Become a TechRadar Insider
When a specialist AI agent company can raise $950m at a valuation above $15bn, enterprise AI has clearly moved beyond experimentation.
Where buyers once asked AI tools to carry out simple tasks such as summarizing documents, answering support questions or drafting code, the question is now whether the technology can own business outcomes, resolve customer issues, prepare claims, reconcile data, plan work and trigger actions across core systems.
CTO and co-founder at Star.
Yet the mood inside enterprises is complicated. The debate is still too often framed around one narrow question: is AI delivering productivity gains?
The answer is yes. In many organizations, it already is. Across technology, operations, marketing, service and back-office teams, AI is helping employees automate workflows and use agents to perform tasks that previously consumed large amounts of human time. Although productivity gains are easy to measure and often the most visible, productivity is not the most important point.
Gartner has predicted that more than 40% of agentic AI projects will be cancelled by the end of 2027 because of rising cost, unclear business value or weak risk controls. The warning points to a deeper issue: many organizations are applying a new technology to an old operating model, putting copilots, assistants and agents on top of workflows that were designed for a slower, more predictable business environment.
When you strip away the hype, an AI strategy comes down to two key priorities – business optimization and business transformation.
Business optimization is about using AI to do what you already do, but better: improving efficiency, reducing redundancy, eliminating manual effort, strengthening existing revenue engines and helping people make better decisions.
Business transformation is about using AI to do something different: creating new products, new services, new revenue models and new ways to generate value that were not viable before.
The move toward autonomous enterprises
For CIOs and CTOs, the immediate question is where the organisation sits on the maturity curve. It’s simple to think about enterprise AI transformation across five levels of autonomy.
L1 – Assisted Automation, defines an initial AI adoption phase which is implemented based on the active use of assistive AI co-pilots. Decisions are still made by people and require human execution. Enterprise systems interaction is also still performed by human operators.
L2 – Partial Autonomy, where AI takes over bounded decisions and execution in clearly scoped domains with guardrails and thresholds in place. Here, people handle exceptions and provide decision supervision. AI agents take ownership of enterprise systems interaction within these automated domains.
L3 – Cross-Functional Autonomy, where multiple agents coordinate across functions, and rely on outcome-driven (as opposed to fixed-workflow based) optimization.
L4 – Near-Autonomous Enterprise, where an enterprise runs itself in purely AI agentic mode. AI agents plan, execute, monitor and correct execution within set policy constraints. People define strategy and ethics and set governing policies.
L5 – Fully Autonomous Enterprise, where AI sets sub-goals, reconfigures organizational execution and autonomously refines strategy within agreed bounds. At this level, people act as the board, ethics, and risk authority.
A realistic goal over the next two to five years is progress from Level 2 to Level 3 in high-volume, well-instrumented domains where data quality, process ownership, controls and ROI metrics are already strong.
Three capabilities
There are three capabilities here that matter.
The first is self-learning, which treats its own operations as a continuous source of intelligence.
The second is self-adapting, which can sense changes in its environment and reconfigure priorities, workflows and resource allocation in response.
The third is self-correcting, which builds feedback loops into everyday work. Actions are measured against outcomes.
When these three capabilities come together, enterprise systems go beyond supporting the business and become part of its adaptive capability.
Rethinking work around intelligent systems
That shift requires leaders to ask which decisions can safely move closer to execution. Considering which tasks can be automated is no longer enough, with leaders needing to focus on the outcomes that can be continuously optimized. Instead of enquiring how many agents have been deployed, CIOs and CTOs should be asking how quickly the operating model learns.
Humans will still be essential for successful organizations, but the nature of work will change. More value will come from setting direction, defining constraints, supervising intelligent systems and making the calls that require judgement, accountability and trust.
Leaders should decide which decisions can safely move closer to execution, which should remain human-led and which outcomes should be continuously optimized. Autonomy without a clear operating philosophy will create risk. Autonomy guided by strategic intent can create speed and resilience.
An autonomous enterprise does not mean humans disappear. There is a growing debate about whether AI will enable the rise of the single-person unicorn: a billion-dollar company operated by one individual and a fleet of intelligent agents. Technically, parts of that vision may become possible. But it should not be the endgame.
The most important question we must ask as a technology community is how AI can help organizations create better value for their employees, customers and society.
We've featured the best IT automation software.
This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.
The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit