 |
|
|
|||||
A new word has entered the business headline writers lexicon over the last month: the SaaSpocalypse. Between mid-January and mid-February 2026, around a trillion dollars was wiped from the value of software stocks. The S&P North American Software Index posted its worst monthly decline since the 2008 financial crisis. Individual stocks have been savaged, with even Microsoft, the ultimate tech blue chip, falling by more than 10%. The panic is real. But is it rational? The catalyst for this turmoil was a series of product launches from AI companiesmost notably Anthropics Claude Cowork tool and its subsequent upgradesdemonstrating that AI agents are now capable of handling complex knowledge work autonomously. The markets interpretation was both swift and brutal: If AI agents can do what enterprise software does, then enterprise software is finished. That narrative is clearly persuasive to those who have been busily dumping stocks. But it rests on a fundamental misunderstanding of what enterprise software is, what it does, and why replacing it isnt the straightforward proposition the market appears to believe. {"blockType":"mv-promo-block","data":{"imageDesktopUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/10\/creator-faisalhoque.png","imageMobileUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/10\/faisal-hoque.png","eyebrow":"","headline":"Ready to thrive at the intersection of business, technology, and humanity?","dek":"Faisal Hoques books, podcast, and his companies give leaders the frameworks and platforms to align purpose, people, process, and techturning disruption into meaningful, lasting progress.","subhed":"","description":"","ctaText":"Learn More","ctaUrl":"https:\/\/faisalhoque.com","theme":{"bg":"#02263c","text":"#ffffff","eyebrow":"#9aa2aa","subhed":"#ffffff","buttonBg":"#ffffff","buttonHoverBg":"#3b3f46","buttonText":"#000000"},"imageDesktopId":91420512,"imageMobileId":91420514,"shareable":false,"slug":""}} More Than a Tool The simple premise behind the market turmoil is that AI agents will, in the not-too-distant future, be able to perform most or all of the tasks that are currently performed by enterprise software. But this vision of the future misunderstands enterprise software at a fundamental level. Enterprise software isnt just a set of tools. It encodes the enterprise itself. Decades of business rules, process flows, governance structures, compliance requirements, data definitions, and role-based permissions are held within these systems. When a company runs on SAP, Salesforce, Microsoft, or ServiceNow products, its not simply using a suite of software that sits on top of the organization. These systems hold the organizations operating architecture in digital formthe institutional memory of how the business actually works in practice, every day, at every level. Replacing enterprise software with a fully agentic enterprise isnt just a matter of swapping one piece of technology for another. The moat around enterprise software isnt the code. Its the accumulated domain knowledge, the business logic, and the deep integration with how organizations actually operate. Three Fallacies Driving the Panic The case for wholesale replacement rests on three assumptions. Each collapses under scrutiny. The first is the change management fallacy. Putting enterprise software in place is not like installing an app; these are often multiyear organizational transformations involving workflow redesign, data migration, retraining, and deep integration across departments. Companies typically change ERP systems every 5 to 10 years, and even routine migrations require months of rigorous preparation. The notion that organizations will undertake wholesale replacement of their entire enterprise architecturenot with new software, but with an entirely different paradigmignores the reality that change management is one of the hardest things organizations can attempt. The disruption involved in even incremental software upgrades creates significant operational risk. A complete paradigm shift involves risks to the business of an entirely different order of magnitude. The second is the economic fallacy. Even if replacement were technically feasible, there is no compelling reason to believe it would be cheaper. Token-based AI pricing is expensive at the enterprise scale, and the world in which running agents across an entire organizations operations could cost less than current SaaS subscriptions is not yet the world in which we live. Token costs will fall over timewe can be sure of thatbut building a case for wholesale replacement on the assumption that they will fall far enough and fast enough to undercut the established economics of enterprise software involves stacking assumption on top of assumption. Token costs are only one part of the equation. The true cost of running agentic systems includes orchestration, integration, data pipelines, monitoring, security, auditability, and the human time required to supervise and correct outputs. The last item is the one most easily underestimated: As agents take on more autonomous and more consequential work, assurance costs will rise, not fall. And even before you reach the question of ongoing costs, the price of the transition itselfthe data