More than three-quarters of U.S. entrepreneurs have woven artificial intelligence into their businesses, but a new spending reality is taking hold: if an AI tool can’t point to a concrete business metric, it risks losing its budget. That’s the central finding from EY’s latest Entrepreneur Ecosystem Barometer, a survey of 500 established U.S. entrepreneurs released July 28 and reported by Forbes. While 76% of respondents now use AI in some form — and 10% have fully integrated it across their processes — over a third have slashed spending on technology tools, including AI, over the past twelve months when the value wasn’t clear. The message from founders to vendors and internal IT teams is unvarnished: show us the money.
A Closer Look at the Numbers
The EY survey paints a detailed picture of entrepreneurial AI usage and expectations. Among those already investing in AI, the top three areas where leaders expect the greatest value are direct and measurable:
- Revenue growth through sales and marketing (57%)
- Cost reduction in operations (55%)
- Customer experience (49%)
Close behind are product and service innovation (48%), back-office efficiency (47%), and workforce efficiency (43%). These are not vague aspirations — they are boardroom-level metrics that can be tracked in dollars, hours, or percentage points. The Forbes article notes a real-world example: a window and door company that invested $10,000 in an AI application for its sales team. The tool listens to sales conversations, automatically creates a quote, and cuts down on paperwork and errors. “It allows my salespeople to talk to more customers and spend less time doing paperwork,” the owner said. That’s the kind of tangible return entrepreneurs now require.
Yet the road to AI payoff is bumpy. The survey identifies data readiness, integration complexity, talent shortages, and security as top barriers to scaling AI. While 99% of respondents say they have some approach to AI security and data-risk management, that broad figure masks uneven maturity — many smaller firms have a policy on paper but lack the enforcement muscle of enterprise-grade tooling.
Crucially, the talent dynamic is shifting. Nearly nine in 10 entrepreneurs (88%) expect their workforce to grow over the next 12 months. Rather than treating AI as a headcount-reduction lever, they are redesigning roles: 42% are reshaping jobs to blend human judgment with automated routine work. At the same time, 46% say a lack of talent and expertise is impeding AI scaling. The need isn’t just for data scientists; it’s for people who can validate AI outputs, map business processes, and manage data safely.
What This Means for Windows Shops and IT Decision Makers
For businesses running on the Microsoft stack — Windows 11 endpoints, Microsoft 365, Dynamics 365, Power Platform, and Teams — the EY data is a call to reexamine every AI investment through the lens of business outcomes. The days of buying a Copilot license or deploying a chatbot simply because it demonstrates generative AI are over.
Consider Microsoft 365 Copilot. It can summarize email threads, draft documents, and generate meeting notes — but to justify its $30 per user monthly cost, a business must identify a measurable benefit. Does it reduce the time a salesperson spends creating a proposal? Does it cut the average resolution time for a support ticket? IT leaders should work with department heads to define those baseline metrics before rollout and then track them religiously.
Power Automate is an even sharper example. Because it integrates with existing line-of-business apps and data sources — SharePoint, SQL Server, Excel, and thousands of third-party connectors — it’s perfectly positioned for the kind of narrow, measurable process automation that entrepreneurs now demand. A workflow that automatically routes invoice approvals, converts a quote to an order in Dynamics 365, or flags customer-service escalations can be equipped with a clear before-and-after story: minutes saved, errors avoided, throughput gained.
The windows-and-doors company story maps directly onto this. The AI tool wasn’t a standalone marvel; it attached to an existing system (the sales conversation), had a clear handoff to a person (the salesperson reviewing the quote), and delivered a result that could be counted: more customer conversations per day. That’s the template. Windows-centric IT teams and managed service providers should look for similar patterns: a narrow workflow with an existing system of record, a defined group of users, and an obvious metric.
Security and governance are no longer afterthoughts. The survey’s 99% figure suggests broad awareness, but awareness isn’t enforcement. In practice, that means moving from a Word document labeled “AI policy” to active controls in Microsoft Entra ID: conditional access policies that restrict which devices can access AI-enhanced apps, sensitivity labels that prevent Copilot from surfacing confidential content from SharePoint or OneDrive, and app governance that keeps unapproved consumer AI services from siphoning corporate data. Small and midsize businesses especially should review third-party AI tool permissions monthly and disable anything not tied to a documented business need.
