AI in Restructuring: The Biggest Shift in Professional Work Since the Spreadsheet

By Tom Goldblatt, CTP, Managing Partner, Ravinia Capital LLC

Featured in the Journal of Corporate Renewal, June 2026 — the flagship publication of the Turnaround Management Association, read by restructuring professionals, lenders, and advisors across the country.

In two years, AI has gone from a curiosity to the most significant change in how professionals work since the spreadsheet replaced the ledger pad. Pick your analogy: email replacing the fax, the smartphone putting a computer in every pocket. This shift is faster, broader, and less predictable than any of those.

The pace keeps accelerating. New model releases in the last few months have made capabilities from six months ago look primitive. What practitioners learned about AI last fall is already outdated. AI has become standard infrastructure in professional services. Not an experiment. Not a pilot. This is how competitive firms now operate.

What AI can produce and assist with is limited only by a professional's creativity. They can envision exactly what they want, a financial analysis structured a particular way, a pitch tailored to a specific audience, a research memo synthesizing a hundred sources, then prompt AI to become the specialist that delivers it.

Restructuring in Uncharted Territory

Here is the uncomfortable truth: there is no great guide to any of this. The space has changed so fast that the best practices of February are the outdated practices of June. The turnaround profession is in a period of experimentation, and the honest professionals are the ones who admit it.

Complicating matters: AI is not there yet. It still makes embarrassingly stupid errors. It fabricates citations with complete confidence. It has frustrating limits on its own memory. It is not obvious how to use it well, and the answers change weekly.

The biggest questions facing every turnaround professional today are simple and profound: what work should humans still do, and what should AI do? Nobody has the definitive answer. But after 18 months of daily use in live deal work, here is a working framework.

The Iterative Workflow

AI is not a vending machine. The best results come from an iterative process, a conversation between your judgment and the machine's output. A simple email might take a single pass. A complex engagement letter might take 15. The underlying rhythm is the same. Here is one framework developed at a boutique investment bank:

Phase	Time	What Happens
1. You Frame	20%	Define the problem. Attach relevant files. Set up context, tools, and connectors. Write the argument in your words. Set voice, structure, outcome.
2. AI Executes	30%	AI researches, drafts, models, formats — drawing on the context you provided.
3. You Review	10%	Review output. Correct errors. Improve structure. Adjust tone and emphasis.
4. AI Refines	20%	AI incorporates your corrections. Rebuilds and strengthens the draft.
5. You Review Again	10%	Second review. Tweak language. Sharpen arguments. Check voice.
6. AI Polishes	5%	Final formatting, consistency, cleanup.
7. You Sign It	5%	Final check for accuracy, reliability, and voice — your name goes on it.

An individual touches the work at every critical juncture. AI never gets the last word. That final 5%, where the user confirms accuracy, checks voice, and verifies that their name belongs on it, is the difference between professional work and professional embarrassment.

When to Use AI, and When Not To

Unlike most business decisions, there is almost no precedent here. Clients expect efficient, high-quality answers, but it reads as inauthentic when the work sounds more like a language model than like a real professional.

For now, the rule is caution. Internal drafts, research synthesis, financial modeling, and analytical groundwork are fair game. Client-facing communications, court filings, and expert reports require extreme care. AI can help to draft, but the final product must be unmistakably a person's. An individual should never submit output that couldn't be defended in front of a creditors' committee.

Work should be read aloud before it goes out. If any sentence sounds like “It is important to note...” or “In today's dynamic environment...,” it must be rewritten. Counterparties can spot AI tells now. In a profession built on trust, sounding like a machine signals you didn't care enough to do the work yourself.

For motivation: In 2025, Deloitte Australia refunded part of a $290,000 government report after fabricated references and a fictitious court quote were discovered, all AI generated, none disclosed. A second Deloitte report in Canada, costing $1.6 million, had identical problems. The Sixth Circuit, in March 2026, imposed $30,000 in sanctions on two attorneys who submitted briefs with more than two dozen fabricated cases. Every one of these failures had the same cause: a professional trusted AI output without checking it.