migration, workflow redesign, retraining, and inevitable disruption to operationswould be enormous. The economic argument for replacement isnt just weak; at present, it barely exists. This isnt to say that its not plausible in some future world. But until we have a convincing map that leads there, its not a serious proposition. The third, and possibly the most important, is the general-purpose agent fallacy. The assumption behind the market panic is that powerful, general-purpose AI agents will take over enterprise functions wholesale. But this doesnt reflect how AI actually delivers value today, and it may not reflect how agents ever deliver value. Research consistently shows that AI works best when its targeted at specific problems with rich contextual grounding. A study conducted by the Australian government found that broad-access AI tools produced significant improvements in basic tasks like summarizing information and preparing first drafts, but that their lack of fit to users specific contexts undermined efficiency gains in more complex work. The result was a productivity paradox: Time saved through automation was consumed by checking and correcting outputs that lacked the domain-specific nuance the work required. This finding has direct implications for the SaaSpocalypse thesis. General-purpose agents deployed to replace enterprise software will face exactly the same problem. Without deep local contextthe profound domain knowledge and specific workflow logic that enterprise software encodesthey will produce generic, unreliableoutputs that require constant human correction. To work effectively at the enterprise level, agents need to be narrow, contextually rich, and tightly integrated with specific workflows. And once you start building agents that way, youre not replacing software as a service. Youre rebuilding it through an agentic lensat enormous cost and with no guarantee that the result will be better than what you already have. What Leaders Should Do None of this means the landscape isnt shifting. AI is changing how people interact with software and how organizations think about their technology investments. But the right response isnt to tear up the enterprise architecture. Its to evolve it. Rather than reacting to the panic, leaders should take three concrete steps. 1. Audit your vendors AI road maps. The strongest enterprise software providers are already integrating agentic capabilities into their platforms. If yours arent, thats a genuine concern, and it may be time to look for vendors who are. The question isnt whether to adopt AI, but whether your existing partners are doing it for you. 2. Invest in data quality and process documentation. The effectiveness of any AIwhether embedded in your software or deployed as agentsdepends on the quality of the data and the clarity of the processes it works with. This is the foundational investment, and it pays off regardless of where the technology lands. 3. Evaluate agentic approaches for genuinely new workflows. Where youre building new capabilities or addressing needs that your current software stack does not serve, purpose-built agentic solutions may be more effective and more flexible than new SaaS implementations. This is where the technologys real greatness lies. Further reading Do you really know what agent means? – Fast Company How AI is changing what it means to be the CEO – Fast Company The Trillion-Dollar Question The SaaSpocalypse makes for dramatic headlines. But the idea on which those headlines are basedthat AI agents will soon be eating the lunch of enterprise software providersis founded on a misunderstanding about what enterprise software does. Its not just a tool that performs tasks. Its the digital encoding of the organizations institutional architecture. That isnt something a general-purpose tool can easily replace. The real risk for business leaders isnt that they will be too slow to abandon their enterprise platforms. Its that they will be stampeded by market panic into undervaluing the systems and institutional knowledge they already have. AI will reshape enterprise softwarethat much is certain. But there is a meaningful difference between a technology that changes how software works and one that makes software unnecessary. That distinction matters. And for the moment at least, the market has lost sight of it. {"blockType":"mv-promo-block","data":{"imageDesktopUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/10\/creator-faisalhoque.png","imageMobileUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/10\/faisal-hoque.png","eyebrow":"","headline":"Ready to thrive at the intersection of business, technology, and humanity?","dek":"Faisal Hoques books, podcast, and his companies give leaders the frameworks and platforms to align purpose, people, process, and techturning disruption into meaningful, lasting progress.","subhed":"","description":"","ctaText":"Learn More","ctaUrl":"https:\/\/faisalhoque.com","theme":{"bg":"#02263c","text":"#ffffff","eyebrow":"#9aa2aa","subhed":"#ffffff","buttonBg":"#ffffff","buttonHoverBg":"#3b3f46","buttonText":"#000000"},"imageDesktopId":91420512,"imageMobileId":91420514,"shareable":false,"slug":""}}
Category:
E-Commerce
