The average Fortune 500 company will deploy more than 150,000 AI agents by 2028, up from fewer than 15 in 2025, according to Gartner predictions cited in a new report from Yugabyte published this week. The article argues that database administrators will spend far less time on hands-on database work and far more time overseeing the autonomous systems that handle it for them. Only 13% of organizations currently think they've got the right governance structures to manage this coming flood of agents, the report notes.
For decades, DBAs have provisioned capacity, troubleshot slow queries, managed migrations, and responded when problems arise. AI agents can now inspect what's happening, decide what requires attention, deploy tools to address it, and verify whether the fix worked, the report explains. Rather than automating a single predetermined task the way scripts and rules do, agents can potentially handle the entire diagnostic and repair cycle. A performance issue that once demanded someone dig through metrics, pinpoint the problematic query, and choose a response could increasingly be investigated by an agent before a human gets involved, according to the article. The DBA doesn't disappear, but the role shifts upward: someone still decides what an agent can do, where human approval is needed, and what happens when it fails.
The report states that not every agent will require its own database, but many will create state, retrieve data, remember past interactions, and share information with other agents. Many will also behave very differently from traditional applications, spinning up quickly, sitting idle for extended periods, and suddenly becoming busy when work arrives. Nobody is hiring 150,000 DBAs to handle them, the article notes. That's the scale challenge Yugabyte is addressing with YugabyteDB AMP, or Agentic Multitenant PostgreSQL, which manages databases as a fleet rather than treating each new agent workload as another database for an administrator to provision individually. The platform consolidates hundreds of small Postgres workloads onto shared distributed infrastructure while isolating their databases, and lifecycle operations including provisioning, branching, scaling, migration, and teardown can be exposed to agents through MCP.
The report explains that agent workloads make conventional capacity planning especially awkward because many are bursty and often idle, meaning permanently provisioned infrastructure for every experimental agent could leave companies paying for databases that spend most of their time doing very little. AMP's approach uses serverless multitenancy and scale-to-zero, where multiple small workloads share the underlying distributed infrastructure and customers pay by CPU minute, with idle agents consuming no compute. Resource governance can impose CPU limits on individual workloads to prevent a single overeager agent from consuming capacity meant for its neighbors. If routine setup, migrations, tuning, and other database operations can increasingly be delegated, a smaller database team can potentially oversee a much larger estate, the article argues. Yugabyte also built specialized agents for setup, migration, performance tuning, and integrations, and has introduced Meko, an agent-native context engine designed for multi-agent AI systems that provides persistent memory, shared knowledge, decision traces, and auditability across multiple agents rather than leaving each agent working from isolated context.
The report concludes that an agent starting as an experiment today could vanish next month or suddenly become a production application used across the business, and building one infrastructure stack for cheap experiments and another for serious workloads risks creating a migration problem every time an experiment succeeds. YugabyteDB AMP lets workloads start on serverless Postgres and transition to fully distributed YugabyteDB as their scale and criticality grow, without rewriting the application or migrating data to a different platform. The safest architectural bet may be one that doesn't require knowing in advance which of today's tiny AI experiments will become tomorrow's critical applications, the article suggests. The DBA remains critical in that world, but the job will look different: managing fewer databases directly while taking responsibility for vastly more of them by overseeing the autonomous systems that do the administering. For organizations still assuming their database teams will continue operating the same way they have for decades, the window to prepare for a fundamentally different operational model is closing faster than most governance structures can adapt. The real question isn't whether agents will take over routine database work, but whether companies can redesign their database organizations before the architecture decisions get made for them.