It's funny where a quick story can lead you.
At a round table conversation at our Forrester CX Forum in New York in June, I spoke with a woman who managed a contact center that handled healthcare insurance claims. She complained about one partner, a dental services organization (DSO) that manages claims for several dentists' offices. This organization built a bot that would call the contact center and enter 30 claims in a single call.
The woman complained that the poor customer service rep who got stuck with that call was exposed to a conversational but rigid and limited bot that sent handle times through the stratosphere and left her frazzled and exhausted. This was a new phenomenon, and the insurance claims company was struggling to find an appropriate response.
As I carried this story with me and shared it with others, I came to realize that this is an opportunity, not a problem. Consumers are already turning to their own artificial intelligence bots for product recommendations. A recent Forrester survey found that 35 percent of U.S. online adults use AI assistants to help them pick products. It won't take long for consumers to figure out that AI can help with customer service issues as well.
I had my own experience where ChatGPT was far more helpful getting my new cable box working than any support I received from my cable company.
As I further shared the story of the bot calling the contact center for batch insurance claims, other contact center managers told me about dealing with consumer-built bots. Some even mentioned dealing with bots calling in to handle B2B bulk transactions. Most are setting rules and procedures to block these interactions, but we can expect bot calls like this to create challenges for years to come.
Building a bot to call a contact center is a hack, a tactical trick to save time. Organizations are going to struggle to rein in consumer bots. At some point there will be productized consumer AI agent platforms that will normalize these interactions. This will require significant consumer adoption, which will take time, even in this moment of crazy AI adoption.
Things are a little different among B2B organizations, which arein a far better position to take advantage of these new capabilities quickly. We have organizations on both sides of the transaction that can benefit greatly from reliable, scalable AI agents to automate interactions, and have developers to build out solutions.
Enter interaction protocols.
It's probably not hard to see where I'm going with this. An AI agent isn't a human; it doesn't need to talk or even chat. Agents will communicate with each other directly. This is why we have the Model Context Protocol and Agent-to-Agent Protocol. In time, I'm sure we will see well-defined procedures and generalized approaches to this sort of agent-driven transaction.
Whenever two organizations waste a great deal of time and money on high-volume transactions over the phone, there is a great opportunity for automation that will benefit them both. These organizations don't need to wait for standards that will serve every type of interaction; they only need to solve for their specific problems, automating simple tasks like filing insurance claims. The protocol is there; the rules can be built specific to the two organizations' needs, and automation can happen quickly. No need for agents to try and talk or mimic human communications. Let agents be agents.
So, one lament from a frustrated customer service manager has led me to believe that we will see fully automated B2B service interactions starting to take off in 2027 while the rest of the world struggles with handling consumer bots.
Max Ball is a principal analyst at Forrester Research.