‘But is that real work? It’s not’ Business leaders still don’t trust AI agents, Harvard survey shows

A survey of more than 600 tech leaders by Workato and Harvard Business Review shows strong intent to invest in agentic AI—86% of companies plan to increase spending over the next two years—even as trust remains limited: only 6% fully trust agents with essential end-to-end processes, 43% will use them only for routine operational tasks, 39% for supervised or noncore complex tasks, and 8% trust them not at all. Respondents cited cybersecurity and privacy (31%), data output quality (23%) and business processes not being ready for automation (22%) as the biggest hurdles, and the report found agents are not yet delivering on expectations for productivity, CX or revenue. Workato is promoting orchestration and standardized protocols to coordinate specialized agents, using agents internally for sales prep and usage monitoring, and expects broader adoption to continue into 2026 while truly trusted, multistep autonomous agents remain two to three years away.

Analysis

A survey of more than 600 tech leaders by Workato and Harvard Business Review shows widespread intent to increase spending on agentic AI—86% of respondents plan to invest more over the next two years—while trust remains limited: just 6% fully trust agents with essential end-to-end processes, 43% limit agents to routine operational tasks, 39% use them in supervised or noncore complex tasks, and 8% reject them entirely.

Respondents identified cybersecurity and privacy (31%), data output quality (23%) and business processes not ready for automation (22%) as the primary adoption barriers, and the report finds agents are not yet meeting expectations for productivity, customer experience or revenue. Workato’s CIO notes many current deployments are simple (for example, creating IT tickets) while the firm itself uses agents to prep sales reps with Salesforce and Gong data and to monitor platform usage.

Workato is positioning orchestration and standardized protocols as a solution; the company expects continued adoption into 2026 but says trusted, multistep autonomous agents are likely two to three years away. Market-impact and sentiment signals are cautious, implying gradual commercial upside rather than near-term transformative revenue for platform vendors, with CRM-related sentiment neutral.

Key Decisions for Investors

• Consider selective exposure to integration and orchestration platforms like Workato that benefit from protocol-driven agent coordination, but size positions modestly until agents demonstrate measurable productivity or revenue lift
• Avoid overpaying pure-play agent vendors or making large thematic bets today given low enterprise trust (6% for end-to-end) and the 2–3 year horizon for trusted multistep agents
• Monitor leading indicators—enterprise trust/adoption metrics, frequency of security/privacy incidents, improvements in data-output quality, and uptake of standardized protocols or integrations (e.g., Salesforce/Gong use cases)—and increase exposure only after persistent improvements
• Maintain neutral or watchful exposure to CRM (Salesforce) positions; look for concrete commercial deployments and monetization tied to agent integrations before adding to positions