AI Chatbots for Business: The Essential Tool You (Don’t) Need

Recently, Beetroot hosted a webinar of the same name designed for business leaders and companies navigating the complexities of AI chatbots. Led by Beetroot’s Account Executive, Nikita (Nick) Tikhomirov, the discussion aimed to cut through the AI assistants hype, uncover their value and (un)importance for business growth, and explore the possibilities to drive tangible business results. 

Below, we collected the key insights shared by our guest experts: Cheryl Hodgkinson, the Chief Product Officer at Kauz.ai, Raphael Cohen, ex Product Director at Google and Founder & CEO at moojo.id, and Vitalii Huliai, Tech/Team Lead at Beetroot experienced in guiding cross-functional teams through complex AI implementation projects.

Why AI Agents Matter for Business: Real Value or Trend?

Many businesses experience a fear of missing out when it comes to adopting AI, especially AI agents. We asked our speakers to share their takes on when AI solutions genuinely add value vs. them being just a trendy addition.

Knowledge Sharing & Automation

Cheryl Hodgkinson, the CPO at Kauz.ai, believes that, as with any other application, it is important to have a clear understanding of what you are trying to achieve with AI and then evaluate whether it will do that for you. We see the most common examples of AI chatbots — or AI agents — in customer service, where they handle large volumes of inquiries and automate repetitive tasks. However, apart from potential cost savings on the service side, there are many other areas where AI applications can deliver value. 

Hodgkinson highlights another common use case — knowledge management — where AI makes the company’s knowledge truly accessible to all employees. Using an AI agent can also help shape a company’s image as tech-savvy and modern. It makes work easier for employees, improving satisfaction while also giving them a chance to develop AI skills that will stay valuable throughout their careers “because AI is not going anywhere anytime soon.” 

Supporting – Not Replacing – Human Roles

Raphael Cohen, the founder and CEO of Moojo.id, agrees that no matter if it’s AI, crypto, cloud, or mobile, the key is to start with the outcome: figure out what you want to achieve first, then decide how to get there. However, the increasing proliferation of AI and its use almost everywhere makes it harder than ever to stay true to this principle, and there are plenty of cases of it not being entirely beneficial. 

“In a business context, the agent will be as good as the data that it is fed with,” Cohen added, noting that many companies struggle to see a clear ROI from AI adoption. While businesses might often feel the pressure to adopt AI out of the fear of missing out, making it truly useful and achieving a positive return on AI chatbot investment depends on getting the data right, which requires serious effort.

Moojo’s CEO shared an interesting perspective: AI agents are most useful for tasks people don’t want to do — the want part is key. A simple example: while people may enjoy picking their own running shoes, they’d likely prefer to offload decisions like choosing a credit card or life insurance plan to an AI. The same goes for leadership roles. An AI agent could mimic a CEO, but when it comes to one-on-one conversations, employees don’t want to talk to a machine. Some things just can’t be replaced.

Addressing Real Pain Points

Vitalii Huliai, the Tech/Team Lead at Beetroot, emphasized that rather than being driven by trends, automating for the sake of automation, or keeping up with competitors, AI adoption should bring value to the company and address actual pain points in daily operations.

Before diving into it, it’s important to define the right use cases and set up a framework to evaluate AI systems. Also, the work doesn’t stop once the system is deployed — it’s an ongoing process. As the expert put it, “You need to constantly evaluate and track its performance, apply different technologies that appear each day to enhance and extend your system with new capabilities.”

What’s the Difference Between AI Agents vs. AI Assistants?

We asked our speakers to explain what makes an “AI chatbot” different from an “AI agent.” While the terms are often used interchangeably, this variation seems to matter to certain professionals.

Different Purposes

Raphael Cohen sees it as a matter of terminology, but ultimately, agents imply a greater sense of agency — the ability to make autonomous decisions, potentially using LLMs, orchestration mechanisms, APIs, or MCPs. AI chatbots and AI agents aren’t just different names for the same thing — they serve distinct purposes. 

Chatbots focus on conversational interactions, while AI agents go further, maintaining context across sessions, making independent decisions based on business objectives, and executing complex workflows. Unlike chatbots, which primarily respond to user queries, AI agents deeply integrate into business processes, adapting over time and working across multiple systems.

Part of the AI Spectrum

Cheryl Hodgkinson believes that “AI chatbot,” “AI assistant,” and “AI agent” often get used interchangeably, partly because early chatbots had a bad reputation for handling only simple tasks. From a marketing perspective, many companies now lean toward “AI agents” to distance themselves from those basic chatbot roots — even if what they’re offering isn’t truly fully autonomous. 

