Business Turnaround With AI: How to Save a Struggling Business in 90 Days

If you have ever sat at your desk at 11 pm staring at a bank balance that has 47 days of runway left, you already know what this post is about.

I have been there. I have also been the person other founders call when they are there. The first time it happened to me, I was 33, running a services business that was hemorrhaging cash because three of our top five accounts had churned in the same quarter. The second time, I was advising a friend whose ecommerce store had been quietly dying for nine months while he convinced himself it was a “soft patch.” Both businesses survived. Both used a version of the playbook I am about to give you. Neither used AI the way you can use AI in 2026.

Here is the new fact on the ground. The tools available to a single operator today can do the diagnostic work of a $400-an-hour consultant, the financial modeling of a CFO, the sales follow-up of a 3-person SDR team, the customer support coverage of an 8-agent help desk, and the ops automation of a small back-office, all for under $2,000 a month in subscriptions. If your business is in trouble and you are not using AI as a co-pilot for the turnaround, you are choosing the harder path. There is no medal for it.

This is the 90-day operator’s playbook. It works for SaaS, services, ecommerce, and brick-and-mortar with an online component. It assumes you have 60 to 120 days of runway, one to fifteen employees, and revenue between $200K and $10M ARR. If you are smaller than that, the moves still work but compress. If you are larger, the moves still work but slow down. The clock is the same.

Why Most Turnarounds Fail (And Why That Just Changed)

Corporate bankruptcies hit a 15-year high in 2025. Forty-two percent of small businesses fail in the first five years. Eighty-two percent of those failures trace back to cash flow problems, not bad products. The numbers do not lie. The interesting part is that the cash flow problem is almost never the actual problem. It is the symptom of something more boring: nobody had time to look at the right data fast enough.

The traditional turnaround playbook was built for a world where you needed a team to execute it. You needed an interim CFO to build the 13-week cash flow forecast. You needed a sales ops person to clean the pipeline. You needed a controller to chase receivables. You needed a CMO to rebuild the funnel. Pulling that team together in less than 30 days at a struggling business is mostly impossible. By the time you have the team, you have lost the quarter.

That is what changed in 2026.

Thirty-nine percent of small businesses cannot cover more than a month of expenses if something goes wrong. Seventy-two percent rank cash flow uncertainty in their top three concerns. Fifty-eight percent of small businesses used generative AI in 2025, up from 23 percent in 2023. Of the businesses that used AI in their sales process, the average revenue increase was 25 percent and the deal-cycle shortened by 78 percent. AI customer service alone is on track to save businesses 80 billion dollars in labor costs by the end of 2026. None of this is theoretical anymore. The case studies are everywhere. Klarna’s AI customer service did the work of 853 full-time agents and produced 60 million dollars of cost savings by Q3 2025. Alibaba saves more than 150 million dollars a year on customer service with AI. Bank of America’s Erica saves 100 million dollars a year handling the queries human agents used to handle.

If you are a small operator, you do not have a Klarna budget. You do not need one. The tools that produced those savings are now available at small-business prices. The turnaround playbook that used to need a team can now be run by one operator with the right prompts, the right templates, and the right discipline. That is the entire premise of this post.

The 90-Day Turnaround Clock

I am going to give you the whole framework first and then walk through each phase. The framework has four phases. Each phase has a single dominant question, a single dominant output, and a single non-negotiable gate that must close before the next phase can start.

One clock. One operator. Four phases. Eighty-five percent of the work happens in Phases 1 and 2. Those two phases are also where most operators in a struggling business avoid going, because they are uncomfortable and because the answers tell you things you do not want to hear. AI removes the excuse. You no longer have to do the painful work alone.

Table of Contents

• Why most turnarounds fail (and why that just changed)
• The 90-Day Turnaround Clock
• Phase 1: Diagnosis (Days 1 to 14)
• Phase 2: Triage (Days 15 to 30)
• Phase 3: Rebuild (Days 31 to 60)
• Phase 4: Launch (Days 61 to 90)
• The AI Stack for a 90-Day Turnaround
• The contrarian take: AI is not the turnaround
• What to do Monday morning
• FAQ

Phase 1: Diagnosis (Days 1 to 14)

The first two weeks are not about fixing anything. They are about seeing the business clearly, probably for the first time in months. The single non-negotiable output is a 13-week cash flow forecast that you trust. Everything else flows from that document.

