On January 20, 2026, Automation Anywhere and OpenAI formalized a partnership that asks a quiet but consequential question: what happens when enterprise software stops following instructions and starts thinking for itself? By pairing OpenAI's reasoning models with Automation Anywhere's Process Reasoning Engine, the two companies are attempting to move business automation from rigid, checklist-bound execution toward something more adaptive — systems that can encounter an unexpected circumstance and decide, rather than fail. The ambition is not to replace human judgment, but to reserve it for the
Automation Anywhere Launches AI-Native Agentic Solutions With OpenAI
Most agentic initiatives fail because they're either too autonomous or too constrained
Mark
So what's actually new here? Automation software has existed for years. What changes with this partnership?
Mimi
The difference is in how the system responds when things don't go according to plan. Old automation follows a script—if the script breaks, the whole thing breaks. These new agents can reason about what to do when circumstances change.
Luke
But that's what the companies are claiming. Do we know if it actually works that way in production, or is this still mostly a capability demonstration?
Mimi
Fair point. The announcement is about the partnership and the architecture. Real-world performance data will come later.
Mark
They mention deploying in weeks instead of months. Is that realistic, or is that marketing language?
Mimi
They're saying the solutions are pre-built for specific functions like finance and HR, so you're not starting from zero. But "weeks" probably means weeks of configuration and testing, not weeks of zero work.
Luke
And the human-in-the-loop piece—is that a feature or a limitation? Are they saying humans need to stay involved because the AI isn't trustworthy yet?
Mimi
Probably both. It's a feature because it keeps humans accountable for critical decisions. It's a limitation if you wanted fully autonomous systems. The companies seem to be saying that's actually the right balance.
Mark
What happens if an enterprise deploys one of these and the AI makes a bad decision?
Luke
That's the question the announcement doesn't answer. They talk about "governed execution" and "human judgment," but we don't know what the actual guardrails are or who's liable when something goes wrong.
Mimi
Right. The architecture exists. Whether it prevents the failures they're warning about—systems that are either too autonomous or too constrained—that's still being tested in the field.
The Pulse
- Traditional automation breaks the moment reality diverges from the script — a fragility that costs enterprises time, money, and trust every day.
- The new partnership introduces AI agents designed to reason through disruption rather than halt at it, targeting finance, HR, IT, and customer service as the first proving grounds.
- Deployment is being promised in weeks rather than months, a direct challenge to the slow, expensive implementation cycles that have long defined enterprise software.
- The central tension the companies are navigating is a known failure mode: agentic systems that are either too autonomous to be safe or too constrained to be useful.
- Their proposed resolution is a deliberate three-way balance — AI reasoning, governed deterministic execution, and human-in-the-loop controls for decisions that carry real consequence.
- The January announcement opens a large-scale test of whether that balance holds when these systems meet the unpredictable complexity of actual enterprise operations.
On January 20, 2026, Automation Anywhere and OpenAI formalized a partnership that asks a quiet but consequential question: what happens when enterprise software stops following instructions and starts thinking for itself? By pairing OpenAI's reasoning models with Automation Anywhere's Process Reasoning Engine, the two companies are attempting to move business automation from rigid, checklist-bound execution toward something more adaptive — systems that can encounter an unexpected circumstance and decide, rather than fail. The ambition is not to replace human judgment, but to reserve it for the moments that matter most.
Automation Anywhere and OpenAI announced a partnership on January 20, 2026, built around a specific frustration with how enterprise automation has worked until now. Traditional systems follow predetermined steps and collapse when circumstances change. The collaboration pairs OpenAI's reasoning models with Automation Anywhere's Process Reasoning Engine — which determines what action should happen next and coordinates work across platforms — to create what the companies describe as a complete reasoning-to-action loop.
CEO Mihir Shukla framed the distinction plainly: this is not faster execution of the same old logic, but a fundamentally different way of working. The targeted domains — finance, HR, IT, and customer service — are areas where adaptive reasoning could deliver immediate value, and Automation Anywhere is promising deployment in weeks rather than the months enterprise software typically demands, with domain knowledge already embedded rather than built from scratch by the customer.
