Appian CEO: Defining Value From AI In Business Processes

Artificial intelligence is worthless. On its own, any instance of an AI on a cloud server, any AI engine built to churn through agentic functions designed to emulate human workplace (or indeed consumer environment) actions, or indeed any AI model built to sit in shining glory to showcase software engineering excellence… isn’t worth that much.

Rather like a pure mathematical model designed to calculate the shape of a rhomboid vector function to be one day powered by some still-prototyping quantum computing service, it is only when algorithmic intelligence is applied to a real world problem or business objective that it becomes valuable. Students study both pure and applied mathematics for a reason; we need to separate theory from the practical empirical application of any toolset… and AI is no different.

AI Injected Into Business Processes

If we accept these core distinctions, then it’s easy to see why now is such an important to bring AI services forward into functional business use. There has of course been a lot of work going on in the labs of Silicon Valley and elsewhere, so now is the time to work out where we are going to put AI to work and quantify its true worth. If we are going to make AI truly valuable in commercial terms, we need to graft it and inject it into business processes.

Although business process mining and management has been around for half a century, much of what we did in the nineties was focused on data mining and dovetailing what we could extract into enterprise resource management platforms. Our modern notion of business process intelligence is exponentially more sophisticated and its business value quotient is now of paramount importance.

CEO of Appian Matt Calkins thinks that despite all the investment, AI is not generating enough revenue because it’s not creating enough business value. Known for his position at the helm of a company well-recognized for its business process platform technologies (and the low-code toolsets that it built its heritage upon), Appian is now further expanding its purview to envelop a wider spectrum of stakeholders into the software creation and execution process. Calkins firmly believes that AI can create more functional business value when it is embedded within existing operational processes inside organizations across every vertical. But for this to happen in practice, we need to examine functions including access controls, data integration and scalability.

The Chatbot Conversation Is Over

We know that the last couple of years has seen organizations across every vertical start to explore the implementation of AI to improve everything from content creation and data analysis to the delivery of better customer service. At the coalface where new services are being applied, many firms’ only exposure to AI is in the form of chatbots and so-called copilots alongside other assistants. The Appian and team suggest that this is somewhat “passive AI” in that it is a service that essentially sits and waits to be asked, ready to jump in when called upon.

“The key to unlocking AI’s full potential is embedding it inside a business process. The process zone is where business happens. It’s where companies make decisions, save and spend money, serve customers and scale business operations,” said Calkins. “When AI operates within processes, it gains purpose, governance and accountability – all factors that are essential to delivering business value from AI. For 25 years, Appian has led the market in process orchestration. No company is better equipped to deploy AI in enterprise processes than Appian.”

Big words indeed, so can Calkins a team break down exactly how, when, why and where AI at the business process level manifests itself and executes?

The proposition here hinges on realities that actually make AI easy to deploy. Because AI is a relative newcomer to the enterprise software landscape and many systems exist in a form that has never had AI architected into their original DNA, we often see AI created as an isolated and disjointed project. This is both costly and complex. According to Appian, by embedding AI within a process, enterprises can access valuable AI capabilities when and where they need them. The company says that process gives AI structure and AI is only as useful as the structure surrounding it.

AI, Alongside Humans

“A process gives AI a set of defined goals in a structured flow of work. AI can work alongside humans and automation tools, escalating issues so humans always maintain oversight and control,” said Calkins. “It is process that gives AI data… and AI in isolation is nothing without data. But despite these truths, most enterprises struggle to feed their AI deployments with complete datasets that span across systems, while still ensuring privacy and maintaining access privileges.

By integrating AI into processes, enterprises ensure AI receives quality, real-time data from all operational systems. These same enterprises can then enforce privacy controls to prevent unauthorized access and optimize data governance to comply with regulations (such as GDPR, HIPAA, etc.).”

Calkins and the Appian team say that a process-centric approach in fact makes AI safe. Because AI is powerful (and should not be left to run) amok, processes provide crucial safety mechanisms, including human-in-the-loop approval steps for high-risk actions and escalation paths to ensure AI errors (bias, hallucinations or other computational errors) do not cause harm while we also have activity logs to make auditing and compliance simple.

“Process makes AI measurable. For many enterprises, AI is a black box. They can’t measure the impact. But a process tracks every AI action, allowing organizations to measure performance, identify bottlenecks and optimize outcomes,” said Calkins. “Process also makes AI scalable. A process provides the necessary infrastructure to scale AI use.

The right tooling puts AI to work with security certifications, enterprise scalability and other capabilities such as process orchestration, automation and intelligence. Processes take AI from a collection of disconnected pilots to an enterprise-wide capability.”

Woven Processes Inside Data Fabric

The company has also this month come forward with its Appian 25.1 platform release. This iteration of Appian introduces additional document processing capacity with AI skills, centralized dashboards for monitoring process key performance indicators and the ability to synchronize 10 million rows per record type in an organization’s data fabric.

As a working architecture layer and information-level toolset, a data fabric approach works to interconnect data that may exist across various disparate systems and ultimately provide a unified view of data resources in a virtualised data layer. Using a data fabric, software engineers can make use of data without the need to migrate it out of or away from where it normally resides.

With a typically modern business locating data across enterprise resource planning suites, a variety of databases and a multiplicity of SaaS applications, Appian says its own data fabric offering plays a fundamental role in end-to-end process automation practices that optimize complex business processes.

With the 25.1 platform’s improved AI architecture, the company suggests that organizations can now classify or extract data from hundreds of millions of “pages” (i.e. workflow pages that exist inside applications, web services and elsewhere that go to make up tasks, roles and jobs) per year with AI skills. Even applications handling high volumes of documents will experience fewer delays and bottlenecks with improved processing capabilities of up to 75 times more documents per hour.

“Appian 25.1 makes AI valuable by combining large language models with Appian’s class-leading autoscale process engine and Appian’s unique data fabric,” said Michael Beckley, CTO and founder of Appian. “Data fabrics are increasingly the preferred data plane in companies’ AI stacks, but most are optimised for read-only access and don’t scale well for writes beyond 2,000 rows per record. For Appian 25.1, Appian’s data fabric natively reads and writes 10 million rows per record, enabling AI to be reliably injected into mission-critical processes in all industries.”

Functional AI Implementations

If we take this platform update alongside Appian CEO Calkin’s determined (and arguably genuine) approach to drive us towards the functional implementation of AI inside business processes, we may perhaps be able move forward significantly.

That progression could take us to a place where we can cut through the AI hype and regard AI services for more than just what they are, as we start to consider them for what they can do for us. Who knows, once we appreciate AI in business processes, we may even stop being afraid of “AI taking our job” and start doing work on our behalf.

While artificial intelligence on its own might be worthless, so is a good sports car until we put a driver in it and find a track. It might still be pretty to look at if it just sits there, but even the shiniest objects tarnish with age… let’s make sure we keep AI on course.

Source: Appian CEO: Defining Value From AI In Business Processes