Ada’s AI agents work in parallel to deliver better results in customer service
AI agents are changing the way companies handle customer service, and Ada just took a major step in that direction.
In February of this year, the platform launched its unified Reasoning Engine (RE), a solution designed to solve a problem many companies face in practice: managing different AI agents spread across separate channels like chat, email, and voice without losing control over how each one behaves.
The idea is simple but powerful. Instead of configuring an agent for each channel separately, Ada lets a single instruction propagate to all of them at once. That means less rework, more consistency, and a much more cohesive customer experience regardless of where the conversation happens.
According to Mike Murchison, CEO and co-founder of Ada, the logic is straightforward: when a company instructs an agent to be more empathetic in email, Ada can propagate that same set of instructions to how the agent communicates over the phone. What used to require replication across five different places now becomes a single effort.
But what really stands out about this architecture is what happens behind the scenes: multiple agents working in parallel, like a well-synchronized team, while the end user sees just one conversation flowing naturally. 🚀
How Ada’s Reasoning Engine works in practice
Ada’s Reasoning Engine isn’t just another layer of natural language processing. It was designed to function as the central brain of the entire support operation, coordinating the behavior of each AI agent based on a unified set of rules and objectives.
In practice, the architecture involves two distinct language models working together. One is the model internally called the talker, which is extremely fast and specialized in conversational dialogue. The other is the thinker model, responsible for deep reasoning and executing complex tasks that involve multiple steps. These two systems communicate seamlessly, without the end user ever noticing the complexity behind each response.
Murchison explained that a company doesn’t need to think about all the different layers of agentic interaction living inside Ada. It simply manages a single agent, as if it were one team member. Similarly, the end customer sees just one unified experience — a single agent they’re interacting with.
When a company updates a refund policy, for example, that change is automatically reflected across all channels at the same time, without the need to edit each agent individually. This kind of synchronization eliminates one of the biggest headaches for anyone working with automation at scale: inconsistency between channels, which ends up creating confusion for customers and rework for internal teams.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
Under the hood, the engine operates with a multi-agent parallel architecture, where each agent specializes in a specific task or channel, but all share the same context and behavioral guidelines. This is quite different from the traditional model, where each bot was treated as an isolated island with no communication with the others. Ada solves this problem by creating an orchestration layer that distributes reasoning intelligently, ensuring the right response reaches the right channel at the right time, with no contradictions or information gaps.
A practical example cited by Murchison illustrates the system well: imagine a customer’s flight gets canceled and they call to rebook. One Ada AI agent maintains the voice conversation, confirming information and keeping the customer updated, while other agents work in the background searching for new flights, handling the rebooking, and sending the confirmation. All of this happens simultaneously, much the same way an experienced human agent would handle the situation.
Another interesting point is that this engine was built with a focus on explainability. Product and support teams can clearly understand why an agent made a particular decision during a conversation. This is critical for companies operating in regulated industries like finance and healthcare, where every customer interaction can have significant legal and operational implications. The ability to audit agent reasoning puts Ada in a differentiated position within the customer service automation market.
Agents working in parallel: why this changes the game
The ability to have AI agents working simultaneously on different fronts of the same conversation is what sets Ada’s approach apart from many competing solutions. While most automated service platforms process one step at a time in sequence, Ada’s agents can split tasks and execute them at the same time.
Murchison pointed out that the scope and complexity of the tasks Ada’s clients are automating are growing significantly. That’s why it’s so important to keep engaging customers, keeping them informed, and resolving other issues while work is being done in the background.
And the ambition isn’t small. Murchison mentioned that later this year, Ada will likely automate the first task equivalent to roughly 14 hours of human work. That kind of milestone shows just how much agentic AI is evolving in terms of autonomous execution capability.
This parallel dynamic is very similar to what happens in a well-run traditional contact center, where a human agent can do multiple things at once: listen to the customer, check internal systems, review policies, and already start preparing the solution. The difference is that Ada’s AI agents can do all of this at scale, serving hundreds or thousands of customers simultaneously without any loss in quality or speed. 🔄
The role of Playbooks in customer service automation
If the Reasoning Engine is the brain of the operation, Playbooks are the script guiding each agent in the field. On Ada’s platform, Playbooks work as sets of natural language instructions that define how AI agents should behave in specific situations — from an initial greeting to resolving a complex technical support issue.
Murchison explained that one of the hardest problems in customer-facing AI is how to harness the creativity of a language model while maintaining the strict compliance and determinism that businesses expect, especially those in regulated environments. Ada is an autonomous agent capable of taking creative actions based on all the tools integrated into it, but it can also operate in an extremely deterministic way when needed — and this doesn’t happen in the same rigid, step-by-step manner of legacy systems.