When a new general-purpose technology emergesbe it railroads, electricity, computers, etc.companies react in predictable ways. A small minority tries to reinvent themselves around it; the majority looks first for ways to cut costs. Right now, in the middle of the most significant technological inflection since the internet, many organizations are choosing the second path. They deploy artificial intelligence to automate call centers, reduce head count in back offices, and squeeze marginal gains out of existing processes. They measure AI ROI in payroll savings and hours reclaimed. It feels rational. It feels disciplined. It feels safe. It is also the fastest way to miss the real opportunity. Innovation waves are not efficiency programs AI is not a new SaaS tool, nor is it merely a workflow enhancement. It is a rapidly evolving general-purpose technology advancing from large language models to agentic systems and toward systems that learn from interaction with environments (the so-called world models that can simulate, plan, and act). When the underlying capability is shifting every few months, optimizing for cost reduction is like trying to improve the fuel efficiency of a car while its engine is being replaced with a jet turbine. The organizations that win in moments like this do not start by asking, Where can we eliminate labor? They ask, What becomes possible that was previously impossible? Those are radically different questions. The productivity paradox should have been a warning In the early 1990s, economists puzzled over a surprising phenomenon: Computers were everywhere, yet productivity statistics stubbornly refused to reflect their impact. In a press article, Nobel laureate Robert Solow famously quipped, You can see the computer age everywhere but in the productivity statistics. That observation became known as the “productivity paradox.” At the time, many assumed the paradox was a failure of technology. My own research from that time examined why the paradox appeared at all, showing that productivity measurement lags widely behind actual transformational change and that the mechanisms of value creation were not captured by conventional metrics. The explanation was obvious only in hindsight. The gains were diffuse, uneven, and entangled with organizational change. Companies had digitized old processes instead of redesigning them. Today we are watching the same pattern unfold with AI. AIs impact wont show up neatly in cost metrics Artificial intelligence does not produce clean, linear productivity gains that fit neatly into quarterly dashboards. Its effects are asymmetrical. One employee using AI effectively may outperform 10 peers. Another may misuse it, degrade quality, or even endanger our corporate cybersecurity plans. Some teams redesign workflows entirely, while others bolt AI onto legacy processes and call it transformation. The result is what researchers now call measurement myopia: the inability of traditional metrics to capture improvements that are real but not directly tied to hours worked or cost saved. Trying to measure AIs value solely through immediate cost savings is like trying to measure the value of electricity by counting candles not purchased. Efficiency is the comfort strategy, but not the opportunity one Cost-cutting is attractive because it fits existing governance structures. CFOs understand it. Boards reward it. Metrics are clear. Exploration is messier. It requires experimentation without guaranteed returns. It demands a tolerance for failure. It produces intangible benefits before visible ones. But in periods of fast innovation, efficiency is often the comfort strategy of laggards who dont yet understand what is happening. If AI is treated primarily as a head-count-reduction tool, organizations will optimize the present and sacrifice the future. They will standardize mediocrity instead of discovering leverage. Exploration, not exploitation, builds capability Advocating exploration does not mean abandoning discipline. It means redefining it. Leaders should be asking: What new products can we build with AI-native capabilities? What decisions can we delegate to systems that learn from feedback? How can we redesign workflows, not just automate them? Companies should mandate controlled experimentation across teams, not restrict AI usage to narrow cost-justification pilots. They should treat AI like an R&D posture rather than a shrink-the-budget posture. Organizations that treat AI as an exploratory layerencouraging teams to test, prototype, recombine, and rethink workflowswill build institutional fluency. They will develop internal champions. They will uncover unexpected value that no top-down cost initiative would have surfaced. The real risk isnt overspending. Its under-imagining The greatest risk in this moment is not overspending on AI. It is under-imagining it. Companies that chase short-term efficiency gains may report modest improvements and declare success. Meanwhile, more ambitious competitors will redesign their operations, products, and customer experiences around capabilities that didnt exist two years ago. Over time, the gap will not be a few percentage points of margin. It will be strategic. In periods of rapid technological change, survival does not belong to the most efficient. It belongs to the most adaptive.
