# Enterprise Autonomy 2028 ## A Scenario *Published by Enterprise Autonomy, 2026* --- ## About this scenario This is a work of grounded speculation. The prologue documents events between 2023 and 2025 that have already occurred. The scenario that follows — 2026 through 2028 — is fictional. It follows two Fortune Global 500 industrial manufacturers as their leadership teams make architectural choices about enterprise AI. One chooses autonomy. One does not. Their trajectories diverge over three years. The scenario is not a prediction. It is a description of what plausibly happens when the choices that are being made in real enterprise environments in 2025 and 2026 are followed to their operational and financial conclusions. The characters are fictional. The companies are fictional. The data cited in the prologue is real and sourced. The trajectory described in the scenario is one of several that are consistent with what is being observed in production enterprise deployments today. --- ## Cast **Maria Reyes** — Chief Information Officer, Meridian Industries. Fifteen years in enterprise IT. Six years as CIO at Meridian. Two failed AI vendor engagements behind her by the end of 2024. Pragmatic. Suspicious of demos. Reports to Katherine Ng, the CEO. **David Chen** — Vice President, AI and Digital Transformation, Vertex Systems. Formerly ran digital transformation at a large industrial peer. Product-oriented. Strong internal advocate for the Microsoft Copilot ecosystem. Reports to Priya Kapoor, the CIO. **Meridian Industries** — Fortune Global 500 specialty chemicals and advanced materials manufacturer. Twelve manufacturing sites globally. Traded on NYSE. **Vertex Systems** — Fortune Global 500 specialty industrial materials manufacturer. Direct competitor to Meridian in three product categories. Traded on NYSE. --- ## Prologue — 2023 to 2025 (documented) By late 2023, the phrase *enterprise AI* appeared in every technology strategy document written on Earth. Every Fortune 500 CIO had received a mandate from the board: use AI, show measurable results, do it now. Consulting firms sold billions of dollars in transformation engagements. Software vendors introduced *AI-native* versions of every existing product. The market attempted to reorganize itself in eighteen months. In November 2023, McKinsey estimated that generative AI could add between $2.6 trillion and $4.4 trillion in annual value across industries. By 2025, that estimate had become the anchor for every enterprise AI business case written. It also became the number every CFO would eventually ask their CIO to defend. The gap between promise and reality opened quickly. In August 2025, MIT's Sloan Management Review published a study of 148 enterprise generative AI initiatives. Ninety-five percent of them had failed to reach production. Three months later, DigitalRoute reported that seventy-one percent of Fortune 500 CFOs said they could not measure ROI on any of their AI investments. The failures were not distributed evenly across use cases. Customer service pilots — where the AI could be constrained to a single conversation with a single user — often succeeded. Anything that required the AI to read something, decide something, and then do something in another system almost always died. That was most business processes. Vendors offered two categories of solution. **Robotic Process Automation** was old but reliable. It executed. It did not understand. It broke the moment anything about the underlying process changed — a form redesigned, a screen field moved, a policy exception introduced. It was scriptable software running in a costume. **Copilots** were new and enthusiastic. They understood. They did not execute. They answered questions and then handed the real work back to a human. In the vendor demos, this looked like productivity. In the analyst reports, this looked like adoption. In the actual Fortune 500 environments, it looked like a chatbot bolted onto a spreadsheet. Neither owned the outcome. By the middle of 2025, a small number of enterprises had begun to experiment with a third pattern. They were building teams of specialized AI workers that coordinated through a reasoning core, connected to enterprise systems through an emerging open protocol called MCP, and executed real business processes end-to-end. The pattern had a name that had not yet stuck: **Enterprise Autonomy**. This is the story of two Fortune 500 companies that reached the same crossroads and chose differently. --- ## Chapter 1 · Q1 2025 · The MIT number The MIT Sloan report crossed Maria Reyes's desk on the morning of March 4, 2025. Ninety-five percent. She read it twice. She had already killed two AI pilots at Meridian in the previous eighteen months — one a customer service copilot that could not access order history, the other a document summarization tool that could not distinguish between three different versions of the same policy. Both had cost between $1.5M and $2M by the time she pulled them. She forwarded the report to Katherine Ng, the CEO, with a single line. *"This is not a Meridian problem. This is a category problem."* Ng read it in the car on the way to the airport and forwarded it to the board. Two hundred miles away, in Vertex Systems' Chicago headquarters, David Chen received the same report from a Gartner analyst. He read the executive summary and skipped to the recommendations. He drafted a Slack message to Priya Kapoor: *"MIT's saying 95% fail. We're doing this right — we have adoption metrics on the Copilot rollout. Not us."