Appian's 'AI needs process' thesis goes mainstream - now it has to prove the math works

There's a particular kind of vindication that comes from watching the rest of the industry arrive at your thesis years after you planted the flag. Matt Calkins, Appian's founder and CEO, is clearly enjoying the moment:

Without a process framework, AI cannot add value to complex work streams o collaborations. Market analysts and researchers like Gartner and MIT published papers on the topic. Customers, prospects and partners have all confirmed the trend. Our competitors shifted their messaging, began talking about workflow and added rudimentary process technology.

That last line is the telling one. When your competitors start borrowing your language, something has shifted. The argument that AI is probabilistic and therefore needs a deterministic governance layer to be trusted with high-value enterprise work is no longer an Appian talking point. It's becoming orthodoxy.

The numbers

Appian just turned in some good quarterly numbers. Q4 delivered cloud subscription revenue of $117 million, up 18% year-over-year, with total revenue hitting $202.9 million, a 22% increase. For the full year, total revenue came in at $726.9 million, up 18%, with adjusted EBITDA of $76.8 million representing an 11% margin. That margin number matters – two years ago, it was negative 8%. The company generated $62.9 million in operating cash flow, compared to a $110 million loss just two years prior.

The headline bookings metric was strong too. Q4 cloud net new annual contract value (ACV) growth was the strongest in almost three years, and the quarter was notably back-end loaded, meaning much of that new business hasn't yet flowed into reported revenue. Cloud net annual recurring revenue (ARR) expansion ticked up to 114%, and the number of customers spending over $1 million annually in ARR grew from 115 to 140.

Calkins points to a 50% increase in the number of customers purchasing over $1 million in software, with the total value of seven-figure transactions nearly doubling. AI usage on the platform grew 14x year-over-year, and Appian is monetizing that growth through its advanced AI license tier, which carries an average price increase of 25%.

Washington comes calling

A $500 million enterprise license agreement with the US Army dominated the narrative - it is already an eight-figure ARR customer. Calkins pitches it as a threshold moment:

This is an important moment in the growth of this organization. It represents a degree of confidence that an agency has not in the past shown in us.

The agreement creates a framework for the Army to purchase Appian software and services over the next decade to modernize legacy systems. Calkins says the conversation was driven heavily by legacy application modernization – a topic he notes stops executive conversations cold whenever it comes up. He intends to use the Army deal as a credential across other government departments and with partners.

More broadly, Appian is benefiting from a structural shift in how the US government procures technology. The current administration's emphasis on efficiency and direct vendor relationships, rather than working through systems integrators, plays to Appian's strengths. Chief Financial Officer Srdjan Tanjga notes that the federal business drove both software and professional services growth, with the latter up 36% in Q4 – its strongest performance in eight years.

Proof in production

A consistent point through the earnings call is how Appian plans to convert the conceptual argument – AI needs process – into repeatable, demonstrable value. Calkins highlights a product called Doc Center, which ingests documents, launches workflows, uploads data and accelerates response times. It's not a conceptually difficult product, but that's partly the point. It works as a proof case because the results are immediate and measurable. One pharmaceutical customer saw an 88% reduction in response times.

Calkins is explicit about the strategy:

Where we see that we've got some of these winning demonstrations that establish how you could make value with AI, we're going to put the pedal down.

Get customers onto the advanced AI tier, deploy a high-visibility use case like Doc Center, demonstrate rapid ROI, then expand. Tanjga explains the upsell mechanics – customers start with proofs of concept, upgrade to the AI tier for production workloads, come back for second and third workloads, and eventually move to higher tiers as new functionality arrives. It's a land-and-expand model built specifically around AI adoption.

The elephant in the room

Calkins also spent considerable time addressing investor concerns about whether AI represents an existential threat to software companies, including Appian. His argument is twofold. First, that AI's probabilistic nature makes it fundamentally unsuited to unsupervised enterprise work – and that this isn't a temporary limitation but an inherent characteristic. Second, that AI-generated code, like open source before it, doesn't threaten Appian's value because customers aren't buying code. They're buying compliance, reliability, community and support. He draws the comparison directly:

This current concern about AI-generated code reminds me of the open source scare years ago. Open source seemed to threaten the pricing power of the entire sector. But in the end, it proved that code isn't the center of value in enterprise software.

This is something that clearly resonates with Appian's customer base – approximately 80% of revenue comes from highly regulated industries and government. No one is going to vibe code a defense logistics system.

My take

The idea that AI needs a deterministic orchestration layer to deliver enterprise-grade reliability is not controversial anymore, and Appian has genuine advantages here – two decades of process technology, deep government relationships, mission-critical deployments at scale. The efficiency turnaround is also real and impressive; going from burning cash to 11% EBITDA margins in two years is not trivial.

But there's a gap between winning the argument and winning the market. Cloud subscription growth is guiding to 16% for the full year 2026, which is solid but not the kind of acceleration you'd expect if the thesis were translating into a demand surge. Some of that is FX – Q1 benefits from a tailwind that normalizes for the rest of the year – and some is timing, given how back-end loaded Q4 was. Tanjga is also clear that the company is returning to investment mode, growing sales capacity for the first time in a while, which should help but takes time to pay off.

Calkins is almost certainly right about AI needing process. What's less certain is whether Appian can move fast enough to own that position before larger platform vendors build or acquire their way into the same space. The competitors who started borrowing Appian's language won't stop there. Calkins acknowledges as much when he describes their process technology as "rudimentary" – but rudimentary has a way of becoming adequate when it's backed by a larger distribution machine.

The $500 million Army ELA and the 14x AI usage growth are strong signals. The next few quarters will show whether they're the beginning of a turning point or just a very good company executing well within its existing orbit.