The companies are candid about where agentic initiatives typically go wrong. SVP Dustin Snell acknowledged that most fail by swinging too far in one direction — either granting the AI too much autonomy or constraining it into uselessness. Their answer is a deliberate blend of AI reasoning, deterministic governed execution, and human judgment at critical decision points. OpenAI's Giancarlo Lionetti described the goal as changing how work fundamentally happens, not merely accelerating existing processes.
Automation Anywhere is positioning this as an inflection point after more than two decades in the automation business. The underlying architecture is designed to be consistent across business functions, so customers can extend solutions without rebuilding core logic or writing code. What the announcement cannot yet answer is whether the balance between autonomous reasoning and human oversight will hold when these systems encounter the full complexity of real enterprise environments — a question the coming months of adoption will begin to answer.
Automation Anywhere and OpenAI announced a partnership on January 20, 2026, designed to bring a new class of artificial intelligence agents into enterprise operations. The collaboration pairs OpenAI's reasoning models with Automation Anywhere's Process Reasoning Engine—a system that determines what action an enterprise should take next and orchestrates work across different platforms—to create what the companies call a complete reasoning-to-action loop.
The distinction the companies are drawing matters. Traditional automation software follows predetermined steps, much like a human following a checklist. When circumstances change, these systems often fail because they have no way to adapt. The new agentic solutions, by contrast, are built to let AI agents reason independently about what should happen next, solve problems as they arise, and adjust their approach when conditions shift. Mihir Shukla, CEO and Chairman of Automation Anywhere, framed this as fundamentally different from what exists in the market today—not just faster execution of the same old logic, but a different way of working altogether.
The partnership targets specific business functions where this kind of adaptive reasoning could have immediate value: finance, human resources, information technology, and customer service. Automation Anywhere is positioning these solutions as ready to deploy in weeks rather than the months traditional enterprise software typically requires. The company describes them as pre-built and deeply contextual, meaning they come with domain knowledge already embedded rather than requiring customers to build everything from scratch.
But the companies are also careful about what they're claiming. Dustin Snell, senior vice president of agentic solutions at Automation Anywhere, acknowledged that most agentic initiatives fail because they swing too far in one direction—either giving the AI too much autonomy or constraining it so heavily that it becomes useless. The solution, he said, is deliberately blending three things: the AI's ability to reason through problems, deterministic execution that follows governed rules, and human judgment. This balance, he argued, is what separates promising experiments from systems that actually work in production.
Giancarlo Lionetti, chief commercial officer at OpenAI, described the partnership as moving beyond task automation to changing how work fundamentally happens. The framing suggests that what's being built here is not just faster versions of existing processes, but a different model for how enterprises organize and execute work at scale. The Process Reasoning Engine handles the execution side—making sure actions happen securely and in the right order across different systems—while OpenAI's models handle the reasoning about what should happen and why.
Automation Anywhere has been in the automation business for more than two decades, but the company is positioning this partnership as a significant inflection point. The agentic solutions are built on what the company calls a consistent agentic architecture, meaning customers can deploy them across different business functions without having to rebuild the underlying logic each time. The architecture includes human-in-the-loop controls, which means humans remain in the decision-making process for critical actions, and the ability to customize and extend solutions without writing code.
What remains to be seen is how enterprises will actually adopt these systems and whether the balance between autonomous reasoning and human oversight holds up in practice. The companies are making a specific bet: that the future of enterprise automation is not about removing humans from decisions, but about giving them better information and faster execution, with AI handling the reasoning work that humans find tedious or error-prone. The January announcement marks the beginning of that experiment at scale.
Notable Quotes
Our agentic solutions are fundamentally different; they're designed for how AI agents work, enabling them to autonomously reason, solve problems, and adapt to changes.— Mihir Shukla, CEO and Chairman of Automation Anywhere
What's different here is that we deliberately blend agentic reasoning, deterministic execution, and human judgment into a single, governed flow.— Dustin Snell, SVP of Agentic Solutions at Automation Anywhere
Want the full story?
Read the original at Europa Press ↗
Technology