Can process automation really help firms save billions?

Process management may not be the showiest side of enterprise tech, but for Nasdaq-listed enterprise tech firm Appian it’s a $2.5bn business.

The ‘low code no code’ US software company automates and simplifies business processes for a variety of customers, whose use cases are as far-ranging as procurement, patent registration and the design of space suits.

End-to-end process automation can help speed up operational processes that might otherwise be viewed as time-consuming, bureaucratic and repetitive.

Appian claims it can do this with ease, using drag-and-drop interfaces, that save users from writing lines of code and make the business of managing processes and creating new ones accessible to a wider range of people than data analysts or devops.

Typical use cases can be listed across the public and private sectors. For instance, the US Army used Appian’s software recently to implement a new contract writing system and, through process automation, was able to deploy its new application to 400 users across a dozen states and locations outside the US in just 23 weeks.

Another customer, UK bank NatWest, recently used Appian to implement an internal governance policy change. By automating almost 50% of its governance process data, the bank claimed that it saw product governance time drop from just under five days to less than 20 minutes.

There are even celestial applications of process automation. Space infrastructure provider Axiom is deploying the tech so that its clients can track the entire lifecycle of one of the firm’s next-generation spacesuits for a future mission to the moon.

Gen AI revolution

In the year of its 25^th anniversary Appian’s four cofounders, Matt Calkins, Michael Beckley, Robert Kramer and Marc Wilson –  who all still hold active C-suit roles at the Tyson’s corner-based company –  find themselves in the middle of data and AI revolution.

How can process automation keep up with what user-friendly Gen AI and LLMs have to offer in terms of their ability to automate and analyse structured and unstructured data internally and externally, within seconds?

Tech trends have come and gone before, acknowledges the vendor’s considered CTO, Michael Beckley. The engineer – an avid bookworm – points out that Appian was founded a few months before the dotcom crash at the end of the Nineties – an event that saw the firm pivot from commercial to government customers almost overnight, before all of its initial funding ran out.

But it’s been the dawn of popularised LLMs like Open AI’s ChatGPT that has forced the company to “go back to school”, in Beckley’s words.

Beckley spoke to TI on the eve of the vendor’s two-day customer and partner focussed tech event Appian World, which saw 2,000 delegates gather to learn new announcements about the firm’s products and to gain insights from customer use case sessions.

This year, the Virginia-founded firm returned closer to home for its event, which took place in mid-April, at a waterfront resort on the calm waters of the river Potomac, a stone’s throw from the US capital.

“What AI is doing now is so different and so compelling that it is changing the way we think about it,” continues Beckley.

“While ChatGPT has not lived up to the hype, large language models have. And yet, for enterprises, we’re not there yet,” he states.

For a whole variety of reasons, he claims, getting practical value for enterprises from AI has been more challenging than it has for students using ChatGPT to cheat on exams or term papers.

“The business world has had to realise that ChatGPT has not been great at replacing everything – but LLMs are going to be important this year.

“LLMs are already delivering practical value, but they are not replacing humans – they are augmenting them.”

The fact is that regulated businesses – and most of the firms’ customer base – financial, legal and governmental organisations fall into this category – must adhere to strict rules, they are accountable by law.

To this end, there are many grey areas that risk data infringements in the process of training and using LLMs, from IP rules to personal data legislation, which means that, no matter how much they improve business processes and outcomes, experimenting with them is considered high risk.

Appian AWS deal

Addressing some of the issues Beckley outlines above, Appian sees value in private AI, applying LLMs to a company’s own internal data, but not training these models on this data.

Training an LLM on company data isn’t necessary, he explains, as an LLM will use its understanding of language to read the material it’s given. This allows companies to use a range of AI models in a secure way.

At Appian World the vendor announced a deal with AWS, to enable customers to harness the capabilities of Amazon’s all-you-can-eat Gen AI service Bedrock and its LLM builder SageMaker in their own private AI environments.

Beckley explained how the vendor’s customers can use its own virtualised data layer, called Data Fabric,  and a technique called Retrieval Augmenting Generation (RAG) to pass only the documents and information it needs to the LLM.

