The Messy Middle of AI
Where Companies Are in 2026
An Insight Global survey of business decision makers on AI adoption, roadblocks, and what success actually looks like.
Download the Report
Introduction
Just four years ago, using AI at work was a nebulous concept. Now, it’s at the center of corporate growth plans. Across industries, companies are moving from experimentation toward real AI deployments, new workflows, and new ways of working. Or at least they’re trying.
Most organizations are learning that adoption isn’t the hard part of their AI journey. Operationalizing it across the company is. It’s transformation, culture, product roll out, workflow disruption and enhancement, and more all wrapped in one.
Thus, many call where we are now “the messy middle,” because so much is still unclear about AI’s impact on people, productivity, and work in general.
We surveyed 501 business decision makers—96 percent of whom said they are the sole decision maker on AI initiatives—to find out where they believe they are in the AI journey, where they’re getting stuck, and how they’re measuring success.
89% of business leaders say an AI initiative has stalled al their company.
Stalling Is the Norm
89 percent of leaders reported that an AI initiative has stalled at their company. Only 11 percent said they haven’t had one stall.
Some of the most commonly cited reasons for this include:
- Deploying AI into systems or workflows—29%
- Compliance or accuracy risk—29%
- Measuring impact—28%
- Building or engineering the AI solution—28%
- Organizational adoption—27%
- Identifying use cases—25%
- Data readiness—23%
How They’re Measuring Success
We also asked leaders how they measure success in their AI initiatives and let them select multiple measures. The spread tells a clear story: businesses want AI to solve a range of problems, and there’s no single definition of “working.”
How Leaders Can Act on AI Strategy
Define Your Version of Success
The data shows that success with AI means something different to each company. What matters is that it’s defined. Start by choosing a few outcomes you can baseline—cycle time, quality defects, customer experience, risk reduction—and align on what “better” looks like across a timeframe—a quarter, a year, and so on. Tie each metric to the workflow you’re changing, not the tool you’re deploying.
Get the Foundation Right
AI use cases work when data quality, security, governance, and cloud strategy are all in good shape. The survey confirms this practice: data quality and security were the top two barriers leaders identified.
Know Where You’re Likely to Get Stuck
Nearly 9 in 10 companies reported having an AI initiative stall, and those stalls usually happen once an initiative meets real-world requirements like adoption, production, or risk review. A good readiness assessment across your organization will reveal potential stall areas before they become real problems.
Plan for a Hybrid AI Environment
Most businesses will adopt a mix of off-the-shelf and custom-built AI tools. Success depends on operational alignment and governance: clear roles for managing, tuning, and deploying AI, connected through shared workflows and platforms.
Get the full picture
Read the full report for a deeper look at AI adoption by industry, where agentic use cases are being deployed, and what the next 12 months look like for enterprise AI.
Resource Center
Why AI Governance Should Start From Day Zero
By
Over the last few years, enterprise conversations about AI have shifted quickly. From building agents, applying models to workflows, and exploring where AI can support decision-making, teams are building systems that support real workflows and…
How Leaders and Organizations Are Using AI in the Workplace
By
Uncover the current landscape of AI in the workplace. Learn which industries are leading the way in AI technology adoption.
What Is Physical AI & How Are Companies Using It?
By
Learn what physical AI is, how physical AI companies use it across industries, and why it matters now as AI moves into real-world action.
From Hallucinations to Trusted Autonomy: Enterprise Agent Tuning at Scale
By
AI agent tuning boosts task completion to 82% and tool accuracy to 96% while reducing hallucinations and improving reliability.
Survey Methodology
Insight Global commissioned Atomik Research to conduct an online survey of 501 executive-level business leaders from VP level, president/business principal, and C-suite roles who have decision-making influence on their organizations’ AI implementation efforts. The margin of error is approximately +/- 4 percentage points with a confidence level of 95 percent. Fieldwork took place between March 27 and March 30, 2026.