Everyone Works with AI Agents, but Who Controls the Agents?
In a recent survey conducted by Salesforce and Deloitte, 1,050 IT decision-makers within enterprise organizations shared their insights on the growing adoption of Artificial Intelligence (AI). While these decision-makers recognize the value of AI, they also highlight significant challenges. As we move into 2026, the adoption of AI is expected to continue its upward trajectory, with agent-to-agent communication becoming increasingly common.
The Complexity of AI Adoption
AI has emerged as a critical tool for enhancing organizational efficiency. However, it remains in its early stages, necessitating clear frameworks for successful implementation. Currently, the average enterprise organization, defined as one with over 1,000 employees, employs about 12 AI agents, a number projected to rise to 20 by 2027.
Current Challenges in AI Utilization
Despite the enthusiasm for AI, decision-makers face several hurdles:
- Integration Issues: The average number of applications within organizations has increased from 897 to 957, with only 27 percent of these applications integrated.
- Outdated Infrastructure: 40 percent of decision-makers report that legacy IT infrastructure hampers their ability to utilize data for AI.
- Data Silos: The persistent issue of data silos complicates AI implementation, with 42 percent of organizations still assessing their risks.
- Lack of Expertise: 41 percent indicate a shortage of AI expertise within their organizations, highlighting the need for increased knowledge.
The Shadow AI Dilemma
Many IT decision-makers are hesitant to adopt AI due to the complexities involved. Concerns range from the investment required to manage various AI solutions to the governance issues they present. Notably, 83 percent of organizations are already using AI across virtually all teams, yet 86 percent of IT decision-makers believe that AI agents could create more problems than they solve. The rise of Shadow AI poses significant risks, as employees may resort to free AI tools, inadvertently exposing sensitive company data.
Embracing Agent-to-Agent Communication
As we approach 2026, agent communication is on the horizon. Protocols enabling agents to communicate and transfer tasks aim to enhance operational efficiency. Currently, 50 percent of AI agents operate in isolation, lacking the necessary context or external data. However, 96 percent of IT decision-makers recognize that seamless integration is essential for success.
Emerging Solutions for Governance
Many large platform providers are acknowledging governance challenges and are actively seeking solutions. For instance:
- ServiceNow launched the AI Control Tower to monitor all agents on its platform.
- Workday aims to integrate AI agents as part of HR databases but lacks management capabilities for third-party platforms.
- MuleSoft has introduced the MuleSoft Agent Fabric, allowing for governance and compliance in data flows between applications and agents.
Actionable Steps for IT Decision-Makers
For IT decision-makers contemplating their next moves in 2026, inaction is not a viable option. Instead, they should:
- Start Small: Implement simple AI applications to achieve significant gains without overhauling existing infrastructure.
- Engage Employees: Conduct surveys to gauge employee sentiment towards AI and identify potential areas for improvement.
- Offer Guidance: Provide training and support for AI tools to enhance adoption and reduce reliance on Shadow AI.
The Future of AI in Organizations
As organizations grapple with the rapid pace of AI innovation, they must adapt to ensure that their AI strategies align with emerging technologies. With the right governance and planning, AI can offer substantial value, but organizations must remain proactive to avoid falling behind.
