The rise of robotic process automation in the late 2010s ushered in a new era of knowledge work automation. RPA bots could emulate human tasks and complete simple workflows through scripted interactions with user interfaces or application programming interfaces.

Although RPA can save companies money by replacing repetitive human activities with less expensive bots, it has two major limitations: bots are vulnerable and increase the company’s technical debt.

In response, RPA vendors added AI to their bots, creating a new market for “cognitive RPA.” However, this was not enough to address these key challenges.

Gartner responded by introducing a new category of “hyperautomation,” shifting the focus of RPA to technology combinations that could address RPA’s limitations.

Then big language models came on the market, powering generative artificial intelligence and revolutionizing RPA, hyperautomation, and the rest of the knowledge work landscape.

It seems that every software vendor today must have a strategy for the AI ​​generation, and vendors in the automation segment are no exception. Today, however, vendors are moving beyond early conversational interfaces and turning to deeper applications of this transformative technology.

I spoke to eight of the most innovative vendors in this market. While they are only a fraction of the incredible number of disruptive forces in the market today, they are representative of the progress vendors are making in tackling the difficult problems of knowledge work automation. (*Disclosure below.)

Here’s what I learned:

From bots to AI agents

The term “agent” refers to autonomous software that achieves certain business goals independently of other software in its environment. However, how autonomous they are and what they actually do depends on who you ask.

The agents of Agents.inc (AGENTS HQ GmbH), for example, connects data sources, interacts with various AI models via prompts, and can generate reports, alerts, and interactive dashboards. Agents.inc is currently focused on market intelligence use cases, but can do much more.

Tines Tines Security Services Ltd., on the other hand, focuses primarily on cybersecurity tasks. Tines makes creating automations so easy that security analysts have the ability to create automations proactively or, when the situation requires it, on-site to solve a new problem.

For his part, Leena.ai Inc. focuses on service management workflows with the goal of eliminating the need for support tickets.

This selection of use cases represents an opportunity that can be easily achieved rather than technical limitations of the respective platforms. As technologies advance, you can expect to see more general AI agents.

Addressing the problem of data integration

One of the reasons RPA bots can be so vulnerable is that they are sensitive to changes in the meaning, format, and availability of data in different domains. Any change to the data in an underlying system can break the bots.

Artificial intelligence (AI) can be particularly useful in addressing such data integration challenges. Bardeen Inc.For example, uses millions of workflows as training data so that its AI can automatically figure out which fields a knowledge worker needs to map and what data formats they are in. As with other AI applications, this ability improves over time.

Agents.inc uses agents to prepare the data and then combines AI with knowledge graphs before feeding it into LLMs. This solves semantic problems and eliminates most “hallucinations” – where first-generation AI makes incorrect guesses in the absence of relevant data.

The basic idea is that most data integrations follow familiar patterns. For example, if someone asks an agent to pull data from Salesforce and enter it into HubSpot, the agent can do a good job of guessing the specifics of each field based on large amounts of historical data.

Creating workflows with AI

Existing AI-powered automation offerings typically provide “next best action” recommendations – essentially an auto-complete for creating workflows, where the AI ​​guesses the next step in an ongoing workflow based on its previous experience with workflows.

Today’s automation solutions also offer this option, but go one step further. For example: Boomi LP provides an AI-based conversation design tool for creating automations and integrations, leveraging training data with more than 300 million examples.

Bardeen provides several pre-built automations and allows users to create new automations through a conversational interface. To reduce the likelihood of hallucinations, the platform converts the workflow’s conversational specification into a domain-specific language representation.

This approach transforms the artificial intelligence-based understanding of intent into a deterministic workflow for improved repeatability, with an additional layer of verification further reducing the risk of hallucinations.

More common, however, are low-code tools for creating workflows that support AI. With Tines, for example, analysts create automations on a visual storyboard by connecting seven basic actions. Analysts can copy and paste third-party APIs into Tines to interact with any third-party tool or app.

SmythOS (INK Content Inc.) also offers a low-code platform for both building AI agents and creating workflows using its application orchestration framework.

SmythOS allows users to create their own agents and workflows, with each agent implementing an LLM-based task. The platform allows users to integrate the actions of different LLMs into customized, often complex workflows.

Sema4.ai Inc.However, it takes a different approach and focuses on document-centric rather than workflow-centric automation.

In Sema4.ai, organizations use an AI assistant to create runbooks with natural language descriptions of the desired behavior. These runbooks, in turn, serve as prompts for the AI ​​to create the agents.

Runbooks describe the specific business requirements for a document-centric task, such as invoice reconciliation. The semantic document intelligence within the platform then leverages business rules encapsulated in documents for validation.

Solving the technical debt problem

RPA increases companies’ technical debt because it does not solve the problem of inflexible and outdated business logic and its representation in code. Instead, RPA adds an additional layer of unstable software that companies now have to maintain – increasing the company’s overall technical debt.

Most of the vendors in this article are leveraging new-generation AI to reduce the maintenance burden of their respective AI agents. Updated LLMs and constantly improving data sets result in agents that automatically get better at what they do – rather than requiring the expensive, largely manual maintenance that burdens RPA bots.

However, dealing with the legacy code problem is another challenge. When there is no API and a custom connector is not cost-effective, most automation vendors default to an RPA bot as the only remaining alternative.

The only exception among the providers I interviewed is Beezlabs (Beez Innovation Labs Pvt. Ltd.).

About Advanced Business Application Programming or ABAP, SAP SEUsing Beezlabs’ proprietary programming language, Beezlabs can interact directly with SAP without depending on the user interface or having to customize the SAP application code itself. In fact, the company is entirely SAP-specific, leveraging its extensive ABAP expertise.

Beezlabs also uses the open source business process orchestrator from Camunda (Camunda Services GmbH) to connect SAP tasks with other platforms, such as Excel and Salesforce. With Beezlabs, companies can run ABAP bots in parallel on the provider’s cloud-native Kubernetes infrastructure. These ABAP bots interact with SAP at an internal coding level rather than through the user interface like RPA bots.

For SAP customers who prefer a “clean core” strategy for SAP (and avoid hard-to-maintain customizations), Beezlabs’ offering provides the best of both worlds: AI agents that run natively on SAP, without the need for customization from SAP itself.

The Intellyx version

AI-based automation technologies like those described in this article are disrupting many markets. By improving resilience and reducing technical debt, AI agents are quickly displacing RPA bots—but that’s not all.

Previous generations of low-code/no-code automation platform vendors used visual metaphors (also known as “boxes and lines”) to provide workflow creation capabilities. Today, natural language interfaces are becoming the norm and are transforming the entire low-code/no-code market.

AI agents also play an important role in enterprises’ cloud-native strategies because each agent can run statelessly in containers, allowing each platform to automatically scale them and deploy as many identical agents as needed for each situation.

The platform, in turn, manages the status of each workflow while providing capabilities for creating and managing workflows based on artificial intelligence.

Hyperautomation is also over. Instead of requiring a collection of different tools, AI agent technology represents a convergence.

New generation AI-based technologies are gradually replacing not only RPA, but also business process automation, low-code/no-code platforms, rule engines, data integration technologies and more.

Stay tuned, as there will be many more disruptions to come as this convergence trend continues.

Jason Bloomberg is founder and president of Intellyx, who advises business leaders and technology providers on their digital transformation strategies. He wrote this article for SiliconANGLE. (*Disclosure: SAP and Tines are customers of Intellyx. Camunda and Sema4.ai are former customers of Intellyx.)

Image: SiliconANGLE/Ideogram