Blog

Contents

    Jun 13

    Deterministic Vs. Agentic Lead Qualification – CloseBot Has the Best of Both

    Agents are all the rage right now, and rightly so… AI agents are cool because they are super simple to use. Give an agent a prompt and some tools and off it goes. 🤯 Compare this to workflows where you have to go through more work to set up the flow that guides the AI working on the flow. Does the emergence of agents mean workflows are dead?

    Agentic Systems

    Pure agentic systems take in your prompt and your tools you give it and it’s free to respond and use tools as it desires. Oftentimes, these agents are often built on top of reasoning models like OpenAI’s o3. These models have a thought process to “reason” through problems, using tools at will.

    The Problem with Agents

    We attended Web Summit Vancouver a couple weeks ago and a quote from the creator of RAG, Douwe Keila stuck out. He said plainly during the presentation:

    “Anyone selling you on pure agents working reliably in production is lying to you.”

    This is pretty bold of Douwe to say despite all the headlines telling us otherwise. So why is this?

    Pure Agents

    On the surface, agentic systems seem great. Just write down your job that you want the agent to perform and connect it to the needed systems and it’s supposed to just work. In reality, AI is easily led astray. For example, AI tends to suffer from being overly agreeable (sycophancy). You may accidentally include bias into your prompt that leads the AI consistently in a direction that you don’t want it to go in.

    Pure agents are great for use in instances where the complexity of the situation is high and the consequence for error is low. Think internet search, for example. If you use an agent to assist with search and summary of online resources (a complex task), the consequence if it doesn’t retrieve the right information is low.

    Alternatively, if you use pure agents in something that’s low complexity and has high consequences, like lead qualification chats, you’re in for a bad time. In production, agents have a difficult time staying on task when qualifying leads. They will falsely confirm appointments often, allow leads to jump forward past qualifying questions, and may even recommend a competitor. 😬

    The True Cost of Agents in Production

    Platforms like GoHighLevel’s AI Employee that allow you to build agents with prompts alone may seem nice, but you’ll think otherwise once you actually put these agents into production to qualify leads. Your clients will likely flood to support as they watch in horror as the agent loses leads for them… leads that cost money. Many CloseBot users were hesitant to adopt AI again because they came from pure agent systems that were difficult to control and troubleshoot.

    Deterministic Systems

    The best system is a system that never messes up. A system where you always get the same output (response) given the same input. These are called deterministic systems. An extreme example would be a chat interface where the customer pushes pre-defined buttons that lead them through a conversation flow.

    These deterministic systems work great for things like support, where the personal touch doesn’t matter as much. If someone wants to get help, they will click the right buttons to be directed to the right resource. Deterministic systems don’t work as well for things like lead qualification, where contacts desire a more personal touch.

    So how do you win?

    Winning with CloseBot

    CloseBot has developed a state of the art job builder for agentic lead qualification flows. Agentic flows… the power of agents paired with the deterministic guardrails of workflows (we call it our Job Flow builder). CloseBot was the first to coin the term objective, but has since expanded to add many other types of nodes.

    CloseBot agents are guided through job flows, but the agent is free to:

    1. Do multiple things at once with a multi-objective as shown HERE
    2. Jump to other parts of the job with scenarios as shown HERE

    Summary

    On top of giving marketing agencies an easy way to re-bill agents to clients, CloseBot also operates as a client dashboard for agent KPI’s, makes it super easy to connect to supported CRMs and much more. See HERE how CloseBot can be connected to GoHighLevel in 48 seconds.

    You could spend hours setting up custom systems for each client with no dashboards, re-billing or support… or you could get started with CloseBot today and spend less time building and more time increasing your agency’s revenue.

    What you should do now

    1. See CloseBot’s powerful AI agents in action.
      Sign up for a free trial.
    2. Read more articles from our blog.
    3. If you know someone who would enjoy this article, share it with them via email, LinkedIn, Twitter/X, or Facebook.