* He hit send. --- ## Chapter 2 · Q3 2025 · The DigitalRoute number By September, Meridian's Copilot deployment had reached 8,000 employees. Adoption reports from the vendor showed 76% of licensed users had used it at least once in the previous 30 days. Maria stared at the number in a QBR slide and asked a question no one on the vendor call wanted to answer. *"What did they use it for."* Silence. The account executive said something about summarization and drafting. Maria asked how many minutes on average a summarization session lasted. The answer was 3.2 minutes. She asked how many business processes had been shortened or replaced. There was no answer to that question. That afternoon, DigitalRoute published its 2025 CFO report. Seventy-one percent of Fortune 500 CFOs said they could not attribute measurable business outcomes to any generative AI investment their company had made. Maria did not need to read the report. She had just watched the interview at 3:42 that afternoon. At Vertex, David saw the same number in a summary. He wrote it in the notes for his upcoming board update as *"category-wide measurement challenge"* and continued preparing his slides. --- ## Chapter 3 · Q1 2026 · A protocol becomes a standard In January 2026, Anthropic's Model Context Protocol — MCP — moved from proposal to de facto standard. Every major AI platform vendor supported it. Every enterprise system with an API published or announced an MCP wrapper. For the first time, the connection layer between AI reasoning and enterprise systems had a single open interface. Maria's platform team briefed her on it in late February. What MCP unlocked was not incremental. It was categorical. An AI system could now, in principle, read from and write to *every* enterprise system it had permission to access. Not through custom integration engineering, but through a single protocol. She asked her chief architect one question. *"Does this change what an AI worker can actually do."* The chief architect thought about it. He said yes. Two weeks later, Maria approved a new pilot. Not a copilot deployment. Not a summarization tool. A single autonomous workflow: quality nonconformance handling on the Baton Rouge line. The AI would read the nonconformance report, pull the historical data from the MES, check the batch records, correlate to any known supplier variance, draft a corrective action, and route it. If it wasn't sure, it would escalate to a human. If it was sure, it would finish the work. At Vertex, David was preparing the Q1 2026 quarterly Copilot expansion. Fifty thousand additional licenses. The rollout would take two quarters. Priya had approved the budget the previous week. --- ## Chapter 4 · Q3 2026 · The first mission runs By July, Meridian's Baton Rouge autonomous nonconformance workflow had been in production for eleven weeks. The system had processed 1,847 nonconformance events. It had resolved 78% of them without human involvement. Of the escalations, 84% had been handled correctly by the reviewing engineer on first pass — meaning the system's escalation criteria were working. Maria did not tell anyone outside Meridian. Katherine Ng noticed the operating cost of the quality organization at Baton Rouge decline by $340,000 in Q2. She asked why. Maria told her. At Vertex, David presented his Q2 board update. Copilot adoption had reached 42,000 monthly active users out of 68,000 licenses. The board asked about business impact. David cited a McKinsey study on time-savings-per-worker. He was thanked for the update. Nothing was decided. Late that month, an analyst at a small research firm published a note titled *"The First Autonomous Enterprises."* It was the first time the phrase had appeared in an industry publication. Maria's platform team sent it to her. She read it and forwarded it to Katherine with three words. *"This is us."* --- ## Chapter 5 · Q4 2026 · The yield question In October, Maria commissioned a broader initiative. Six autonomous workflows across three plants. Nonconformance was the pilot. Now: batch scheduling, feedstock allocation, unplanned maintenance triage, supplier quality escalation, customer order status handling, and CAPA (corrective and preventive action) drafting. The pattern was consistent. Each workflow was designed for a specific business outcome — not a productivity metric. Each was owned by a business leader with a P&L, not by IT. Each ran on Meridian's own infrastructure. Each was connected to the enterprise stack through MCP. By the end of the year, Meridian's twelve manufacturing sites had adopted at least one autonomous workflow. Plant-level yield improved on average 1.4% year-over-year. That number would appear in Meridian's Q4 earnings call in February 2027. At Vertex, David had a difficult Q4. The Copilot vendor's own measurement study showed a 23% reduction in average time spent on document drafting among heavy users. The board asked whether that had translated into any measurable change in operating margin. David explained that it was hard to isolate the AI variable. The board thanked him. Priya Kapoor did not thank him. --- ## Chapter 6 · Q1 2027 · The board asks the hard question Meridian's Q4 2026 earnings call was held on February 8, 2027. Analyst David Foulkes at a bulge-bracket bank asked Katherine Ng about the 1.4% yield improvement figure that had appeared in the operating supplement. Was it structural, or one-time? Katherine's answer was measured. She used the phrase *"autonomous operational infrastructure"* for the first time on a public call. She said Meridian had been investing in it for eighteen months and expected the improvements to compound. The stock moved up 2.3% that afternoon. Two Meridian competitors — one of them Vertex — moved down. At Vertex, David spent the next morning drafting a memo to Priya. The subject line was *"AI strategy — recalibration recommended."