How AI agents transform UCaaS workflows and ROI

AI is impacting enterprises in many different ways, and workplace collaboration is a fundamental use case. Unified communications as a service has become the hub for how work gets done, and IT leaders need to understand how integrating AI is driving change not only in worker productivity but also in automating workflows across the organization.

Let's explore how AI agents make workflows better and best practices for using AI agents with UCaaS.

How AI copilots and agents bring benefits to UC workflows

By now, enterprises have learned how AI copilots help workers automate tasks, but the emergence of agentic AI takes collaboration to another level. Both types are AI-driven and offer distinct benefits, and IT leaders need to support them both. Whereas copilots automate specific tasks as directed by workers, AI agents operate autonomously, with little or no human supervision.

This broadens the scope for how AI enhances workplace collaboration, going beyond task automation to executing workflows end-to-end. Workflows may incorporate discrete tasks handled by copilots, but they also orchestrate tasks across lines of business apps and workflows across the organization.

With the focus here being on AI agents, there are several benefits for IT leaders to consider when automating workflows. The following are some leading examples.

• Omnichannel communications support. By integrating with UCaaS, AI agents get a complete picture of the communications activity related to a workflow, such as voice, email, messaging, video and documents. With today's speech-to-text capabilities, AI agents can get accurate transcripts of related voice or video conversations.
• Cross-platform integrations. AI agents can integrate with other business-level platforms such as CRM, ERP and HR. This enables AI agents to automate workflows tied to specific departments or lines of business.
• Flexibility when things change. Simple workflows don't necessarily require AI for automation, especially those that are based on a fixed set of rules. The autonomous nature of agentic AI means handling complex workflows, where ambiguity and uncertainty are more likely to arise. Deterministic models are not built for this, and this is where agentic AI can take on a wider range of workflows.

High-value workflows for AI agents

All forms of AI need to show ROI at some point. With agentic AI, the stakes are higher because these agents are executing on workflows and making decisions along the way. Conversely, the negative effect can be significant if execution is poor or bad judgment leads to wrong decisions.

Deploying AI agents for workflows implies a level of trust that better results can be achieved than with humans. It's important to note that this is a higher level of trust than with AI copilots, which just automate specific tasks. Workflows are more complex, but when AI agents can perform as advertised, the effect is much greater -- and that's what drives ROI.

To that end, IT leaders need to identify workflows that provide high value in two ways. First would be the benefits of the workflow on the business, whether for a specific department or overall operations. Second would be the additive effect that AI brings with autonomous automation for that workflow.

Since AI works best with large datasets and discrete forms of data, the initial focus should be on workflows that play to those strengths. To help identify those workflows, here are some defining characteristics:

• Workflows that generate and/or consume large volumes of data. This includes tracking measurable activity over time or requiring computation across a wide range of variables.
• High degree of manual involvement. These kinds of workflows are prone to human error. With AI being able to detect anomalies faster than humans, AI agents will reduce these errors.
• Rules-based parameters. The more structured the workflow, the more effective AI agents will be. This is where initial deployments should focus, but note that AI can process unstructured, ambiguous inputs. IT can move on to those workflows as trust develops.
• Processes are repeatable. The more predictable, the better AI agents will perform. For recurring workflows, such as processing renewals, planning for monthly meetings or submitting quarterly tax payments, where the inputs are known and the steps well-defined. Autonomous automation will manage these workflows faster and more efficiently than humans.

Governance and guardrails

Governance is a must-have for any form of agentic AI that autonomously manages tasks or workflows. When AI bots are making independent decisions -- large or small -- there are direct factors to consider where AI must have a high degree of trust.

1. Compliance. Ensure workflows are compliant, especially for industry-specific requirements such as HIPAA, SOX and PCI, and supported by a clear audit trail.
2. Security. This is largely related to cybersecurity to protect workflows from malicious threats or autonomous actions that could trigger security breaches and expose data to bad actors.
3. Privacy. IT must protect employees, customers and partners when their data is connected to workflows that are being automated by AI agents

These are the basic guardrails for establishing trust with AI agents. A governance structure is needed for all AI use cases, not just for AI agents. This framework will constantly evolve, but requires a clear set of rules that define what AI agents can and cannot do when automating workflows. Not only that, but since each workflow is distinct, there should be policies specific to the workflow rather than being based on generic policies.

Central to all of this is human in the loop. AI may be advancing quickly, but it cannot be fully trusted to autonomously automate workflows. Unlike task automation, workflows have multiple points of failure, and humans need end-to-end visibility to detect potential AI mishaps, such as incorrect routing decisions, sharing unauthorized files, providing access to workers who don't have permissions for this workflow and missing steps in a predefined process.

Best practices for using AI agents

With AI agents still new, a universal set of best practices has not yet emerged, but a guiding principle is to start off with simple workflows that already have well-established steps and processes. That way, human in the loop will be more effective with AI agents to produce the expected outcomes, especially to support ROI. To help mitigate risk and build trust, here are two best practices to build around.

1. Establish success metrics

At a high level, three types of metrics will go a long way to defining success when using AI agents to automate workflows:

• Financial. Driving revenue, cost savings and improving margins.
• Productivity. Faster decisions, short timelines to complete projects and handling more inquiries.
• Quality. Fewer errors, improved accuracy, reduced reworking and less duplication of work.

To help develop these, team leaders responsible for the automated workflows should be involved. They have the hands-on familiarity to identify what those specific metrics should be and to map out the complete workflow, ensuring AI agents handle the entire workflow and not just part of it.

2. Integrating AI agents into workflows

This is another best practice that will help AI agents have maximum effect on workflows. Again, the focus would be on the workers and departments most connected to the workflow. Aside from having a working knowledge of the workflow, they can also play a direct role in getting those integrations just right.

The latter is now possible due to the prevalence of low-code/no-code programming for building AI applications. Technical expertise in AI is not needed, meaning these workers are in the best position to properly integrate AI agents with workflows. Otherwise, these teams would have to rely on integrations built by IT, system integrators or AI vendors, who would lack the localized expertise of the teams tied to the workflows.

By encouraging team members to be hands-on this way, not only will workflow automation results be better, but it will also build more buy-in from workers for using AI. Rather than viewing AI agents as a threat to their jobs, they will see them as tools to enhance their own performance and value to the business.

As a follow-on benefit, when workers successfully automate workflows on their own, IT's workload will decrease. Overall, this helps accelerate AI adoption across the enterprise.

Jon Arnold is principal of J Arnold & Associates, an independent analyst providing thought leadership and go-to-market counsel with a focus on the business-level effect of communications technology on digital transformation.