Bridging the AI divide: why 95% of AI projects stall, and how to close the gap
Bridging the AI divide: why 95% of AI projects stall, and how to close the gap
TLDR
→ A July 2025 MIT Media Lab report found that 95% of enterprise AI pilots deliver no measurable P&L impact, despite $30-40 billion invested. The divide is not about model quality. → The core barrier is a learning gap: most AI agents do not retain feedback, adapt to context, or improve between sessions. Users who encounter the same failure twice tend not to try a third time. → Infrastructure monitoring tracks uptime and latency. It cannot tell you whether employees are using your AI agent effectively, which teams have stopped engaging, or where the agent consistently fails users. → The organisations closing the ROI gap measure user outcomes: task completion rate, intent resolution rate, hours saved by department, and return rate. These connect AI usage to business results. → Organisations that build feedback loops into their AI agents from day one improve their products significantly faster, and convert pilots into production at a measurably higher rate. Updated: 24th June 2026
A July 2025 report from MIT's Project NANDA put a number on something many enterprise leaders already sensed. Despite $30-40 billion in enterprise AI investment, 95% of AI pilots delivered no measurable P&L impact. Only 5% of integrated systems created significant value. AI Wiki
The report named this split the AI divide. On one side, a small group of organisations extracting millions in productivity gains and cost reduction. On the other, a large majority with working technology and no business results to show for it.
The divide is not about which model you chose. The core barrier is not infrastructure, regulation, or talent. It is learning. Most AI systems do not retain feedback, adapt to context, or improve over time. (Campus Technology)
The learning gap in practice
60% of organisations evaluated enterprise-grade AI systems. Only 20% reached pilot stage. Just 5% reached production. Enterprises led in pilot volume and allocated the most staff to AI initiatives. They also converted the fewest pilots into production systems.
The reason surfaces consistently in user behaviour. Users often prefer consumer AI interfaces for drafts, but reject enterprise tools for mission-critical work due to lack of memory and persistence. One user quoted in the MIT report explained it directly: the tool was useful for early-stage work, but it repeated the same mistakes session after session and required extensive context input every time. For high-stakes work, that is not sustainable.
When the official AI agent does not meet user needs, employees find alternatives. While only 40% of companies have official AI subscriptions, 90% of workers reported daily use of personal AI tools like ChatGPT or Claude for work purposes. The demand for AI assistance is present. The enterprise supply is not meeting it. (AI Wiki)
What infrastructure monitoring cannot tell you
Most enterprise AI deployments are measured the same way as infrastructure: uptime, latency, error rates, token consumption. These metrics tell you the system is running. They cannot tell you whether your employees are using it effectively, which departments have quietly stopped engaging, or where the AI consistently fails to complete the tasks users actually need.
The organisations closing the ROI gap measure something different. They track what users ask, where they abandon, whether they return, and which tasks resolve successfully. Every interaction becomes a signal. That signal is what drives continuous improvement.
Consider a concrete example. If 30% of employees who try an internal AI agent escalate to a colleague or fall back to email within two interactions, that is a measurable failure pattern. If one department uses an AI agent heavily while another with the same tool barely touches it, the usage data points to a workflow mismatch that can be fixed. Neither of these signals appears in an infrastructure dashboard.
The metrics that connect AI to business outcomes
The teams seeing real ROI track outcomes, not just activity. The relevant metrics are:
Task completion rate. Did the user accomplish what they came for, without escalating or abandoning?
Intent resolution rate. Did the AI address the actual question behind the user's request, not just produce a response?
Hours saved by department. Which teams are gaining measurable productivity, and by how much?
Return rate. Do users come back because the AI is reliably useful, or do early interactions drive them away?
Abandonment timing. Where in specific flows do users stop engaging? That is where the agent needs redesign.
These metrics make ROI visible to leadership in a way that token counts and uptime figures cannot. They also change the conversation from "is anyone using this?" to "where do we scale this next?"
The feedback loop that drives improvement
Mid-market companies that crossed the AI divide moved faster and more decisively, reporting average timelines of 90 days from pilot to full implementation. What they shared was not a better model. It was a tighter feedback loop between user behaviour and product improvement. (Campus Technology)
Treating an AI agent like a product, rather than a deployment, means instrumenting it for user behaviour from day one. It means acting on what the data surfaces. An AI agent that failed to answer compliance questions last month and visibly handles them better this month builds user trust. Users who see improvement become regular users. Regular users generate more signal. The loop compounds.
This is the mechanism the 5% are using. It is also replicable.
Nebuly
Nebuly is the ROI platform for enterprise AI. It connects to the AI agents your business runs on, the assistants your customers interact with, and the tools your employees use every day, including Claude, ChatGPT, and Copilot, and translates that activity into business value. How much time is being saved across teams. What revenue your AI is influencing. What adoption and AI proficiency look like in practice, across departments and geographies. All aggregated at the organisational level, never tied to individuals.
If you need clarity on what your AI investment is actually delivering, book a demo.
FAQs
Why do most enterprise AI projects fail to deliver ROI?
According to the July 2025 MIT NANDA report, the primary reason is a learning gap, not a technology failure. Most AI agents do not retain feedback, adapt to user context, or improve between sessions. Users who encounter the same failure repeatedly stop engaging. Without a mechanism to capture that signal and act on it, the deployment plateaus.
What separates the 5% of AI deployments that deliver measurable value from the rest?
The organisations seeing real ROI treat every user interaction as a source of intelligence. They measure what users ask, where they abandon, and which tasks fail. They connect those signals to business outcomes like hours saved and task completion rates. That feedback loop is what drives continuous improvement and sustained adoption.
What is the difference between AI infrastructure monitoring and user analytics?
Infrastructure monitoring tracks system performance: uptime, latency, error rates, and token consumption. It tells you whether the AI agent is running. User analytics tracks user behaviour: what people ask, whether they complete their sessions, which flows they abandon, and whether they return. Infrastructure monitoring cannot tell you whether your AI is delivering value. User analytics can.
How do you measure the business value of an enterprise AI agent?
The metrics that connect AI usage to business outcomes include task completion rate, intent resolution rate, hours saved per department, return rate, and abandonment timing by flow. These measure whether users accomplished what they came for, not just whether the system responded. Tracking them over time against pre-deployment baselines is what makes ROI visible to leadership.
What is shadow AI and what does it reveal about enterprise AI programmes?
Shadow AI is when employees use personal AI tools for work rather than the organisation's approved systems. The MIT report found this occurs in over 90% of organisations. It is a direct signal that the demand for AI assistance is real, but the official deployment is not meeting user needs. Treating it as a compliance problem misses the point. The usage patterns in shadow AI tell you exactly what employees actually need from an AI agent.
Stay up to date on what we're learning, building, and seeing as enterprise teams deploy and measure AI agents in production.