* He proposed forming a working group. Priya replied by email. *"Come see me at 2."* --- ## Chapter 7 · Q2 2027 · The consulting engagement Vertex hired a Tier 1 strategy firm in April to review its AI strategy. The engagement cost $2.8M. It ran for eleven weeks. The final deliverable was a 168-page report and a set of recommendations. The recommendations included: consolidate the Copilot investment; hire a Chief AI Officer; establish an "autonomous AI Center of Excellence"; run six pilots by end of year in autonomous workflows selected for maximum measurable impact; and evaluate three vendors. David presented the recommendations to Priya. She read the executive summary and asked one question. *"Where are we going to find the engineering capacity to run six autonomous AI pilots by end of year, when we have not yet finished the Copilot expansion?"* David did not have an answer. Priya asked him to come back in a week with a proposal. At Meridian, Maria had by this point brought thirty-two autonomous workflows into production across the enterprise. She had hired one additional platform engineer that year. The workflows had been designed and deployed by the business functions that needed them. --- ## Chapter 8 · Q3 2027 · The gap becomes visible In August, an industry research firm published its inaugural *State of Enterprise Autonomy* report. The report used data from 800 Fortune 2000 companies. It defined a metric it called the *Autonomy Multiplier* — the throughput per employee of enterprises that had adopted autonomous workflows versus those that had not. At the median, autonomy-first enterprises were operating at a 1.9× multiplier by mid-2027. The report projected the multiplier to reach 3.4× by the end of 2028 in leading deployments. Meridian was mentioned by name in the appendix. Vertex was not. The report was cited by twelve analyst firms in the following four weeks. It appeared in the *Wall Street Journal* on August 22. Katherine Ng was quoted. She used the phrase *"we did not invent this — we recognized it earlier than most"* — which became the report's most-quoted line. At Vertex, David watched the coverage. He forwarded the *WSJ* article to Priya with no note. --- ## Chapter 9 · Q4 2027 · An acquisition attempt In November, Vertex's CEO called Meridian's CEO. He proposed a strategic partnership. Katherine deflected politely. Two weeks later, Vertex's board authorized $340M for the acquisition of an autonomous AI platform vendor. The vendor was a mid-stage private company with 63 employees and $22M in ARR. Vertex acquired it for a 15× ARR multiple. The acquisition closed in February 2028. Integration began immediately. The autonomous AI platform team was retained but reported to Priya, not to David. David was asked to focus on the Copilot ecosystem. At Meridian, Maria received an invitation to speak at Davos. She declined. She said in an internal email that she was busy running Meridian's business. --- ## Chapter 10 · Q1 2028 · Margins diverge Meridian's Q4 2027 earnings, reported in February 2028, showed operating margin expansion of 380 basis points year-over-year. Katherine Ng credited *"autonomous operational infrastructure across our twelve manufacturing sites"* and *"a step-change in the efficiency of our commercial operations."* Vertex reported the same quarter three days later. Operating margin was flat. The stock declined 6.2% on the call. Priya Kapoor announced a restructuring in the technology organization. Sixteen roles were eliminated. The digital transformation function was consolidated under the newly-appointed Chief AI Officer, who reported to the CEO, not to Priya. David's role and title were preserved. His scope was not. He wrote in his personal notes that evening — a habit he had kept for fifteen years — that he had been *"lapped."* --- ## Chapter 11 · Q2 2028 · Meridian on Bloomberg Katherine Ng appeared on Bloomberg on the morning of April 18, 2028. The interview was intended to be about Meridian's Q1 results. The interviewer opened with a different question. *"You've been described in the trade press as running the first fully autonomous Fortune 500 enterprise. Is that accurate?"* Katherine's answer became one of the most-quoted business statements of the year. She said: *"We are not fully autonomous. We are architecturally ready to be autonomous where it makes business sense. Which is most places, and more places every quarter."* The clip was shared 340,000 times in the following 72 hours. Meridian's stock closed up 3.4%. Two analysts upgraded the company that afternoon. At Vertex's Chicago headquarters, David watched the interview on the treadmill in the executive gym. He got off the treadmill, showered, and drove to the airport. He did not come back to Vertex. He submitted his resignation on the flight. --- ## Chapter 12 · Q3 2028 · The sector reprices By August 2028, seven of the twenty largest specialty industrial materials manufacturers had either publicly committed to autonomous operations or completed acquisitions of autonomous AI platform companies. Trading multiples in the sector had begun to bifurcate. Companies with public commitments to and evidence of autonomous operations were trading at premiums of 22% to 34% relative to their prior-year multiples. Companies without were trading at discounts. The bifurcation was noted in a research note from a bulge-bracket bank on August 14. The note used a phrase that became the title of a follow-up book: *"The Autonomy Premium."