In a recent Gartner study of 6,000 consumers, 64 percent preferred companies not to use AI, and 53 percent disliked AI so much they’d consider switching to a competitor.
Additionally, 90 percent of people in a SurveyMonkey poll said they’d rather deal with a human than a chatbot.
These are alarming stats. But here’s the good news: the issue isn’t the technology itself; it’s how it’s being deployed.
Recognizing this, Matt Dickson, COO of Eclipse Telecom, joins host Justin Robbins for the second of a six-episode CX Today miniseries: Contact Center Talk.
In the following video, they discuss Dickson’s the deadly sins of AI and automation, serving up seven best practices for implementing contact center AI in 2025.
A written rundown of each best practice is also available below.
1. Consider Where Automation Fits Within Your Brand
Think about how much effort goes into training contact center agents on your brand voice. Is it playful, authoritative, empathetic?
Yet, when it comes to chatbots, many organizations deploy them straight out of the box, making them sound like every other brand using that platform.
To fix this, consider tools that allow the CX leaders to personalize tone and tenor through techniques like prompt engineering.
Also, bring more people into the fold, particularly the marketing team, who are the business’s brand voice guardians.
Just remember, collaboration is key. Marketing and the contact center share the same goal: delivering exceptional customer experiences.
So, don’t be afraid of inviting them into the process early. It’s not about turf wars; it’s about leveraging expertise to create consistent, impactful customer interactions.
2. Understand Your “Exit Rows”
In 2024, the White House raised the issue of customers getting stuck in “doom loops”, where experiences are designed to frustrate people into giving up.
Examples include overly complex processes for canceling services, with SiriusXM getting hit by a lawsuit for not allowing customers to delete an account simply.
Yet, ultimately, almost everyone has yelled “agent” into the phone or pounded “zero” on the keypad. Customers need an easy way to exit “AI jail.”
Contact centers must clearly label exit options in their automated systems to avoid this.
Moreover, consider using sentiment analysis to identify when customers get frustrated and proactively offer human assistance.
Other top tips include:
- Starting conversation automation initiatives with high-volume, low-complexity use cases that are predictable and less prone to emotional volatility.
- Maintaining context when transitioning customers from AI to a human agent. No one wants to repeat themselves. It’s one of the quickest ways to tank satisfaction.
- Monitoring long-running interactions and proactively escalating before frustration peaks.
Lastly, remember that not all AI is generative AI. But that’s where many organizations stumble. Generative AI has its place but is prone to issues like hallucinations, where it “makes stuff up.”
Indeed, Stanford research shows that the longer a generative AI interacts, the higher the likelihood of errors. That’s why – to the last bullet point – monitoring back-and-forth exchanges is critical.
If a conversation drags on, it’s time to get a human involved.
3. Locate Your “Circuit Breakers”
Not every use case is a good fit for AI. Sometimes, a business can prevent the issue before it occurs and remove the need for customer contact altogether.
Moreover, even after building an automated journey for a specific use case, there will be times when the customer shouldn’t use it.
For example, consider a customer who has endured ongoing service issues with their internet provider. The provider sent out technicians, who consistently found that the problem stemmed from outside the house.
However, every time the customer called for support, they entered an AI-led flow that forced them to go through the same troubleshooting steps despite the technicians knowing the issue wasn’t with their equipment.
That example spotlights a missed opportunity for a “circuit breaker” approach. After multiple visits and a known external issue, the provider could have bypassed the initial troubleshooting steps and escalated the case directly to Level 2 or Level 3 support. Instead, the customer was subjected to unnecessary delays and frustration.
The lesson is to evaluate use case fit and determine when to bypass automated processes.
Generative AI and automation tools aren’t the answer to everything. Sometimes, based on a customer’s history, the contact center should route them directly to a human for better service. Having a holistic view of the customer journey is crucial.
4. Think About Agent Enablement Before Automation
The best way to improve customer experiences is to enable better agent interactions. Indeed, a Forrester study highlighted three key ways to enhance revenue through better customer service:
- Answering All Customer Queries – Customers won’t buy if you can’t answer their questions.
- First Contact Resolution (FCR) – Resolving issues the first time is critical.
- Empowered Agents – Agents who can resolve issues without checking with supervisors or following rigid scripts.
So, rather than just prioritizing automation for cost savings, focus on tools that empower your agents to perform at their best.
Consider solutions like agent-assist and copilots to make it easier for reps to access information, receive real-time coaching, and solve problems efficiently.
Investing in these tools also uncovers insights that can improve contact center operations.
For example, by analyzing unstructured contact center data with generative AI, companies can optimize FAQs, update product manuals, or even refine product labels to reduce customer confusion.
That dual benefit of improving agent efficiency and gaining business insights is where AI shines.
5. Make Your Mistakes Where Nobody Sees Them
Too many companies start implementing AI from the top down, deciding on high-level use cases and automating processes prematurely. A better approach is to start at the ground level by deploying AI tools to agents.
Agents provide valuable feedback on whether the tools work. They’ll also pinpoint when AI performs well and when it fails.
By iterating and refining in this “sandbox,” the contact center can perfect its solutions before rolling them out directly to customers.
This bottoms-up approach avoids the pitfalls of rolling out unproven AI to customers and ensures customer service is automating processes that are ready.
Moreover, the strategy will stop organizations from over-engineering solutions without fully understanding the problem.
By creating safe testing environments and engaging the workforce in iterative experiments, contact centers can uncover the best use cases for AI.
Just remember: start small, learn quickly, and scale intelligently.
6. Take a Broad View of Contact Center AI Use Cases
Many contact center leaders focus narrowly on automation and generative AI, missing out on other impactful applications. Consider the following two underrepresented use cases.
The first is automated quality assurance (QA). Historically, contact centers review only one or two percent of calls, which is not a statistically valid sample.
However, AI can review every interaction – digital or human – for compliance, accuracy, and performance. This capability is invaluable in highly regulated industries like healthcare and finance, where adhering to protocols is non-negotiable.
The second is mining unstructured data from the contact center. By analyzing why customers make contact and identifying points of frustration in their journey, AI provides insights far beyond what call deflection through automation can achieve.
Understanding intent data holistically enables contact center teams to uncover opportunities for significant improvement.
Yet, ultimately, before any AI project, leaders must first clarify what’s most important to achieve. It’s easy to be reactive to urgent issues, but focusing on long-term priorities drives sustainable progress.
7. Continue Monitoring & Enhancing Your AI
Plan for the ongoing care and enhancement of AI systems. Why? Because, unlike traditional technologies, AI tools require continuous evaluation and improvement.
As such, businesses must assign resources – whether a new role or upskilled supervisors – to regularly monitor and refine these tools.
Neglecting this step typically leads to frustration or abandonment of the technology.
Recognizing this, organizations should ask:
- Who is responsible for ensuring the AI performs optimally?
- What training do they need?
Having a plan from day one to dedicate time and resources to AI maintenance is crucial. After all, AI is not a “set it and forget it” solution.
AI should help organizations become more predictive and proactive in the long term, but that requires ongoing involvement.
Don’t miss an episode of our Contact Center Talk miniseries. Sign up for the CX Today Newsletter.
