AI Agent Consolidation: Why Fewer, Smarter Agents Beat Sprawl

The shift from agent proliferation to consolidation

As enterprises move AI agents from pilot projects to production environments, a counterintuitive pattern is emerging: organizations are finding greater success by consolidating capabilities into fewer, more versatile agents rather than deploying specialized agents for every workflow.

At the recent Snowflake Summit in San Francisco, technology leaders from Synopsys and Fanatics described how they're rethinking their approach to agentic AI deployment. The central insight: adding "skills" to existing agents often proves more agile and cost-effective than creating entirely new agents, each with its own governance requirements and cost structure.

"You have to be careful in 'skills' versus 'agents' thinking," said Sriram Sitaraman, CIO at Synopsys. "Do you want to automate something, or do you want to actually create an agent, which is a different cost structure, usage pattern, and governance and all those things?"

Why it matters

This consolidation trend has significant implications for enterprise AI strategy and budgets. As companies race to deploy AI agents across departments, the temptation to create a new agent for every use case can lead to management overhead, integration challenges, and escalating costs. The skills-based approach offers a more sustainable path to scaling AI capabilities while maintaining control.

From specialized agents to knowledge agents

Synopsys initially deployed purpose-built agents—a revenue agent for finance reports, a debug agent for data center ticketing. Now the company is evolving toward what Sitaraman calls "knowledge agents" that can simultaneously optimize across multiple dimensions: quality, timeliness, and cost-effectiveness.

This represents a departure from traditional trade-offs where organizations had to sacrifice one dimension to excel at two others. "By focusing on data, we could move all three metrics more positively," Sitaraman explained.

Maddie Want, vice president of data at Fanatics, echoed this skills-first philosophy. "Skills have turned out to be a more agile and smaller unit of currency," she said, noting that many requests for new agents can be satisfied by codifying specific knowledge and sharing it across the organization as a skill rather than building a standalone agent.

Agents improve with scale

Unlike traditional software that can become unwieldy as it grows, AI agents demonstrate a different scaling characteristic. "It doesn't matter how much data volume you throw at it, because AI is truly a linear scale," Sitaraman said. "The more data it has, the better decisions it makes."

At Fanatics, Want observed that agent quality has improved while supervision requirements have decreased. "Over time, the degree of investment we had to make in the context layer is decreasing," she said. "And the degree of supervision an agent needs before its able to start autonomously answering questions is decreasing."

The governance challenge

While agents can expand their capabilities, this flexibility introduces new management challenges. Sitaraman cautioned that AI agents can quickly exceed their initial scope. "You may have a sales-ops agent, but there's nothing stopping it from being a sales analyst agent, and a sales-something-else agent," he noted.

This scope creep isn't necessarily problematic, but it requires frameworks to manage intent and context. Want observed that lines are blurring between analysis-focused agents and those handling operational use cases, as users increasingly want to act immediately on the information agents provide.

These insights were first reported by Joe McKendrick for Forbes, based on discussions at the Snowflake Summit.