Key Considerations for Businesses Automating Work With AI Agents

By Harvey Hu, Founder & CTO at General Agency

Caption: Getty Images

Businesses are rapidly adopting AI-first strategies, with companies like Duolingo and Shopify leading the charge. According to McKinsey, 78% of companies already use AI in their operations. However, before jumping on the automation bandwagon, leaders must understand five critical aspects of AI implementation.

1. Predictive vs. Generative AI

Previous AI waves relied on predictive models, which use historical data to forecast specific outcomes like ad clicks or video completion rates. These models were narrow in scope and expensive to develop. Generative AI, however, creates new content—text, images, or audio—without strict targets. While versatile, generative AI is prone to "hallucinations," where it confidently produces incorrect information.

2. Horizontal Agents vs. AI Workflows

Most "AI agents" today are actually AI workflows—predefined sequences for specific tasks, often built on tools like Zapier. True horizontal AI agents autonomously decide actions and adapt to broader tasks. These are emerging in 2025 and promise better user experiences with advanced reasoning.

3. Target High-ROI, Repetitive Tasks

A Zapier study found that 94% of knowledge workers waste time on repetitive tasks like data entry or report formatting. Automating these can recover ~250 hours per employee annually, boosting productivity and morale.

4. Data Security and Privacy

Treat AI like a new hire. After Samsung engineers leaked code via ChatGPT, the company banned public chatbots. Apply the "intern test": restrict access to sensitive data.

5. Human Supervision

AI errors can be costly. Lawyers were fined $5,000 for citing fake ChatGPT cases. PwC advises .