* At Vertex, the newly-installed CAIO announced that the company had completed its first autonomous workflow — a customer service escalation triage system — in production. She said it was the first of forty planned by end of 2029. The stock did not respond. At Meridian, Maria promoted her platform engineer to Chief Enterprise Autonomy Architect. There were now sixty-eight autonomous workflows in production across Meridian's twelve sites and its commercial operations. She stopped counting because the number was no longer strategically meaningful. Katherine Ng was interviewed by *Fortune* in September for their annual list of the most influential CEOs. She was ranked fourteenth. Maria was not interviewed. --- ## Chapter 13 · Q4 2028 · The board of Vertex asks a different question In November, Vertex's board convened for the annual strategy review. The Chief AI Officer presented her forty-workflow roadmap. She projected the Autonomy Multiplier for Vertex to reach 1.8× by end of 2029. The board asked her what Meridian's multiplier was. She said 3.1×. The board asked what it would take for Vertex to reach 3.1×. She said that Meridian had started in Q1 2026. Vertex had started in Q1 2028. The gap was two years of compounding architectural investment. It was possible to close, but it was not possible to close quickly. She said this in a way that was intended to be honest. The board thanked her for the update. Two months later, the Vertex board initiated a strategic review that would culminate, eleven months after that, in the sale of the company. --- ## Epilogue — 2029 and beyond By 2029, the phrase *Enterprise Autonomy* had reached the level of common usage that *cloud native* had reached by 2017 and that *digital transformation* had reached by 2015. Every enterprise software company claimed to enable it. Every consulting firm sold engagements to implement it. The category was no longer new. The 5% of enterprises that had reached the pattern first had, by then, moved on to a different problem. Their autonomous workflows were compounding. Their operating leverage was expanding. Their competitors were spending capital to catch up. Their advantage was not permanent, but it was not disappearing quickly, either. Maria Reyes remained CIO of Meridian Industries. She gave one keynote per year at an industry conference and otherwise avoided public commentary. In private, she told colleagues that the hardest part of the last three years had not been the technology. It had been getting her business functions to accept ownership of the workflows their AI was running. Katherine Ng was named *Fortune's* CEO of the Year in 2030. In her acceptance interview, she was asked what Meridian had done that other companies had not. She said: *"We didn't do anything special. We just took the architecture seriously. And we did it earlier than most."* David Chen took an operating role at a mid-cap materials company in Ohio. He wrote a book about his experience at Vertex. The book was thoughtful and generous and moderately successful. Vertex Systems was acquired by a European industrial conglomerate in June 2030. The acquisition was described in the press releases as a *"combination of complementary specialty materials platforms."* Analysts described it as *"Vertex could not close the operational gap fast enough."* Both descriptions were correct. Sometime in 2031 or 2032, a management scholar somewhere will write a paper about what happened to Vertex, and to the fifteen or twenty other companies that made the same choice. The paper will probably conclude that the difference was not the technology and not the leadership. It will conclude that the difference was the *architectural choice made in 2026*, and that everything downstream — the yield improvements, the margin expansion, the analyst upgrades, the acquisitions — was a consequence. Whether the paper is right will depend, in the end, on whether Meridian's advantage lasts. And that will depend on decisions Katherine and Maria and their successors have not yet made. The reader is left with the question the scholar will eventually ask. *Which company would you be?* --- ## Sources cited in this scenario - **MIT Sloan Management Review, "The 95% Problem: Why Enterprise Generative AI Pilots Fail to Scale," August 2025.** - **DigitalRoute, "The CFO's AI Dilemma: 2025 Report on Enterprise AI Monetization," November 2025.** - **McKinsey Global Institute, "The Economic Potential of Generative AI," June 2023.** - **Anthropic, "Model Context Protocol (MCP) Specification," 2024–2026.** The remaining events, characters, and companies in the scenario are fictional. Any resemblance to specific enterprises is illustrative. --- ## About Enterprise Autonomy Enterprise Autonomy is an independent publication covering the emerging category of autonomous AI in the enterprise. Founded in 2026 by Ajay Malik. Seeded with support from StudioX. Editorial decisions are made independently by the editorial team. To submit an article, propose a topic, or contribute a press release, visit **enterpriseautonomy.ai/submit**. --- *This report is available in PDF, audio, and machine-readable Markdown formats at enterpriseautonomy.ai/report.*