The big differentiator here is that these scripts are written in natural language, which means support teams can create, edit, and adjust Playbooks without needing any programming knowledge. In practice, you can just drag and drop an old standard operating procedure or a flow diagram from a legacy IVR system — whether it’s a PDF or even a screenshot — directly into Ada’s Playbook environment. The platform then automatically generates a pre-configured Playbook with the actions the company has already defined. This really democratizes the setup process and reduces the dependency on engineering teams for tasks that should be operational.
In practice, a Playbook can cover anything from simple flows like answering frequently asked questions about business hours, to more elaborate journeys like guiding a customer through a service cancellation process — collecting information, offering alternatives, and escalating to a human agent only when necessary. This flexibility is what makes Playbooks such a powerful tool within a customer experience strategy.
On top of that, Playbooks work in tandem with the Reasoning Engine, which means agents don’t just follow a rigid script. They can interpret the context of the conversation, adapt the tone and direction of the interaction based on the customer’s history and real-time information. With Ada’s Playbooks, customer service automation stops feeling robotic and starts to resemble a real human conversation much more closely. 💬
Observability and oversight: how Ada makes sure everything works
Automating customer service at scale without visibility into what’s happening is a recipe for trouble. Ada understood this and developed a robust set of observability tools that allow companies to monitor every conversation in detail.
According to Murchison, the platform uses language models to annotate each conversation, including a complete record of the reasoning behind the agent’s decisions. Clients have full visibility into questions like: at what point was the Playbook followed correctly? If an API call failed, at which step exactly did the breakdown occur?
But it goes beyond simple passive monitoring. Ada also has an adherence supervisor agent that follows every conversation involving a Playbook, actively verifying whether the explicit steps are being completed. This information is then displayed on a dashboard that lets teams take corrective action quickly and with the right context.
Beyond the adherence supervisor, there’s a specialized reviewer model focused on understanding resolution. This model annotates each conversation to identify whether the interaction was relevant, whether the information provided was accurate, whether the conversation was safe, and most importantly, whether the customer’s issue was actually resolved.
This layer of automated oversight is especially valuable for companies dealing with massive support volumes, where manually reviewing every interaction would simply be impossible. The combination of real-time supervision with actionable dashboards lets support teams identify patterns, fix Playbook issues, and continuously improve the quality of service delivered by the AI agents. 📊
Simplification versus complexity: the philosophy behind Ada
One of the most interesting aspects of Ada’s approach is the deliberate decision to hide complexity from both the end user and the client company. Murchison acknowledged that technically, there are swarms of agents under the hood communicating with each other to execute different tasks. But the platform makes a point of removing that complexity from the customer experience.
This design philosophy is especially relevant at a time when many companies are facing the challenge of coordinating dozens — and in some cases hundreds — of different agent types that need to be orchestrated. Ada simplified all of this into a system where the company manages just one customer-facing agent that’s deployed and managed centrally.
For technology and operations teams, this represents a significant reduction in management overhead. Instead of dealing with multiple platforms, configurations, and dashboards, everything converges into a single point of control. And for the end customer, the experience is equally streamlined: they interact with one agent that seems to understand the full context of the conversation, regardless of the channel being used.
Why this approach matters for customer experience
The customer service market has gone through many transformations in recent years, but the arrival of AI agents with real reasoning capability represents something different from previous innovations. We’re no longer talking about bots that follow decision trees or simply search for keywords in a knowledge base. We’re talking about systems that can understand the intent behind a message, consider the context of the customer journey, and make decisions that actually make sense for that specific situation.
This completely changes how a customer feels when interacting with a company, because they sense they’re being heard and understood — even if the other side of the conversation is an automated system.
The cross-channel consistency enabled by the unified Reasoning Engine also has a direct impact on customer satisfaction. When someone starts a conversation via chat, continues over email, and finishes by phone, the last thing they want is to repeat everything from scratch at each transition. With Ada’s architecture, the conversation context travels with the customer regardless of the channel. This kind of continuity is something consumers already expect from the most digitally mature brands, and delivering it consistently at scale is exactly where most companies still fall short.
From a strategic perspective, investing in a platform that centralizes agent reasoning and behavior also brings clear benefits for internal teams. Support teams can better monitor each agent’s performance, identify where Playbooks need adjustments, and act proactively before an issue becomes a full-blown customer relationship crisis.
Well-implemented customer service automation doesn’t replace the human touch where it’s needed — it frees people up to focus on the interactions that truly demand empathy, creativity, and judgment. And it’s exactly this combination of artificial intelligence and human intelligence that defines the future of customer service. 🤝