Category:
E-Commerce
For decades, formative assessment has been a silent engine for learningpowering insights about student progress and worker readiness. But lets be honest, in a world where technology is evolving faster than human skills, its time to ask questions about traditional teaching and learning models, and in many cases, modernize them. So, lets talk about formative assessment in the age of AI. Formative assessment is the ongoing process educators and workplace trainers use to understand where students are in their learning and how to adjust instruction accordingly, through homework, essays, quizzes, and short writing assignments. Eighty percent of educators rate formative assessment as extremely or very important. Unfortunately, but understandably, the arrival of generative AI has made it difficult for instructors to determine what students genuinely understand, as AI tools can produce polished work instantly. THE FUTURE OF ASSESSMENT DESIGN While administrative policy can help address improper AI use, the real potential for progress comes from evolving assessment design itself. When assessments are built to prioritize the thought process rather than just the product, AI becomes far less disruptive and far more beneficial. Asking students to make their thinking visiblethrough reflections, revisions, or short explanations of how they approached a taskrestores the instructional signal that AI might otherwise obscure. For educators, this means being able to spot misconceptions earlier, tailor feedback more precisely, and differentiate support without increasing workload. This shift isnt about adding complexity. If anything, its about adding clarity. And its an opportunity to modernize assessment in ways that mirror the world students are entering. In most professional environments, AI assistance is not only allowed; it is expected. Success comes from knowing how to use these tools responsibly: checking sources, critiquing the quality of generated outputs, and adapting insights to novel contexts. Assessments that emphasize reasoning, analysis, and the ability to apply knowledge to new situations better reflect these real-world demands. They prepare students not just to complete tasks, but to think with AI in ways that enhance their learning and judgment. TEACHER BENEFITS For instructors, thoughtfully integrating GenAI within formative assessment can also reduce friction. Welldesigned tools can automate repetitive tasks such as generating varied practice items, suggesting targeted feedback language, or providing examples at different proficiency levels. This allows educators to spend more time on the highvalue interactions that deepen learning and provide individualized support. In an era of rising expectations and constrained capacity, that shift matters. There is another often overlooked benefit: insight. When AI helps surface patterns in student work, it gives educators a clearer starting point for instruction. With better visibility, teaching becomes more adaptive, and learning becomes more personalized. This is especially powerful in large classes, hybrid formats, or virtual learning environments where realtime insight can be harder to access. Recent Pearson research reveals strategies for schoolteachers and higher education instructors to evolve their formative assessments in a GenAI era. Of course, none of this happens automatically. Bold, collaborative action is required across school and highereducation leadership, administrators, and policymakers to ensure formative assessment evolves in meaningful and sustainable ways. Together, these groups play a critical role in providing a clear AI strategy, supporting educator training, and shaping an ecosystem that aligns curriculum, instruction, and assessment with responsible GenAI use. This transition also requires assessments that reward thoughtfulness over polish, reasoning over rote, and application over replication. And it requires a shared understanding that AI is not a shortcut to learning but a catalyst for insightone that can elevate the quality of teaching when used intentionally. A LOOK AHEAD The future of formative assessment isnt about outsmarting AI or pretending it doesnt exist. Formative assessment must remain fundamental to good teaching and effective learning. Ensuring AI strengthens reflection, feedback, and understanding will allow it to become a partner, rather than a substitute for learning. With thoughtful action, the integration of AI into teaching and learning can move us closer to what education has always aspired to deliver: deeper learning, clearer understanding, and better outcomes for every learner. Tom ap Simon is the president of higher education and virtual learning at Pearson.