“In all fairness, though,” Hodgkinson added, “I do think that agentic AI is kind of a scale — it’s not that something is fully agentic and something is fully not.” Technology is moving fast toward more autonomous capabilities, and society is gradually warming up to the idea of agentic solutions, though not everyone is ready to hand over control entirely.

Chatbots as a Gateway to AI

As Vitalii Huliai sees it, the key difference lies in capabilities: AI systems can handle complex tasks, plan ahead, and process vast amounts of data, while traditional chatbots are mainly for simple Q&A interactions. People still associate chatbots with basic, sometimes frustrating experiences and limited responses.

In reality, chatbots are just a delivery medium for AI systems. They can operate through WhatsApp, Telegram, web pop-ups, or even emails, but the real power comes from AI-driven, hybrid human-AI systems that provide continuous, intelligent responses.

Scaling AI: How (and If) Chatbots Can Help Businesses Grow

Many companies successfully launch an AI chatbot but struggle to scale it. We asked our guests about the most common scalability challenges and how to solve them, and gathered their insights.

Quick Wins vs. Long-Tail Issues

Raphael Cohen highlights that AI agents can deliver quick wins but face significant hurdles when scaled across an organization. Referencing the 80/20 rule, he explained how companies can generate a lot of value fast, but fully perfecting the solution at scale can take much longer. 

Data is the biggest bottleneck: even though the “brain” behind AI keeps improving, making enterprise data digestible for an agent is a formidable task that needs a deep understanding of one’s business processes, workflows, customers, and users. 

The expert cautions that as you expand, people pay less attention to the early successes and focus on the tail-end issues. It’s easy to wow everyone with a demo, but the real work — and the real challenge — comes when you try to handle all those edge cases at scale.

Small Start, Big Data Hurdles

Cheryl Hodgkinson agrees that data is a huge hurdle to scaling AI. Many companies start small with a proof of concept (POC) developed by an experienced and motivated team using carefully chosen data, and it works fine initially. But once they try to expand, they encounter inaccessible or unsuitable data — sometimes stored on old drives or locked away in chats — and the challenge of integrating with multiple systems. 

Adding system administrators and data security further complicates things, especially when moving from a simple knowledge-management chatbot to an agent that actually performs tasks. Hodgkinson also points out that specialized AI solutions often don’t work across different departments, forcing businesses to implement numerous tools instead of one scalable platform.

Data, Costs, and Human in the Loop

Vitalii Huliai sees data management and real-time responsiveness as major hurdles for scaling AI. He emphasizes that simply feeding information to a model isn’t enough; context windows, near-real-time data processing, and system constraints all come into play. 

Huliai also points out the cost factor: it’s easy to build a POC on top of large off-the-shelf models (like OpenAI or Gemini), but at scale, those fees can add up quickly. While open-source options might seem free, they also bring their own infrastructure demands and require skilled teams to manage servers and workflows.

As the expert notes, AI systems involve more moving parts than traditional software. Even after launch, they need continuous performance tracking, user feedback, and refinement. He notes that stuffing too much information into an AI assistant can cause it to lose focus — or even get stuck in an endless loop. There need to be clear points of human intervention to avoid user frustration, allowing people to step in when the AI hits its limits.

Key Metrics to Measure AI Chatbot Success

Talking about the business angle beyond simple engagement metrics, our experts shared their recommendations on what other measures businesses should track to evaluate their AI chatbot’s impact. 

Cheryl Hodgkinson ties this back to an earlier point, “define your goal and then track whatever measures the success of that goal.” For example, if you want a customer service chatbot to cut costs, track engagement rates, the frequency of human escalation, and any related call center expenses.

If the focus is internal AI adoption, keep tabs on how often employees use it, for what tasks, and — most importantly — ask them why they might not be using it. Hodgkinson stresses that qualitative feedback is essential for uncovering hidden hurdles and guiding employees through the change process because not all challenges can be measured in numbers.

Raphael Cohen underscores that businesses shouldn’t invent new metrics just for AI. The same North Star metric that defines value for customers should remain central. Adopting AI agents doesn’t change what your business is trying to achieve, so metrics like engagement on their own aren’t enough. 

Instead, Cohen advises measuring the true impact on business processes, customer journey improvements, economic value creation, and organizational learning. An AI agent is just another tool — and it should be held to the same standards as any other solution that aims to deliver value to your customers.

AI Chatbots or Human Support: How to Decide?

We asked our speakers about finding the right balance between AI chatbots and human-led customer support — specifically, how companies decide which tasks to automate and which ones still need a human touch. After all, AI isn’t a silver bullet, and these tools are called “assistants” for a reason. 

Given the range of possible use cases and varying data maturity levels in different organizations, what do they recommend for businesses looking to draw that line?