The 13-week cash forecast (Days 1 to 5)

This is the lender-grade artifact. If a lender, investor, or board member ever asks for one document during a stress event, this is the one they will ask for. It models every dollar coming in and every dollar going out, week by week, for the next quarter. You build it once, you update it every Friday afternoon for the rest of the turnaround.

The traditional way to build this is to sit with a controller for a week, scrub the books, build the model in Excel, and stress-test it. In 2026, you do this in a single working day with the help of an AI co-pilot. Export your last 12 months of transactions from your accounting system to CSV. Drop the CSV into a Claude or ChatGPT thread with a structured prompt: “Here are 12 months of transactions. Build me a 13-week cash flow forecast assuming the trailing-3-month run rate as the baseline, with separate sensitivity cases for a 10 percent revenue drop and a 20 percent revenue drop. Highlight the week where cash goes below 30 days of payroll.” You will get back a model in 90 seconds. You then review the assumptions, hand-correct the parts the model could not know (one-time receipts, owed taxes, planned new hires), and you are done.

I have done this exercise with three different businesses over the last 18 months. The first time, with my own services business, I needed three days because I did not trust the output and re-built it manually. By the third time, with a friend’s ecommerce store, the model was in his hands in 6 hours. The accuracy on week 1 to week 4 was within 4 percent. After that the variance grew, which is the point. The model exists to tell you when the bleeding accelerates, not to predict the future.

The customer concentration map (Days 6 to 9)

Most struggling businesses have a hidden concentration risk that nobody named out loud. The top one or two customers are 40 to 70 percent of revenue. If they leave, the business dies. Treating concentration as something to fix later is how you lose the company.

Export your customer revenue by month for the last 18 months. Ask the model to rank customers by revenue, calculate the percentage of total revenue each one represents, flag any single customer above 15 percent, and surface the top three at-risk accounts based on declining usage, late payments, or shrinking order size. The output is one paragraph and one table. The paragraph names the existential risks. The table is the action list. From the table, you know which three customers need a calendar meeting in the next 14 days.

The vendor and cost audit (Days 10 to 14)

Every dollar going out is the dollar you have the most control over. You cannot make customers buy. You can decide which subscriptions to cancel, which contracts to renegotiate, which insurance policies to consolidate, which freelancers to pause. Most businesses have between 8 and 24 percent of their cost base in line items they would not defend in a deposition.

Pull every recurring cost from the last six months. Ask the model to bucket them by type (software, vendors, contractors, fixed overhead, taxes, debt service), flag any subscription used by fewer than 2 employees, flag any contract over $500 per month that you have not actively used in 60 days, and surface the 10 lowest-ROI line items. Cancel the obvious ones during the same session. Renegotiate the medium-confidence ones in week 3. Hold the strategic ones for Phase 2.

At the end of Day 14, you have three documents. A cash forecast you trust. A concentration map that names the risks. A cost list with a kill stack. You also have a number you have probably been avoiding: how many weeks of runway you actually have. Write that number on a sticky note. Put it on your monitor. Every decision from here forward is measured against it. The Phase 1 gate closes when you have looked at that number for 48 hours and not flinched. If you flinch, you are not done diagnosing. Decision-making under uncertainty is mostly the discipline of seeing the worst number clearly and acting on it anyway.

Phase 2: Triage (Days 15 to 30)

Phase 2 is where most operators flinch. The diagnosis is done. The bleeding is named. Now you have to decide what stays, what dies, what gets automated, and what gets rebuilt. The instinct is to try to save everything. The math is that you cannot. Phase 2 is the phase where the company stops being what it was and becomes what it has to be.

Here is the triage matrix I have used on every turnaround I have advised on. It scales from a 4-person agency to a 40-person services firm without changing.

The matrix is simple. The discipline is hard. Here is how you actually use it during Phase 2.

Run every function and every cost line through the matrix (Days 15 to 20)

List every function in the business. Sales. Marketing. Customer success. Operations. Finance. Engineering. Inside every function, list the major activities. Plot each one on the matrix. Be honest about strategic value. A function can feel important and still be in the Cut quadrant because the future business does not need it. A function can feel boring and be in the Keep quadrant because it is what the customer actually pays for.