Category:
E-Commerce
IBM stock was down 10% on Monday afternoon after Anthropic published a blog post about how its Claude Code tool can be used to modernize software written in the COBOL language, which handles large-scale batch transactions. Many of the software systems used by the federal government, banks, and airlines are written in COBOL (“Common Business-Oriented Language”), and most of those systems run on IBM mainframes. IBM also generates revenue from servicing, modernizing, and consulting on those mainframes. If COBOL code were converted to a more modern language, the systems would likely migrate to newer cloud servers. But modernizing COBOLwhich was developed 67 years agois a slow and expensive process, largely because the code can be difficult to understand and easy to break. It often reflects decades of institutional knowledge and workflows, and is frequently poorly documentedmeaning its true intent can only be uncovered through close analysis. These challenges are compounded by the shrinking pool of programmers who know COBOL. Most university computer science programs no longer teach it. Anthropic says this analysis phase is the most time-consuming and costly. Thats where Claude Code comes in. The tool can uncover and document workflows hidden within the code, identify dependencies across different parts of a code base, and give engineers insights into how to redesign systems. With AI, teams can modernize their COBOL code base in quarters instead of years, the company writes in the blog post. IBM says the analysis phase is not the hardest part. “Translating COBOL is the easy partthe real work is data architecture redesign, runtime replacement, transaction processing integrity, and hardware-accelerated performance built over decades of tight software and hardware coupling,” an IBM spokesperson said in an email. “That is the problem IBM has spent decades learning to solve, and AI is the most powerful tool we have ever had to do it.” COBOL was developed in 1959 via a public-private partnership that included the Pentagon and IBM, with the goal of creating a universal, English-like programming language for business applications. But private-sector companies have largely moved away from it. The code is difficult and costly to maintain and was designed for batch processing, making it poorly suited for modern cloud-based and real-time applications. (Anthropic and IBM did not immediately respond to requests for comment.) The U.S. government, despite repeated modernization efforts, continues to rely on COBOL-based mainframe systems to manage a wide range of financial transactions, including tax payments and refunds, Social Security benefits, and Medicare reimbursements. Anthropics blog post comes in the middle of a separate dispute between the company and the government. Anthropic CEO Dario Amodei is expected to meet with Defense Secretary Pete Hegseth to explain why the company has not removed all safety guardrails from its AI models for Pentagon use. Anthropic has drawn the line at providing AI for autonomous weapons or systems that mass-surveil American citizens. At the moment, Anthropics models are the only ones approved for government use with classified information. Anthropic says its blog post about COBOL modernization is unrelated to its friction with the government. “The timing here isn’t related to a new product or any events,” a company spokesperson said in an email. “This is part of an ongoing series of content we’ve been publishing around code modernization and Claude Code.” And Anthropic’s blog post may not be the only factor affecting IBMs stock. Investor concerns about the speed and breadth of AI deployment have depressed enterprise software stocks more broadly. The market may also be reacting to uncertainty surrounding new global tariff announcements, which could affect tech companies and their supply chains.
Category:
E-Commerce
The Big Gulp might have some new competition in the realm of giant beverages from an unlikely dark horse: Dunkin‘. Over the weekend, Dunkin’ customers in New Hampshire and Massachusetts began posting head-turning images of giant coffee buckets on the menu at their local stores. While some commenters doubted the veracity of these reports, a Dunkin’ spokesperson confirmed in an email to Fast Company that the donut chain is indeed testing out a 48-ounce collectible bucket at select stores after noticing buzz around coffee buckets taking off on social media. A coffee bucket is exactly what it sounds like: a giant iced latte served in a plastic container that looks more like a garden tool than a cup. The novelty beverage took off this summer and appears to have been sparked by several different small businesses, including Noctua Coffee in Missouri, Dulce Vida in Oklahoma, and Wicked Southern Coffee in Connecticut, all of which attracted thousands of views on social media. Dunkin’ is no stranger to jumping on a trend, so it makes sense that the brand would arrive at this moment in the social media zeitgeist with a bucket in tow. In the past few years, Dunkin has experimented with wacky concepts like an alcoholic drink line, a donut deodorant, and a horny Halloween donut. Heres what to know about its latest launch: Where can I find the Dunkin bucket? Dunkin’ told Fast Company that the coffee bucket test is taking place at select stores in New Hampshire and Massachusetts, but the company did not provide an official list of locations. Internet sleuths and coffee fanatics have uncovered a few stores that reportedly carry the bucket, according to a cursory search of social media. We have requested the full list of participating stores from Dunkin’ and will update this story if we hear back. What comes in the bucket? According to the Dunkin’ spokesperson, guests can fill their coffee buckets with classics like iced coffee, iced lattes, or Dunkin’ refreshers. Also available are three featured drinks: the blueberry cobbler iced latte, caramel coco iced coffee, and strawberry dragonfruit lemonade refresher. (We shudder to imagine the nutritional contents of these creations.) Customers report paying between $7 and $10 for their buckets. How are customers reacting? So far, customers main complaint for this behemoth of a beverage appears to be the impossible prospect of transporting it. One Instagram Reel with almost 85,000 likes from creator Elijah Boivin shows Boivin cradling the bucket above the caption, Me holding my Dunkin bucket because I dont know where to put it because it doesnt fit in the cup holder. A modern conundrum, indeed.
Category:
E-Commerce