Talent-wise, the message is to stop chasing unicorn data scientists and start upskilling the people who already understand your operations. An accounts-receivable clerk who knows why an invoice gets stuck can be trained to build a Power Automate flow that fixes it. A customer-service lead can learn to review AI-generated responses for accuracy and tone. These are the roles that bridge the gap between AI capability and genuine business value.
How We Got Here: The Hype Cycle Gives Way to Accountability
AI adoption didn’t happen overnight. A few years ago, the conversation was dominated by excitement over large language models and what they might do. Microsoft launched Copilot to great fanfare, and businesses scrambled to experiment. According to the EY survey, that experimentation phase has delivered real penetration: three-quarters of entrepreneurs now use AI in some form. But the honeymoon is ending.
What’s driving the pivot is a familiar pattern in tech spending. New tools get funded on promise, but sustained investment requires proof. The survey’s most telling statistic isn’t the 76% adoption — it’s that more than a third of these leaders have already cut technology budgets, including AI, when the returns weren’t evident. High-profile AI failures and underwhelming pilots have made boards cautious. Inflation and economic pressure have added urgency. The result is an entrepreneur class that, as Forbes put it, is saying “show us the money.”
This shift is especially relevant in the Windows ecosystem because many of the AI tools that have been pushed to users — Copilot in Edge, AI-powered search in Windows, automated transcription in Teams — are often adopted without a clear business case. The survey suggests that tolerance for such “nice to have” features is evaporating. If it can’t be tied to a deal closed, a cost saved, or a customer retained, it may not survive the next budget cycle.
What to Do Now: Building an AI Budget That Survives the Cut
For IT leaders and business owners, the path forward is straightforward but demanding. Here’s a practical five-step plan to align AI spending with the new ROI reality.
-
Pick a workflow, not a tool. Start with a business process that has a clear owner, a defined outcome, and a metric people already track. Examples: the quote-to-cash cycle, the customer-support ticket pipeline, or inventory-reorder triggers. Don’t begin by shopping for AI software — begin by mapping the steps and the pain points.
-
Establish a baseline. Before touching a single AI setting, measure the current state. How long does it take a salesperson to generate a quote? What’s the error rate on manual data entry? How many support tickets require a human callback? Write it down.
-
Rigorously pilot one AI enhancement. Deploy a narrow AI capability — a Copilot prompt that drafts a quote from a CRM record, a Power Automate flow that routes approvals, a Teams transcription that saves a summary to a Planner task — and measure the same metrics over a defined period (ideally 60–90 days). Compare to the baseline. If you can’t see a clear difference, don’t scale.
-
Enforce active governance. Move beyond documented policies. In your Microsoft 365 tenant, turn on Microsoft Entra ID conditional access to enforce managed devices for AI apps. Apply sensitivity labels that restrict Copilot access to confidential SharePoint sites. Use cloud app security to detect shadow AI usage and block unauthorized consumer services. Set up automated alerts for unusual data access patterns by AI services.
-
Reskill the people you have. Identify two or three employees who deeply understand a key process and train them on low-code AI tools: Power Automate, Copilot Studio, or even the AI Builder in Power Apps. Their operational knowledge is the ingredient that turns generic AI into a business-specific asset. Pair them with IT staff who understand data governance and integration, and make this cross-functional team responsible for the pilot’s success — and for proving that ROI.
Outlook: AI Becomes Part of the Business Fabric, Not a Separate Experiment
The EY data suggests that AI is not retreating — it’s maturing. With 76% adoption and 88% of entrepreneurs expecting workforce growth, the technology will continue to reshape how work gets done. But the era of blank-check AI budgets is over. In its place is a healthier cycle: businesses that can demonstrate a measurable return will invest more deeply, while tools that can’t will be quietly shelved.
For Windows-powered organizations, the pieces are already in place. Microsoft’s Copilot stack, Power Platform, and Dynamics 365 offer a rich toolkit for the kind of process-bound, metric-driven AI that entrepreneurs now demand. The differentiator isn’t access to models — it’s the discipline to tie every AI action to a business result. As one entrepreneur told Forbes, the lesson is simple: “It allows my salespeople to talk to more customers and spend less time doing paperwork.” That’s the standard. Meet it, and the budget will follow.