Routine Task Automation

Vitalii Huliai believes the easiest place to start automating is with repetitive, routine tasks — especially those employees don’t enjoy. From his perspective, a hybrid approach works best, with AI and human expertise complementing each other. In practice, that means AI chatbots can greet customers, gather basic information, and provide context before handing over more complex or critical issues to human agents. 

People are better at handling tasks that involve high stakes, emotional intelligence, or domain-specific expertise. Beetroot’s Tech Lead also stresses the importance of having humans in the loop to guard against “hallucinations” (errors) that are an innate, inherited problem of large language models.

Journey Mapping & Escalation Paths

Building on the previous comment, Raphael Cohen advises starting from a practical perspective: mapping the entire customer journey to pinpoint friction points and then assessing each interaction’s complexity-to-volume ratio — or, more simply, figuring out where human emotional intelligence provides the most value, as opposed to high-volume, repetitive inquiries with predictable patterns. 

Cohen also emphasizes the need for clear escalation paths from AI to a human, particularly for interactions where the relationship itself is the goal. Tasks people don’t want to do — high-volume, low-value issues — should be replaced and scaled with agents. While finding the right escalation path can be tricky, he believes it’s entirely doable.

Personal Connection and AI’s Impartiality

Cheryl Hodgkinson offers another perspective involving company culture — specifically, deciding where a personal touch is essential in customer-facing interactions. She mentions complex tasks that AI can’t handle, situations with security concerns, or high-value products where it’s critical to build trust through a human advisor. 

At the same time, she notes that acceptance of AI chatbots is growing for simpler tasks, as they can be faster and more efficient. Interestingly, Cheryl points out that in certain cases — like when customers feel ashamed or worry a human might push a more expensive product — they might actually prefer talking to an AI, seeing it as a neutral party that won’t judge or up-sell them.

Biggest Mistakes in AI Chatbot Implementation

We asked our speakers about the most common missteps businesses make when launching AI chatbots — and how to steer clear of them. Here are their insights and some cautionary tales:

Vitalii Huliai recommends establishing clear goals and boundaries for each AI component. He warns against building a chatbot “just for the sake of chatting” and stresses the need for an easy escalation path so that human agents can step in to resolve the issue. Because AI is evolving so fast, Vitalii believes unpredictability is inevitable, so having humans on standby prevents minor issues from becoming major headaches.

Cheryl Hodgkinson argues that the biggest mistake would be not leveraging AI at all. However, she cautions against using AI without a clear purpose. Set concrete objectives and choose a platform accordingly, but remember that AI isn’t a “set-it-and-forget-it” system and requires ongoing attention. The expert suggests dedicating an internal team to continuous development; otherwise, the chatbot may quickly become obsolete and no longer useful.

Raphael Cohen noted that the most common error is assuming you made it once you’ve rolled out an impressive demo. In reality, you haven’t even started. Cohen stresses that proper planning and a long-term strategy are critical if you want lasting success rather than a short-lived buzz.

By integrating these perspectives, businesses can avoid common pitfalls: define clear objectives, ensure a smooth human handoff, plan for continuous updates, and never stop at the first “wow” moment.

Chatbot Readiness on a Scale from 1 to 10

We got a simple yet revealing question from the audience: “On a scale of 1 to 10, how ready are people to embrace AI chatbots?” Here’s how our speakers answered:

• Vitalii Huliai split his answer: 8 or 9 for advanced users who understand AI’s limits, but 5 or 6 for average users or businesses lacking AI expertise.
• Cheryl Hodgkinson generally agrees with 5 or 6, pointing to misinformation and limited public education as key barriers. Yet she notes rising acceptance — especially with the growing use of tools like ChatGPT.
• Raphael Cohen compared AI chatbots to traditional methods. He sees them as more inclusive and accessible, placing them around 8 or 9 from a user experience standpoint while acknowledging potential drawbacks such as misinformation.

Taken together, these viewpoints suggest that while power users and forward-thinking companies may be near the top of the scale, everyday users still hover around the midpoint. Yet all the speakers believe adoption will likely improve as chatbots become more capable and better integrated into existing workflows.

Wrapping Up the Conversation

Throughout this discussion, our experts emphasized one clear message: AI chatbots and agents hold exciting possibilities for businesses — whether it’s automating repetitive tasks, enhancing customer engagement, or scaling up more efficiently. Yet, realizing these benefits requires thoughtful planning, ongoing support, and a willingness to adapt as technology evolves.

If you’re looking for a partner to tackle the AI or ML side of your projects, we’re here to help. Our team specializes in end-to-end development solutions, from initial strategy to deployment and beyond.

Lastly, keep an eye out for our next events. We’ll be inviting more fantastic speakers and exploring new perspectives on AI, technology, and innovation. Stay tuned!