“So now a bank or a government agency can have a LLM that’s private,” says Beckley.” They can use our Data Fabric to ensure that they don’t need to even train that model.

“You want to ask an AI: Tell me the highest performing region within these regions I told you about? I don’t need to train the models on my regions, the model just knows how to use its understanding of language to read the material I give it and give me the best answer.”

Process makes perfect

In 2021, Appian acquired Berlin-based start up Lana Labs, which specialises in process mining  – a method of applying data science to discover, validate and improve workflows.

With this acquisition, into the Appian fold came the start up’s three cofounders: Rami-Habib Eid-Sabbagh,  Thomas Baier, and its energetic managing director Karina Buschsieweke, who is also now Appian’s director of product strategy.

During Appian World, Buschsieweke, a former Thai kick boxer who also helps run her family-founded online furniture brand, explains that, at Lana Labs, the aim was to automate some of the work that consultants do to try and streamline and analyse business processes to make firms more efficient and productive.

“They come to your organisation, they look at your workflows or do some interviews and recommend some solutions to help you with your business. We wanted to make this process digital with a tool that looks at how processes are running,” she explains.

She gives an example of ordering shoes online. “When this happens, in the background a process has kicked off: Your order is passed onto the warehouse. It’s distributed; it’s shipped, the payment needs to be taken. But maybe you end up with the wrong shoe. Maybe you don’t get the payment back.

She continues: “Sometimes when you have thousands of orders you don’t have numbers of how often it happens or why it happens. But process mining helps you spot patterns in the data that can tell you why certain process patterns happen.”

While Buschsieweke uses the example of retail, the tool can also be applied to insurance, to look at the way claims are processed, or in manufacturing where “saving seconds can save you millions,” she notes, adding that Mercedes-Benz used Lana’s product for this purpose.

Lana’s process mining product was able to analyse data, but not automate the process in near-real time. Following the acquisition and two-and-a-half years’ worth of development,  testing and validations, however, Appian fully integrated Lana Lab’s product into the Appian Platform and introduced Process HQ, a combination of process mining and enterprise AI unified with the Appian Data Fabric.

Unveiling the new feature ‘Process HQ’ as the latest integration into the vendor’s platform on the first day of Appian World, CEO Matt Calkins delivered the pitch, claiming that the feature was capable of enterprise-wide navigation, enabling users to measure any delay or momentary blip in a heads-up display. The tool also optimises, suggesting where improvements can be made to increase efficiency – this latter feature is what Lana Labs brings to the party.

The vendor also claims that its low code no code approach allows customers to make reports or answer questions quickly, without needing to rely on a data expert or developer to build a report.

AI-sprinkled procurement

With over 200 public sector clients worldwide, and many more from enterprise, Beckley believes that Appian “ brings commercial best practices to government and government best practices to the commercial world”.

The fact is, he adds, governments have so much data that in the world of AI sometimes they’re able to innovate in ways that the private sector can’t.

At its event this year, the vendor used AI on government data to launch a free tool aimed at simplifying the procurement process for US federal officials.

“It’s targeting the US government but it’s a sign of things to come and it’s really applicable to other domains in the future,” Beckley trails.

He adds that the product is targeted at the US government first, but the vendor hopes to roll it out globally eventually.

With no Appian licence needed, ProcureSight claims to use semantic search to help users track past procurement data; to gain insights from past procurements as well as accelerating procurement creation and documents.

“What we’ve done is used AI to search through the mass of government data on everything that’s bought – on the solicitations of what was wanted ; and then who won the contract, how was it purchased; was it a firm fixed price contact or a time and materials contact; what type of contract was it, and who won it,” Beckley explains.

“So now if I need to buy more pens, I can see who else has bought pens and who they used, what vendor they bought from and what they paid and whether they are happy with that delivery.

“That’s a trivial example – most of what the government buys is very complicated and so being able to follow in the footsteps of acquisitions that have gone before allows for a much better definition of requirements and better buying and better outcomes,” he adds.

Noting that the US government buys “$2 trillion worth of stuff” every year, he adds that even to do a little bit better “is going to save billions.”