Choosing a Model

Each of the major platforms, Anthropic's Claude, OpenAI's ChatGPT, Google's Gemini, and Microsoft's Copilot, has strengths. Some organizations invest in industry-specific tools like Harvey, the legal AI platform now valued at $11 billion and used by over half the Am Law 100. Others build proprietary systems.

After working with several models, one boutique investment bank chose Anthropic's Claude as its foundational platform. The technical merits mattered: document processing, financial reasoning, and writing quality. But what sealed it was the company behind the model.

Anthropic was founded in 2021 by Dario and Daniela Amodei and colleagues who left OpenAI over concerns about how powerful AI was being developed without adequate safety work. Dario Amodei was OpenAI's VP of Research. He helped build GPT-2 and GPT-3, then left to create a public benefit corporation with safety, accuracy, and integrity embedded in its charter. The Anti-Defamation League's January 2026 AI Index tested six models across more than 25,000 conversations on bias detection. Claude scored 80 out of 100. The next closest scored 57.

AI Model	Score (out of 100)
Claude (Anthropic)	80
ChatGPT (OpenAI)	57
DeepSeek	50
Gemini (Google)	49
LLaMA (Meta)	31
Grok (xAI)	21

Source: ADL AI Index, January 2026

Best Practices After Making a Selection

Picking a model is the beginning. Turnaround executives need to train their teams, not just on the interface, but on how to think with the tool. At the firm that chose Anthropic, they mandated training on Claude's chat and Cowork environments.

More importantly, that firm defined its “skills,” custom instruction sets loaded into every session that encode the firm's standards. The professional communication skill opens with: “Every piece of writing that leaves this firm reflects on our credibility with clients, lenders, referral partners, and counterparties.” The firm prohibits AI-sounding language and requires verification of every factual claim. The quality checklist ends with: “Would you be comfortable if this ended up on a judge's desk?”

Beyond standards, learn to organize projects. Persistent context across engagements makes AI dramatically more useful than one-off conversations. And know the limitations: memory across sessions is imperfect.

Daily applications at the boutique investment bank include tailoring pitch materials to specific audiences in minutes, improving email clarity and tone, reviewing standard documents like NDAs, transcribing meetings and generating summaries for team members who missed them, creating customized interview questionnaires, and drafting social media content.

Where This Is Going

AI will get smarter and make fewer mistakes. The models being introduced this quarter are meaningfully better than last quarter's, a trajectory that is not slowing.

Every professional field will be reshaped. Legal billing by the hour will come under enormous pressure when work that takes a junior associate four hours can be drafted in seconds.

Investment banking analyst work will be increasingly absorbed by AI. The next wave is agentic AI, systems that execute multi-step workflows on their own, pulling data, drafting analysis, and looping in a human only at decision points. Much of what analysts and junior diligence staff do today will shift to work that senior turnaround professionals direct rather than perform. But the core functions from managing director on up will remain with humans for a long time. A sale-of-business transaction requires enormous trust, wisdom, and judgment. Deal sourcing will still be closed by people whose integrity is known. Deals will still be run by professionals who can judge the character of buyers and keep transactions from breaking.

Turnaround work, in particular, will always need seasoned professionals. Every distressed situation is different. The financials may rhyme, but the human dynamics never do: the lender who needs to save face, the founder who can't separate identity from the business, the creditor committee with undisclosed agendas. No AI reads a room. No AI earns a frightened business owner's trust at midnight.

The floor on standard business practice has broken. All turnaround professionals are in fresh territory. Those who master these tools without losing the judgment, integrity, and human connection that define our work will be the ones still standing.

The rest will be wondering what happened.

Tom Goldblatt, CTP, is managing partner of Ravinia Capital LLC, a Chicago-based investment bank specializing in distressed and middle-market M&A. He is a two-time M&A Advisor National Distressed Dealmaker of the Year and was the first investment banker to receive the TMA CTP of the Year award. He holds a J.D. from the University of Chicago Law School and an MBA from Northwestern's Kellogg School of Management and is a CPA.