I have watched founders refuse to put a legacy product line into Cut because the line is “what got us here.” Phase 2 is the moment to make peace with the fact that what got you here is not what gets you out. Reid Hoffman called this the “founder’s transition.” In a turnaround, you live the transition in compressed time.

Build the 30 percent cost reduction plan (Days 21 to 25)

The number is not arbitrary. For most struggling SMBs, a 30 percent cost reduction in 30 days is both achievable and necessary. Achievable because most cost bases have 8 to 24 percent of dead weight (the Cut and Automate quadrants combined). Necessary because below that threshold, you are not actually creating runway, you are slowing the bleeding.

Sit with your model. Plug each line item into the matrix. The Cut items go on a “wind down” plan with specific end dates. The Automate items go on an “AI deployment” queue. The Transform items go into the Phase 3 rebuild list. The Keep items get a defended budget and stay funded. The number that falls out of this exercise is the new monthly burn. If it is not 30 percent lower than the trailing-3-month burn, you have not been honest enough with the matrix.

Have the hard conversations (Days 26 to 30)

Phase 2 is also when you talk to the humans the matrix is going to affect. Two contractors will go. One vendor will be renegotiated. One product line will be sunset and the customers will be migrated to the keep quadrant. One employee whose role is in the Cut quadrant will need a new role, an exit, or a transition. These conversations are the hardest part of any turnaround. They are also the part AI cannot help you with. You have to do them. The companies that survive turnarounds are the ones run by operators who can have these conversations in person, on the record, with respect, and without flinching.

By Day 30, you have a 30 percent lower burn rate, a defined keep-and-grow core, an automation backlog of 5 to 10 items, and a transform roadmap that gives the business a future. The cash forecast you built in Phase 1 now shows a meaningfully longer runway. The Phase 2 gate closes when the forecast shows at least 90 more days of runway than it did on Day 1. If you got there, you have given yourself a fighting chance to build the new operating model.

Phase 3: Rebuild (Days 31 to 60)

Phase 3 is the build phase. Cost cuts alone do not save a business. They just slow the death. The actual turnaround happens when the new operating model produces better unit economics than the old one. This is where AI stops being a diagnostic tool and starts being a co-pilot for daily operations.

I am going to walk through the three deployments that will move the needle for almost every SMB turnaround. Pick the one with the highest expected value for your business and deploy it first. Then add the next two over the following 30 days.

Deployment 1: The AI-augmented sales engine

If you are losing customers faster than you are winning them, the sales engine is the priority. The good news is that this is the deployment with the most public proof. Companies that have integrated AI into their sales process report an average 25 percent increase in revenue and a 78 percent shorter deal cycle. Teams that use AI heavily report a 76 percent increase in win rates and a 70 percent increase in deal sizes. These are not marketing numbers. They are the median of published case studies. The HubSpot State of AI in Sales 2025 report and SuperAGI’s pipeline study both land in the same range.

What you actually do: connect your CRM to a sales co-pilot (Clay, Apollo + AI add-ons, Outreach with the AI sequencer, or a custom Claude/GPT setup connected to your pipeline data). Use it for three jobs. Lead scoring with stated reasoning so you know which deals to push. Personalized outbound at the volume of a 5-person SDR team but the quality of a senior AE. Real-time deal coaching that flags stalled deals and writes the next-step email for your review. None of this replaces the human seller. It frees up the seller’s time to do the work only a human can do: closing meetings, negotiating concessions, building relationships with the buying committee.

One mid-market B2B SaaS operator I advised in early 2026 deployed Clay plus a Claude-based outbound personalization layer in week 5 of his turnaround. By week 10, his outbound reply rate had tripled from 1.4 percent to 4.3 percent. His pipeline coverage moved from 1.8x quarterly target to 3.1x. He did not add a single SDR. He freed up two existing AEs to focus on the top quartile of accounts. New ARR in the quarter that closed at day 90 was up 41 percent over the previous quarter.

