It’s estimated that there are around 15 million customer service representatives (CSRs) in the world. Many of them spend their days doing things that humans are overqualified to do: answering simple product questions, scheduling appointments, or confirming delivery date. Their managers — often cautious, operations-minded contact center leaders — find themselves on the tip of the AI spear with one of the best use cases for modern AI capabilities such as generative AI, LLMs, and agentic frameworks.

In The Forrester Wave™: Conversational AI Platforms For Customer Service, Q2 2026, I got to see many cool product capabilities, but the key lessons from this Wave came from my conversations with 27 practitioners who delivered self-service applications using these tools.

Managing The Change To AI-Driven Self-Service

Traditional customer self-service applications are well-known and often loathed by consumers everywhere. By contrast, modern self-service applications please customers with their conversational skills and are taking on ever more complex tasks. It’s not easy to get the technology properly in place, but the bigger challenge is the work involved getting these applications into production. Driving organizational alignment, managing a project with few examples to learn from, designing open-ended customer conversations, and reassuring CSRs who feel their jobs are at risk are some of the many challenges keeping most contact centers on the sideline for this transition to date.

Here are three key things that these 27 practitioners had in common. Specifically, they:

  • Built a “coalition of the willing” from day zero. These leaders connected with key stakeholders, such as the legal and security departments, from the very start. Partnering with these stakeholders requires bringing a shared challenge, taking input, and communicating consistently throughout the deployment process.
  • Followed a crawl-walk-run approach to deployment. I have spoken to some organizations that successfully deployed a “big bang” initial deployment. But the safest approach is to start small and grow. Successful practitioners started with small deployments of limited scope and continued to add capabilities and size gradually, learning at each step. There isn’t a lot of room for a “fail fast” mentality when your applications are customer facing, but the tools allow you to launch to 1% of users and only grow as success is proven, which allows you to try things and minimize any negative impact.
  • Faced significant change management challenges. Reference customers in this Forrester Wave repeatedly said that they wished they had spent more time on change management, specifically on helping their CSRs to understand and accept their new normal. While much of what CSRs do is work that they are overqualified for, they will feel threatened when they see automated bots taking on more of the customer interaction load. What made a difference for some of these organizations: bringing CSRs into the conversation early, getting their input on handoffs from bots, and making changes based on their feedback.

Building The Change To Manage

My previous conversational AI Wave was published in the first half of 2024, a little over a year after OpenAI’s announcement of ChatGPT. In an amazing group sprint, all the vendors in that Wave had refactored their platforms around generative AI, and the result was a new “art of the possible” for customer self-service. In 2026, the leap is to agentic frameworks. These new platforms can reason, plan, take actions, use tools, and coordinate workflows to achieve a goal with a degree of autonomy.

For the foreseeable future, no contact center will give full autonomy to AI agents that are interacting with customers. To meet the needs of their customers, the vendors implement guardrails and controls that restrain the AI agents from full autonomy and provide the accountability that customer service teams require. The result is a platform that grows with organizations as they mature and take on more automation but stays in check and under control.

There are many new capabilities that conversational AI vendors offer for customer self-service today, including:

  • Full multilingual capabilities. Some of the reference customers in the Wave interact with customers in 15 or even 30 different languages, all running the same core application.
  • AI to identify trending use cases that will be valuable to automate. Most of the Wave vendors have tools that identify specific conversations that represent a large portion of customer interactions. In some cases, the system will estimate the percent of the brand’s overall interactions that would be covered by that use case.
  • Automated application development. Nonprogrammers can build applications that the system identifies, like order status for certain order types or installation difficulties with certain products. All that is required is a document that describes the interaction, including required data sources, application flows, and customer conversations. Conversational AI systems take documents like this and turn them directly into applications; no coding is required.
  • New testing capabilities. That application you just vibe-coded will not be perfect, particularly if it has to integrate with backend legacy systems, but the vendors provide bots that automate the testing process to identify and iron out issues. Conversational AI platforms automate testing with bots that are instructed to take on synthetic personas spanning a range of customer needs, emotions, and expertise levels. Beyond simply executing the application flow, these more specific personalities show that the application can properly serve a brand’s actual customers. Testing can be done on a small scale during development and ramped up to high volume during regression testing.
  • Multimodal capabilities. Multimodal enables a single conversation between a bot and a consumer to happen across multiple channels at once. Imagine being able to talk to a customer through a troubleshooting exercise while they share a video of the configuration screen of their system. Or imagine allowing customers to select a shirt from a carousel in WhatsApp while confirming verbally that this is the one that they want.

For the full insights, please see the research, Lessons Learned From The Forrester Wave™: Conversational AI Platforms For Customer Service, Q2 2026. To discuss your questions about conversational AI in your organization, please schedule a guidance session with me to dig in on these topics.

Share