you are mapping the next phase of your firm's AI roadmap, start a Perspective or see the conversational intake product.