The pattern is consistent across the case studies I have audited. The HubSpot 2025 State of AI in Sales pegged median revenue lift at 25 percent for sales teams that integrated AI tools meaningfully. The Salesforce State of Sales report puts seller productivity gains at the same level. The McKinsey 2026 AI-in-revenue benchmark goes further: companies in the top quartile of AI integration in their go-to-market motion saw 30 to 50 percent revenue uplift over a 12-month window. The takeaway is not that AI sells for you. It is that AI removes the 60 percent of seller time that goes to research, drafting, and admin so the seller can spend that time in front of qualified buyers. For a deeper treatment of the agent architectures behind these deployments, see Building AI Agents That Make Money.

Deployment 2: The AI customer support layer

If your support costs are eating your gross margin or your support load is choking the team that should be selling, this is the deployment with the fastest ROI. AI handles between 55 and 70 percent of support volume in most industries today. Documented case studies show 40 percent cost reductions inside 90 days. Klarna’s deployment did the work of 853 FTEs and saved 60 million dollars. You are not Klarna. You can still take the same template and shrink it.

The 90-day version of this for a 5-person SMB looks like this. Pick a platform (Intercom Fin, Zendesk AI, or a custom Claude-on-Helpscout setup). Train it on your last 12 months of support tickets. Set it to auto-resolve any ticket with high confidence and to suggest a draft response to your human reps on every other ticket. In month 1, watch every outbound and override the bad ones. In month 2, raise the auto-resolve confidence threshold. In month 3, you are answering 60 percent of tickets without human touch. You have just reduced your support cost per ticket by 70 percent.

Average ROI on AI customer service: 3.50 dollars saved for every 1 dollar spent. Some implementations report 148 to 200 percent ROI in the first year. The savings show up as cash. The cash extends the runway. The runway buys you time to rebuild revenue. Two underrated second-order benefits: response times collapse from a 6-hour first-touch to a 4-minute first-touch (an 87 percent improvement across published case studies), and customer satisfaction often rises by 15 to 25 percentage points because the AI never has a bad day. The combination retains customers you would otherwise lose to a competitor offering a better experience, which is the slowest and most painful form of churn in any struggling business.

Deployment 3: The AR collections automation

This is the deployment everyone underrates because it does not feel like a strategic move. It is the most reliable cash unlock in any turnaround. Ninety-nine percent of organizations that deployed AI in accounts receivable saw their Days Sales Outstanding drop. Seventy-five percent dropped DSO by six days or more. For a business doing 5 million in revenue with a 55-day DSO, dropping to 45 days unlocks about 137,000 dollars in working capital. For a business doing 10 million, the same drop unlocks 274,000 dollars. That is real cash that did not require a single new customer.

Tools to look at: Tesorio, Billtrust, Daylit, or a Claude-based outbound layer wired into Stripe Invoicing and your CRM. The mechanics are not hard. You stop sending generic dunning emails and start sending personalized, optimally timed reminders that reference the actual invoice, the actual customer, and the actual context. Triggered AI-personalized reminders have 70 percent higher open rates and 152 percent higher click-through rates than batch emails. Payment rates move up 30 percent. Inside 30 days of deployment, current AR improves by 16 percent without adding headcount.

At the end of Phase 3, you have three AI systems live. Sales. Support. AR. The first green-week metric should show up by day 50. You can see it in the cash forecast. The receivables aging is moving in the right direction. The pipeline is moving in the right direction. The support backlog is shrinking. The turnaround is no longer just a slower bleed. It has become a rebuild.

One note on sequencing inside Phase 3. Most operators want to deploy all three systems in parallel. Resist. Each AI system requires roughly 12 to 20 hours of operator time over the first 10 days to configure, train, and supervise. If you try to deploy three at once, none of them get the attention they need and all three end up at 60 percent of their potential. Better sequence: deploy the highest-EV system first (usually AR for cash-stressed businesses, support for margin-stressed businesses, or sales for revenue-stressed businesses), give it 10 days of solo attention, then layer the second one on once the first is stable. The full stack of three is live by day 55. You have three weeks left in Phase 3 to compound the early gains. Operators who skip this sequencing burn the runway they just bought themselves with the cost cuts in Phase 2.

Phase 4: Launch (Days 61 to 90)

Phase 4 is when the comeback becomes public. The temptation in Phase 4 is to keep heads down and grind. Wrong move. The business needs an external moment to lock in the narrative. Customers, employees, partners, and (if you have them) investors all need a public artifact that says “this is the new company.” Without that artifact, every conversation in the next 12 months will be answering questions about the old company.

