Automation Anywhere says 61% of Q4 software bookings came from AI. That matters because it shows buyers are moving budget toward agentic automation tied to real business outcomes, not just experimentation.
Article text
If you want to know whether AI demand is real, stop asking how many executives say they are interested.
Look at where the bookings are going.
That is why Automation Anywhere’s latest number matters: the company says 61% of its fourth-quarter software bookings came from AI-powered offerings.
That is not a vanity adoption metric. It is not a waitlist number. It is not another “leaders believe AI is important” survey.
It is budget.
And budget is a much better signal.
According to the company’s April 2026 results, that demand showed up alongside 23% growth in enterprise customers with more than $1 million in ARR, more than 2x growth in its agentic customer base, and double-digit growth in ARR and remaining performance obligations. Automation Anywhere also says it reached its 10th consecutive quarter of non-GAAP profitability.
Taken together, that paints a clearer picture than most AI headlines do.
The market is starting to reward AI that is attached to operational work inside real departments.
That is the important part.
What the Number Actually Tells Us
A lot of AI coverage still treats demand like a vibes question.
Are executives excited? Are teams experimenting? Are companies exploring pilots?
That is interesting, but it does not tell you much about where the market is maturing.
Bookings do.
If 61% of new software bookings are coming from AI-powered products, it suggests buyers are no longer treating AI as a sidecar. They are increasingly buying it as part of core operational infrastructure.
Automation Anywhere frames this as a shift from experimentation to deployment. I think that is the right read.
The company’s CEO, Mihir Shukla, said customers are “moving past pilots and into production” by using AI solutions with fast time-to-value and a clear path to scale. He also described adoption at the department level: autonomous ITSM, autonomous finance, autonomous support.
That language matters because it points to a more practical buying pattern.
Companies are not just buying “AI.” They are buying:
AI for IT service operations
AI for finance workflows
AI for support operations
AI for healthcare administration
AI for high-volume back-office process work
That is where budgets get serious.
This Is What Real Market Pull Looks Like
Real market pull is not when a new AI feature gets applause.
Real market pull is when buyers commit money because the software is close enough to a measurable business outcome that the purchase makes sense.
That usually means one of five things is true:
The workflow already hurts.
The owner of the problem is obvious.
The value can be measured.
The deployment path is credible.
The system fits into an existing operational stack.
Automation Anywhere’s results suggest it is checking enough of those boxes for a meaningful slice of enterprise buyers.
The company says it is selling both pre-built, pre-validated agentic solutions and a platform for custom extensions. That matters more than it sounds.
The pre-built side reduces the distance between interest and production. The platform side gives larger customers room to adapt the system to their own environments.
That combination is usually where durable software demand starts to show up.
If you only sell a flexible platform, buyers worry about implementation drag. If you only sell rigid pre-built workflows, buyers worry it will not fit their environment.
The middle ground is where vendors can move from pilot energy to booked revenue.
Why This Is Bigger Than One Vendor’s Quarter
I do not think the main lesson here is “Automation Anywhere is winning.”
The bigger lesson is that the market is becoming more selective.
For the past two years, a lot of AI products got attention simply for proving they could generate something: text, summaries, support replies, draft code, analysis, recommendations.
That phase is ending.
The next phase is about operational fit.
Can the system reduce handling time? Can it resolve requests end to end? Can it work inside governed enterprise environments? Can it attach to a department budget with a clear owner? Can it show value fast enough that procurement does not kill it?
That is a much harder standard.
It is also a healthier one.
The companies that keep growing in this market will probably be the ones that connect reasoning to execution inside existing business processes.
That is also why Automation Anywhere’s earlier January announcement with OpenAI matters in context. The company is pairing OpenAI’s reasoning models with its own Process Reasoning Engine, which it describes as the governed execution layer for enterprise work. Whether or not you buy the branding, the strategic direction is clear: reasoning alone is not enough. Enterprise buyers want a path from model output to controlled action.
That is what software budgets are starting to reward.
The Most Important Detail: This Is Department Money
One easy mistake in AI analysis is talking about “the enterprise” like it buys software as a single brain.
It does not.
Budgets show up when a specific team has a specific pain:
IT has too many repetitive service requests
finance has slow, error-prone reconciliation and process work
support has too much queue volume and too much manual triage
operations has too many handoffs and too much swivel-chair work
healthcare administration has too much scheduling and data-management friction
Those are not abstract innovation goals. They are operating problems.
That is why I think this number is useful beyond large enterprises.
It tells smaller companies something important about how AI buying is maturing.
The safest way to create demand is not to lead with intelligence. It is to lead with throughput, cost, response time, or service quality inside a workflow that already matters.
That lesson travels down-market very well.
The Signal Behind the 23% Growth in $1M+ ARR Customers
The 61% bookings number is the headline, but the 23% growth in customers above $1 million in ARR is also worth watching.
That usually points to expansion, not just curiosity.
Big customers do not keep growing spend on workflow software because they liked the demo. They do it because the software is landing in additional functions, handling more process volume, or proving economic value strongly enough to justify broader rollout.
Automation Anywhere also says its agentic customer base more than doubled.
If that claim holds up over time, it strengthens the same thesis: buyers are moving from isolated experiments to scaled deployments where AI is expected to participate in the operating model.
That is the real story.
The AI market is slowly shifting from feature novelty to budget durability.
Why Time-to-Value Is Quietly Doing Most of the Work
One line in the company’s framing stands out: fast time-to-value.
That phrase is easy to skip past, but it is probably doing a lot of heavy lifting.
AI products do not win just because they are capable. They win because they can get from evaluation to operational value before the buyer loses patience.
This is especially true in automation.