This week’s AI developments highlight how organizations are finding more specific applications for artificial intelligence while confronting the challenges of measuring results and maintaining oversight. Businesses are using AI to automate IT support, identify cybersecurity vulnerabilities, and improve customer service, while new research raises questions about whether productivity gains are translating into meaningful financial savings. These developments illustrate the importance of selecting practical use cases, maintaining employee involvement, and evaluating AI according to measurable business outcomes.
1. Gartner Finds AI Productivity Gains Are Not Translating Into Expected Cost Savings
New research from Gartner reveals a disconnect between executive expectations and the financial results of AI adoption. Although 77% of board members expect AI to generate productivity improvements and cost savings, only 3% of organizations are exceeding their cost-saving goals through productivity gains. Gartner suggests that companies may find greater value by redirecting employee time toward new services, business capabilities, and revenue opportunities rather than relying exclusively on efficiency improvements. For business leaders, the findings reinforce the importance of measuring AI through outcomes such as customer satisfaction, service quality, and business growth alongside traditional productivity metrics.
2. Lucid Motors Uses AI to Modernize Employee IT Support
Electric vehicle manufacturer Lucid Motors is using AI-powered workflow automation to modernize its internal IT operations. According to Reuters, Lucid migrated from ServiceNow to automation provider Serval in just 45 days. Serval’s technology uses AI agents to manage employee access requests, troubleshoot technical problems, and oversee software and hardware assets. The company also separates agents used to create workflows from those employees use to complete approved tasks, while retaining human approval for sensitive actions. The deployment provides an example of how AI can help IT departments reduce administrative work and improve employee support without eliminating established security controls.
Read the full story on Reuters
3. Mistral Releases an Open AI Model Designed for Cybersecurity
French AI company Mistral released the weights for its Large 4 model, making it possible for organizations to download and operate the technology within their own environments. According to Reuters, the model is designed to help cybersecurity teams identify and address software vulnerabilities, and the public release follows testing with government and corporate users. Unlike AI systems that require access through an external provider, a locally deployed model can give organizations greater control over where information is processed. Potential applications include vulnerability analysis, security investigations, and assistance with remediation, particularly for businesses with strict data management requirements.
Read the full story on Reuters
4. Salesforce Introduces Automated Shutdown Controls for AI Agents
Salesforce expanded its Agent Fabric capabilities with tools designed to evaluate AI agents and stop them when their performance falls below established requirements. According to VentureBeat, the technology allows organizations to monitor agent behavior and trigger shutdowns when systems fail evaluations. This approach has practical applications for companies using agents to update CRM records, respond to customers, or complete transactions across connected applications. Rather than depending entirely on employees to identify errors, businesses can establish testing criteria and intervention procedures that operate continuously.
Read the full story on VentureBeat, Salesforce Agent Fabric coverage
5. Companies Are Rehiring Workers After Discovering AI’s Limitations
Some employers are reconsidering workforce reductions after discovering that AI systems cannot reliably perform every responsibility previously handled by employees. The Associated Press reports that Klarna resumed customer service hiring after relying heavily on chatbot technology, while Ford hired or promoted approximately 350 experienced engineers when AI quality-control tools proved insufficient. These examples demonstrate the importance of evaluating complete job responsibilities before automating them. AI may perform repetitive activities efficiently, but employees often contribute institutional knowledge, judgment, customer relationships, and problem-solving capabilities that are difficult to reproduce through software.
Read the full story on Associated Press
6. New Research Reveals Security Gaps in Autonomous AI Deployment
New research from cybersecurity company Illumio suggests that organizations are giving AI agents more authority than their monitoring systems can adequately support. According to TechRadar, 70% of surveyed organizations have delegated decisions previously made by employees to AI, but only 49% can audit agent actions. The study also found that 76% experienced at least one AI-related security issue during the previous year. Reported problems included unauthorized data access attempts, unapproved connections, and risky recommendations. For IT leaders, the findings reinforce the importance of limiting agent permissions, documenting actions, and ensuring employees can intervene before an automated decision creates operational or security problems.
Read the full story on TechRadar
Why It Matters
- AI productivity improvements do not automatically translate into financial savings, making broader business performance measures increasingly important.
- Practical IT deployments demonstrate how AI agents can automate employee support while maintaining human approval for sensitive actions.
- Locally deployed AI models may give cybersecurity teams greater control over sensitive information and vulnerability analysis.
- Automated evaluation and shutdown capabilities can help organizations manage AI agents operating across enterprise applications.
- Companies are discovering that human expertise remains necessary when automation cannot account for complex responsibilities and exceptions.
- AI security practices must evolve as agents gain more authority to access information and make decisions.
Stay Ahead with the AI Insiders Newsletter
Join more than 55,000 AI executives and innovators for exclusive insights, strategies, and trends shaping the future of AI leadership. Subscribe to the AI Insiders Newsletter.