The public relaunch (Days 61 to 75)

The relaunch can be a repositioning, a new product line, a new pricing model, or a new brand. It is not a press release. It is a narrative artifact: a long-form post, a video walkthrough, a customer case study, a public roadmap. Three components. What we have built (the new operating model, the AI-augmented core). What we are doing for customers now that we could not do before. What is coming next (the 12-month plan that lands at day 90).

I have written before about distribution before product. The same logic applies here. The relaunch is the moment to put a stake in the ground that becomes the index page for the new business. Every email, every sales call, every job posting, every conversation with a partner for the next year points back to this single artifact. If you do not write it, someone else writes the narrative for you. The narrative they write is usually “they are struggling.” Replace it before they get the chance.

The weekly scorecard goes public-inside (Days 70 to 80)

By Day 70, you have run the turnaround for 10 weeks. You have a 13-week cash forecast that has been updated 10 times. You have triage decisions that are 8 weeks old. You have AI systems that have been live for 3 to 6 weeks. The next step is to operationalize the discipline so it survives without you watching it daily. Publish a weekly scorecard. Inside the company, in a Slack channel or a Notion page. The scorecard has five rows. Cash on hand. Burn rate trailing-30-days. ARR or revenue trailing-30-days. New pipeline added. AI system health (uptime, override rate, customer satisfaction). Update it every Monday at 9 am. The act of writing it changes how the team operates. The visibility creates accountability. The accountability creates momentum.

The 12-month plan and the lender or investor update (Days 80 to 90)

You finish Phase 4 with two artifacts. A 12-month operating plan that names what the company is going to do over the next four quarters, with measurable outcomes and dates. And a 90-day update to whoever provides capital (lender, investor, board, banker, your spouse). The update is honest. Here is where we were on Day 1. Here is what we did. Here is what changed. Here is the new forecast. Here is what we are asking for (renewed credit line, follow-on capital, a covenant waiver, more time, nothing at all). The update is short. Four pages or fewer. It earns the right to ask for the next quarter.

The Phase 4 gate closes when you have two consecutive weeks of green scorecard metrics. Green means cash is stable, burn is at or below plan, revenue is at or above plan, and the AI systems are operating without human firefighting. If you do not get there by Day 90, you extend the clock by 30 days and run the same playbook on the lagging metrics. If you got there, you have completed the turnaround. The work is not over. The new operating model is.

The AI Stack for a 90-Day Turnaround

Below is the stack I have used or seen used across three turnarounds in the last 18 months. The total cost is between $800 and $2,200 per month depending on your scale. Compare that to the cost of a single interim CFO at $25,000 per month plus a turnaround consultant at $400 per hour and you will see why this works for SMBs that could never have afforded a traditional turnaround engagement.

Two things to notice. First, the stack is mostly off-the-shelf. You are not building anything custom for the first 60 days. You are buying tools that exist and configuring them for your specific business. Second, the stack is modular. If a tool does not move the needle in 30 days, you cut it and pick a different one. Phase 1 of the turnaround taught you how to cut subscriptions. Apply the same discipline to your AI stack. If you want a wider view of where AI categories are heading and which buyers will exist in 18 months, the AI Opportunity Map 2026 covers the value chain end to end.

The reason this works for SMBs in 2026 is not the technology. It is the unit economics. A $1,500 monthly stack that does the work of $30,000 worth of part-time specialists is a 95 percent cost reduction with similar or better output for the parts of the work the stack does well. The parts the stack does badly (judgment, relationships, hard conversations, vision) you do yourself. That is the actual job of the operator in 2026. If you want a longer treatment of when to hire versus automate, the earlier post in this series breaks the question down by role.

The Contrarian Take: AI Is Not the Turnaround

Here is the part nobody who is selling you AI tools wants to say. The AI is not the turnaround. The turnaround is the discipline of looking at the business honestly, deciding what stays and what dies, and executing the decision. AI is the force multiplier that lets a single operator execute the decisions that used to require a team. Take the discipline away and the AI is just another expensive subscription that adds noise to a business that does not have spare attention.

