The current landscape
5 min readHow organizations are deploying AI agents today.
More than half of businesses are deploying AI agents with multi-step workflows
Trend summary
Organizations are deploying AI agents for work that goes well beyond chat interfaces and single-step automation. More than half (57%) now use agents to handle multi-stage workflows, while 16% have progressed to cross-functional or end-to-end processes spanning multiple teams or business functions.
| Response | Startups & SMBs | Mid-Market | Enterprises | Total |
|---|---|---|---|---|
| Agents that perform single-step tasks | 5% | 11% | 9% | 10% |
| Agents that perform multi-step workflows within a single department | 32% | 28% | 28% | 29% |
| Agents that perform multi-step workflows across functions or end-to-end processes | 13% | 15% | 20% | 16% |
| Agents that perform multi-step workflows and operate autonomously on complex tasks with limited human oversight | 8% | 12% | 15% | 12% |
| Agents that perform multi-step workflows (aggregate of the above categories) | 53% | 55% | 62% | 57% |
Why this matters
The shift from task automation to process orchestration represents a fundamentally different use case—and a different value proposition. Organizations that master multi-stage and cross-functional agent deployments can unlock advantages in speed, consistency, and scale that simple automation can't. This is where AI moves from incremental efficiency gains to enabling new ways of working.
Nearly all organizations are adopting coding agents
Trend summary
More than 9 in 10 organizations now use AI to assist with coding. The vast majority (86%) have moved beyond experimentation and are deploying AI coding agents for production code, with enterprises leading adoption at 91% compared to 83% for small and mid-sized businesses. And 42% of organizations trust these agents to lead development work with human oversight, signaling a meaningful shift in how engineering teams are structured and how code gets written.
| Response | Startups & SMBs | Mid-Market | Enterprises | Total |
|---|---|---|---|---|
| To write small code snippets and get code suggestions | 28% | 34% | 32% | 32% |
| To write boilerplate code and suggest code snippets | 27% | 37% | 34% | 35% |
| To build complete features with limited human review | 32% | 44% | 39% | 40% |
| To guide critical architecture decisions | 27% | 34% | 41% | 35% |
| To lead development, with humans in the loop to review code and set strategy | 39% | 39% | 47% | 42% |
| We don't use AI agents at my organization, coding or otherwise | 27% | 15% | 9% | 14% |
| We use coding agents across the SDLC (aggregate of the above categories) | 73% | 86% | 91% | 86% |
Why this matters
AI coding agents have moved from experimental to mainstream, with the majority of organizations already deploying them in production environments. Organizations that embrace these tools strategically are accelerating delivery timelines, optimizing engineering resources, and freeing developers to focus on higher-value architectural and problem-solving work. The split between organizations that trust agents to lead versus assist reveals significant upside for those who invest in building expertise and establishing best practices early.
AI coding agents boost developer productivity
Trend summary
AI agents are increasing productivity across the entire development lifecycle, not just code generation. Organizations report time gains in four key areas at nearly identical rates: code generation (59%), research and documentation (59%), code review and testing (59%), and planning and ideation (58%).
| Response | Startups & SMBs | Mid-Market | Enterprises | Total |
|---|---|---|---|---|
| Bug fixing / debugging | 45% | 56% | 54% | 54% |
| Code generation | 40% | 63% | 62% | 59% |
| Planning and ideation | 45% | 59% | 61% | 58% |
| Reviewing and testing code | 44% | 61% | 61% | 59% |
| Code refactoring / modernization | 40% | 55% | 61% | 54% |
| Codebase navigation | 35% | 53% | 56% | 52% |
| Research and documentation | 49% | 62% | 57% | 59% |
Why this matters
The impact spans every phase of software development, which means teams can improve both engineering velocity and code quality simultaneously. Organizations that integrate AI agents across the full development process can compound these gains, turning what might be a 10-15% improvement in coding speed into meaningful acceleration of entire project timelines. The nearly even distribution across activities also suggests teams are finding value wherever they apply these tools, making this less about picking the "right" use case and more about systematic adoption.
Leaders favor hybrid approaches to agentic development over building from scratch
Trend summary
Most organizations (47%) take a hybrid approach to AI agents, combining off-the-shelf solutions with custom-built components. About one in five (21%) rely entirely on pre-built agents, while a similar share (20%) build their own using APIs, open-source models, or developer toolkits that require coding expertise.
| Response | Startups & SMBs | Mid-Market | Enterprises | Total |
|---|---|---|---|---|
| We both buy off-the-shelf solutions and build ones internally with APIs or developer toolkits | 33% | 47% | 54% | 47% |
| We primarily purchase off-the-shelf solutions | 21% | 21% | 20% | 21% |
| We build agents using APIs or developer toolkits | 25% | 20% | 18% | 20% |
Why this matters
The hybrid model dominance suggests no single approach delivers everything organizations need. Off-the-shelf agents get teams running quickly but often lack the customization required for specific workflows or proprietary systems. Fully custom builds offer control and differentiation but require significant engineering investment. Most organizations find value in the middle: using pre-built agents where they work well and investing development resources only where customization creates meaningful advantage.