All About Process Guides: Your Agentic AI Agent GPS

Take a look at the following example of a conversation with a typical AI agent. Does it feel familiar?

Rather than ask clarifying questions to help solve the customer’s issue, the chatbot immediately responds with a help desk article on how to fix an item that may or may not match the device in question. As soon as it hears the article isn’t helpful, it routes the customer to a live agent, and the conversation must start all over again.

Frustrating, right? I couldn’t agree more. 

Enter Process Guides: One of the key tools we use at Quiq to power truly agentic AI solutions that replicate experiences customers have with top-tier human agents. Our AI agents are capable of understanding conversational context, switching topics, and reasoning across company policies and procedures to deliver exceptional service that builds satisfaction, trust, and lifetime value.

Never heard of a Process Guide? No sweat, because by the end of this post, you’ll practically be an expert. Let’s go.

Pre-defined Conversation Flows vs. Process Guides

Before we get into what Process Guides are and how they work, it’s helpful to first understand how previous-generation chatbots like the one in the example above work. These chatbots are rules-based and use traditional Natural Language Processing (NLP) to try to match users’ questions to specific, pre-defined intents and responses. 

In other words, companies must anticipate the types of questions users will ask and the phrases they will use to build out rigid if/then conversation flows for the chatbot to follow. This obviously creates a world of problems, including the inability to focus on more than one question or request at a time, or switch topics without forcing the customer to repeat themselves. 

A chatbot’s assistance is limited to the subjects and phrasing it’s been explicitly designed to understand and process. I like to compare this to a train on a track. There is one path to the chatbot’s destination, and it has to stay on the rails to get there. If something unexpected blocks the track, it either goes off the rails (provide a nonsensical or insufficient answer) or has to cut the journey short (escalate to a human).

Some AI for CX vendors claim their chatbots use the most advanced GenAI. However, they are really using only a fraction of an LLM’s power to generate a single response from a knowledge base, rather than reasoning and helping to disambiguate and solve complex issues. Because these simple GenAI bots are designed for single turn questions and answers, they still struggle to actually help users solve real problems, as you can see in the following example:

Rather than simply sending a link to the wrong article like the first-generation bot does in the example at the beginning of our post, this simple GenAI chatbot generates an answer. However, it does so using information from the same incorrect article. Think of it as the same train but with a smarter conductor: better at retrieving and communicating knowledge, but still stuck on the same rails.

In contrast, Process Guides are designed to take advantage of the full reasoning power of the latest large language models (LLMs), enabling truly agentic or autonomous decisions and actions within clearly defined boundaries. Rather than giving the AI agent a specific set of step-by-step rules it must follow, Process Guides provide instructions, best practices, and tools AI agents can reference to achieve certain goals.

Compare this to a car (self-driving, of course) following a GPS. You give a GPS your final destination, but you can also tell it you want to avoid tolls or highways, and it can re-route you in real-time to avoid road closures and traffic along the way. And if you enter a new destination, it will simply re-direct you from your current location without making you “backtrack.”

Similarly, while agentic AI agents are objective-driven, they also comprehend and consider the full conversational context around these goals. This allows them to answer multi-part questions and seamlessly switch topics without losing sight of their “final destination,” or change goals altogether without forcing customers to “backtrack” and repeat themselves.

This approach is especially critical for enterprises, like Roku and Spirit Airlines, for example,  with nuanced policies and procedures, where incorrect information or negative experiences can have high stakes consequences. These workflows are often complex and require rigor and structure in addition to reasoning, problem solving, and sensitivity. Process Guides help provide this essential framework, enabling companies to control AI without dictating it, giving it freedom to provide exceptional service within the necessary constraints.

Let’s Take a Closer Look… 

A Process Guide is a goal-focused document that defines what to collect, which tools to use, and which actions to take. It gives an AI agent clear steps and decision rules to follow during a customer conversation, including which data sources to consult and when to execute specific tasks. These could be anything from a knowledge base to a product catalog to the account details needed to answer a given question. 

Process Guides vary based on what an organization does and the kinds of inquiries it wants its agentic AI agent to handle. However, a few common types of guides are:

• Agent Escalation
• General Support
• Device Trouble Shooting
• Booking or Scheduling
• Reservation Modification
• Account Management
• Order Management
• Rebooking or Rescheduling
• Return or Exchange
• Subscription Modification
• Product Recommendation
• Password Reset or Recovery

Let’s take a look at an example of a real Process Guide for Device Troubleshooting:

As you can see, the Process Guide begins with a clear objective that tells the AI agent when to use this guide. It then provides a list of steps and relevant supporting details for the AI agent to reference. These steps and details also vary by company, which means no two Process Guides are exactly the same, even if they share an objective.

Notice that while the Process Guide may tell the AI agent that specific information (like a user’s email, for example) is required to achieve the objective, it does not specify exactly when to collect it. Similarly, it does not dictate exactly how the AI agent should respond outside of adhering to specific company communication policies, such as keeping responses under a certain length. (However, note that when exact legal or compliance-grade wording is required, Process Guides can direct AI agents to select and send pre-approved messages from a catalog.)

Instead, it offers examples, resources, considerations, and general instructions the AI agent can choose to leverage or follow given the context of the conversation and the knowledge or information it has already obtained. This is essentially the same way companies educate and enable human agents. In fact, we usually collaborate with our clients to adapt their existing human agent training manuals into Process Guides.

Hopefully you have a better understanding of what a Process Guide is than you did at the top of the post. While Process Guides are extremely important, it’s the way they’re leveraged in our AI Studio platform that makes them special and sets our agentic AI agents (and human-facing AI assistants) apart…

Quiq’s AI Studio and Cognitive Architecture

Every one of Quiq’s Quiq agentic AI agents are built and managed using Quiq’s AI Studio, which gives companies the best of both a “buy” and “build” approach to agentic AI lifecycle management. Quiq handles the maintenance, scalability, and ecosystem required to power an agentic AI agent, while teams can choose the level of observability, flexibility, and control they want.

It also strikes the perfect balance between AI capabilities and business logic, enabling enterprises to create safe, consistent, and repeatable AI experiences that harness the full power of AI while keeping humans in the driver’s seat. This is accomplished largely in AI Studio’s cognitive architecture, which enables the AI agents to make sense of information, assess context, reason, and respond and act thoughtfully — very similar to how humans think intelligently. 

Below is a high-level diagram of what Quiq’s cognitive architecture looks like. Every user input goes through this entire architecture before the AI agent responds:

As you can see, before the AI agent even gets to the “Guide Consultation” step, it goes through a “Planning” stage that takes not only the user’s latest input, but also the context of the entire conversation into account, to determine the appropriate Process Guide to invoke. This is what allows the AI agent to seamlessly switch between objectives or Guides without losing context, while orchestration steps like post-answer “Guardrails and Verification” ensure accurate, on-brand responses and effective guide execution.

You’ve Arrived at Your Destination

Now that you have a deeper understanding of Process Guides and how they’re used in Quiq’s AI Studio, let’s revisit the example conversation from the beginning of this post. The following image shows how the conversation would progress with one of Quiq’s agentic AI agents, including how it identifies and references our sample Device Troubleshooting guide behind the scenes:

Better, right? I couldn’t agree more.
If you’re interested in learning more about Quiq’s Process Guides and AI Studio, feel free to reach out to me on LinkedIn. Or, book a demo to see Quiq and our agentic AI agents in action.

Frequently Asked Questions (FAQs)