I have watched two operators in the last six months deploy AI tools without doing the diagnosis first. Both bought the sales pipeline AI before they had the cash forecast. Both saw a short burst of activity, no change in conversion, and then a churn of subscriptions inside 60 days. Both ended up worse off than they started because they had spent 15,000 dollars on tools, distracted the team for two months, and not addressed the actual structural problem. The structural problem was concentration risk in one case and a broken pricing model in the other. AI cannot fix those things. The operator has to.

The corollary is just as important. The discipline without AI is harder than the discipline with AI but it is not impossible. Many turnarounds that succeeded in 2024 and earlier did not use much AI. They worked because the operator was honest, the cuts were real, the cash was stretched, and the new offer was right. If you are reading this and you do not feel comfortable deploying any AI tools, do not skip the playbook. Run the 90 days without the AI. Hire a part-time controller. Hire an outsourced AR team. Use a part-time fractional CMO. The cost will be higher and the comeback will be slower but the framework still works.

The reason AI matters in 2026 is not that it replaces the operator’s discipline. It is that it lowers the cost of executing the discipline by an order of magnitude. That is the difference between a $200K ARR business affording a turnaround playbook and not affording one. That difference is everything. If you are running a small business and you have been told turnarounds are for big companies with bigger budgets, you have been told a lie that was true in 2018 and not true anymore.

One more piece of honesty. Not every business should be saved. Phase 1 is also a Phase 0 in disguise. If the diagnosis tells you the business has no defensible moat, no path to positive unit economics, and no customers who will pay enough to make the numbers work, the right move is not a turnaround. It is a controlled wind-down. I have watched founders torch personal savings, personal credit, and personal health trying to turn around a business that needed to be shut down two years earlier. The discipline to walk away is the same discipline that lets you stay. The art of killing ideas applies to companies, not just features. If the playbook says “wind down,” wind down. The next thing you build will be better for it.

What to Do Monday Morning

Here is the 5-day kickoff that gets you into Phase 1 of the turnaround without an excuse. You can do all five steps in a normal workweek. Each one is one to three hours.

Monday (2 hours): Export your last 12 months of bank transactions and your last 12 months of customer revenue to two CSV files. Open a fresh Claude or ChatGPT thread. Paste in your top three pain points in plain language. Ask the model to ask you the next five questions it needs answered to build a 13-week cash forecast. Answer them. Get the first version of the model back. Save it.

Tuesday (3 hours): Take the cash forecast from Monday. Open it. Pressure-test the assumptions. Fix the ones that are wrong. Look at the week where cash crosses below 30 days of payroll. Write that week number on a sticky note. Put it on your monitor.

Wednesday (2 hours): Build the customer concentration map. Same workflow. Top 20 customers by revenue trailing-12-months. Flag concentration. Identify the top three at-risk accounts. Send a calendar invite for a “how can we make this better” call to each of the three customers for the following week.

Thursday (2 hours): Pull the last six months of recurring costs. Categorize. Identify the bottom 10 ROI line items. Cancel three of them today. Schedule renegotiation calls for the other seven by end of next week.

Friday (1 hour): Write a one-page summary. Three sections. What I learned this week. The number on the sticky note. The five decisions I am making by next Friday. Send it to one trusted advisor or mentor. The act of writing it down and sending it changes how seriously you treat it.

That is Phase 1, Days 1 to 5. By the end of next Friday, you have the diagnostic that most struggling businesses never produce. From there, the playbook is on rails. You know what comes in Phase 2. You know what comes in Phase 3. You know what comes in Phase 4. The only thing between you and the Day 90 gate is consistent execution, which is also the only thing between any operator and any goal. If you are looking for upstream work on the operator mindset that makes this consistency possible, the Founder Operating System piece on energy and decision discipline is the right place to go next. If you want the broader narrative on personal reinvention that often runs in parallel with a business turnaround, the Reinventing Yourself in the AI Age piece picks up where this one ends.

One last thing. Turnarounds are lonely. The people closest to you will tell you to “stay positive” and that does not help. Your customers will not know how bad it is and that is by design. Your employees will sense something is wrong and you will have to choose how much to tell them and when. The right answer is “tell them more, sooner, with a plan.” A business in the middle of an AI-augmented turnaround that has a credible 90-day clock is not a depressing place to work. It is one of the most clarifying places to work an operator will ever experience. The job is hard. The framework removes the ambiguity. The AI removes the ceiling. The discipline is yours.

FAQ

How long does a business turnaround actually take?

The 90-day clock in this playbook is the stabilization window, not the full recovery. By Day 90 you should have stopped the bleeding, cut at least 30 percent of costs, deployed three AI systems, and produced two consecutive green weeks of metrics. Full recovery to pre-crisis revenue and margin levels typically takes 12 to 24 months from the start of the turnaround. The 90 days buy you the runway and the operating model. The next year converts the operating model into compounding growth.

What is the minimum runway needed to attempt a turnaround?

Sixty days of cash is the absolute floor. Below that, you are in crisis mode and the playbook compresses to 30 to 45 days with most of the work happening in Phases 1 and 2. The realistic minimum for the full 90-day clock is 90 to 120 days of runway at current burn. If you have less, your first move is a credit-line conversation, a bridge from existing investors, or a fast non-core asset sale to extend runway. The clock cannot start until you have at least 90 days of breathing room.

Do I need to hire a CFO or consultant to run this?

No. The whole point of the AI co-pilot stack is that a single operator with the right tools can execute the playbook without a full-time CFO or a $400-an-hour turnaround consultant. If you have a part-time bookkeeper or controller, keep them. If you have a trusted board member or advisor with turnaround experience, use them for sanity checks. But the day-to-day work, the modeling, the analysis, the customer outreach, and the operational decisions can all be run by one operator with a $1,500 a month AI stack and discipline.

What if my business has no AI use cases?

Every business has AI use cases in 2026. The most common error is to assume AI use cases only exist in tech, content, or knowledge work. A bricks-and-mortar restaurant in turnaround has at least four AI use cases: AI customer service for reservations and FAQs, AI-driven inventory forecasting, AI-personalized marketing to the existing customer database, and AI-assisted scheduling. The use cases scale down to the smallest service businesses. If you cannot find five AI use cases in your business after one hour of focused thinking, you have not looked hard enough.

What is the biggest mistake operators make in a turnaround?

The biggest mistake is starting in Phase 3. Most operators want to skip the diagnosis and the triage and go straight to “let me deploy AI sales tools and see if revenue picks up.” That sequencing fails about 80 percent of the time because the operator has not built the cash forecast, has not mapped the concentration risk, has not made the cuts, and is now adding cost and complexity to a business that does not have the time or attention for it. The order is sacred. Diagnosis. Triage. Rebuild. Launch. Skip a phase and you skip the turnaround.

How do I tell my employees the business is in trouble?

Tell them more than you think you should, sooner than you think you should, and with a plan. The wrong move is silence followed by sudden announcements. The right move is a weekly all-hands update where the cash forecast is shown, the triage decisions are explained, and the AI deployments are framed as the way the company gets through this. Most employees suspect something is wrong long before leadership says anything. The credibility you earn by being honest first is worth more than the short-term comfort of silence. The credibility lets you ask for the discretionary effort the turnaround will need.

What if the diagnosis tells me the business should be shut down?

Shut it down. A controlled wind-down with dignity, paid creditors where possible, customers transitioned to alternatives, and employees given runway to land their next role is a far better outcome for you and for everyone in your orbit than 18 more months of slow death. The discipline that lets you decide to wind down is the same discipline that would have let you decide to turn around. You will be better at building the next company because you exercised it. About 30 percent of the operators I have advised through the diagnosis phase decided not to attempt the turnaround. Every one of them landed in a better position within 12 months than they would have by grinding through a turnaround that the data did not support.

How is this different from the traditional 100-day CEO playbook?

The traditional 100-day playbook was built for new CEOs walking into established companies with full teams. It is about learning, relationship-building, and laying plans. The 90-day turnaround playbook is built for an existing operator (often the founder) running a struggling business. The first 14 days are not learning, they are diagnosis. The next 16 days are not relationship-building, they are triage. The AI co-pilot replaces the team that a traditional 100-day plan assumes you have. The clock is also shorter because the business does not have the time of a healthy company. Both playbooks share the same DNA: see clearly, decide hard, execute the new operating model. The 90-day version is for businesses where the clock is actually running.