From conversational magic to real automation
Martijn Bosschaart, Senior Solution Architect, DevRev EMEA
On April 10th, at the DevRev Executive Leadership Circle in Amsterdam, we had the opportunity to showcase and demonstrate something that’s been a big focus for DevRev: Conversational customer experience powered by DevRev's AI Agent.
The live demo illustrated how a customer from a fictitious company NeuralPay (imagine a customer of your organization) can resolve real problems through a natural conversation without ever needing to talk to a human.
From disputing transactions to requesting new features and even evaluating financial products like credit cards, everything happened in a single, seamless flow.
A 360° experience. And while the front-end was intentionally magical and simple, the real story lies in how it works within the Agentic AI platform of DevRev, AgentOS.
AgentOS is a platform that integrates humans and AI agents into a single system, providing them with the necessary data and tools for efficient task completion.
It connects modern computing with legacy systems, allowing users to maintain existing tools while benefiting from new technology. AgentOS consolidates data from various sources into a smart knowledge graph, facilitating easy information retrieval through semantic search and enabling real-time automation with a serverless workflow engine.
Additionally, it features an in-browser analytics engine to help AI agents manage complex tasks, enhancing productivity and customer satisfaction.
This blog describes the tech behind the demonstration in detail, and serves as an invitation to explore this capability in your own organization's environment.
Above: A screenshot of the NeuralPay banking demo showing the interaction between the end user and DevRev’s PLuG Conversational Agent.
PLuG conversational Agent is a part of the DevRev AgentOS platform designed to facilitate communication and understanding between users and customer-facing teams.
It includes features such as Search, Chat, and Observability, enabling effective customer support and engagement.
PLuG can be integrated into websites and configured to appear on specific pages, allowing users to start conversations and access relevant information.
It also supports email integration and can escalate complex queries to support agents.
Recap: What the demo showed during the Executive Leadership Circle event
Here's a quick overview of what we demonstrated:
- A logged-in user on a banking interface opens a conversation with "PLuG," our AI-powered assistant.
- The user disputes a transaction via natural language. The AI:
- Retrieves the relevant transactions via an API.
- Understands the user's request and submits a formal dispute.
- Updates the transaction status in real time.
- Creates an internal ticket for human review.
- The user gives feedback ("I wish I could do this from the mobile app"), and PLuG:
- Automatically routes the request as a feature request to the product team.
- The user inquires about rising car expenses. After this, PLuG:
- Analyzes spending by category.
- Suggests a credit card promotion.
- Calculates potential savings.
- Signs the user up for the card—instantly.
Everything happens in natural language, with backend operations triggered as if by magic.
So, how does this magic work?
Let’s break down the two key backend components that made the demo possible:
1. Workflow Builder
A Workflow Builder is a capability as part of AgentOS that allows users to create, configure, and manage workflows. It typically involves setting up a series of triggers, actions, and conditions to automate tasks and processes. This capability is designed to be flexible, catering to both high-code and no-code users, and often includes pre-built nodes for common tasks to simplify the creation of complex automations. The goal is to enhance productivity and efficiency by automating repetitive tasks and providing precise control over various processes.
The Workflow Builder in DevRev enables non-engineering teams to design and maintain backend logic without writing extensive code. It’s what allows the AI agent to:
- Retrieve data from APIs (e.g., user transactions).
- Trigger downstream systems (like refund processing).
- File tickets, notify teams, or update a CRM.
- Send confirmations and summaries back to the user.
Every step the user saw in the demo was powered by workflows that we designed visually—and that you can, too.
2. Agent Manager
This is the AI brain of the operation. One of the most empowering features of the Agent Manager is that you can define its behavior using natural language prompts—not code. Instead of writing rules or logic in a programming language, you simply instruct the agent in plain English what to do, how to act, and which skills to use.
It combines:
- Authentication awareness: Since the user is logged in, the agent knows who they are and what they’re allowed to do. Any authorizations needed for backend operations are passed onto the agent, like API keys, or tokens generated by the banks’ website when the user logs in.
- Skill routing: Based on intent detection, it selects and activates the right skills (e.g., “DisputeTransaction,” “ListTransactions,” “UpgradeCard”).
- Multi-turn memory: The agent can continue a conversation naturally, link previous actions, and drive toward outcomes.
Knowledge retrieval: It taps into product documentation or knowledge base articles in real time to support answers with relevant facts. In fact, the terms and conditions document for the Gold Card was Airdropped, DevRev’ data synchronization (one way or bi directional) into DevRev’s Knowledge Graph. This allowed the AI to accurately answer questions about fees, cashback percentages, and perks, as if it had read the fine print itself.
Why this matters
This becomes a blueprint for how modern customer experience (CX) can work and help both your organisation and your end users/ customer
- Low-friction UX: No menus, no app switching, no long wait times.
- Automated back office: No hand-offs to overwhelmed agents.
Closed-loop intelligence: Customer asks → system acts → feedback routed → product evolves.
Traditional CX vs. Conversational Agent
To highlight the difference between conventional customer support and DevRev’s agent-based approach, here’s a side-by-side comparison of core capabilities:
Feature | Traditional CX | DevRev Conversational Agent |
|---|---|---|
User Interface | Menus, forms | Natural language |
Backend Actions | Manual handoffs | Automated workflows |
Feedback Loop | Survey-based | Real-time, in-convo routing |
Dev/PM Involvement | Delayed | Instant ticket generation |
Time to Resolution | Hours–Days | Seconds–Minutes |
The result
A fully automated Pre-Sales Assistant working in natural language—turning research and outreach into a single flow.
From chaos to clarity: Behind the scenes of our Sales AI demo
Richard Lansdorp, Senior Solution Engineer DevRev
While in the previous Blogpost from Martijn Bosschart “ From Conversational Magic to Real Automation” we described automating customer service, we now describe a sales representative —navigating data sprawl, disconnected tools, and the constant chase for context.
We demonstrated how DevRev turns messy enterprise data, both structured and unstructured, into decisive, AI-powered action—starting with sales, but extendable across every function, all supported by our AgenticAI Platform, AgentOS.
Setting the stage: The demo story explained
In the presented scenario, we followed a sales representative at a fictional company, NeuralPay, preparing for outreach to Heineken, a global strategic account buried among thousands of leads.
With just one question; “What are our top leads?" our sales representative kicked off the workflow. Instead of digging through reports, hopping between multiple tools, or waiting on RevOps, he got an instant, prioritized answer via DevRev AI Search Agent.
Heineken surfaced as a top opportunity—based not only on CRM data, but also on Support tickets, Product signals, and Engagement history. A full 360 view.
That’s where the magic started for the Sales Representative in this demo scenario;
Step 1: Fast, permission-aware research
The Sales Representative asked the system to “research Heineken.” In seconds, they were served a comprehensive view:
- Company profile
- Key stakeholders
- Tech stack
- Past interactions across support and product
Without switching tabs and without any permissions chaos. Just clear, contextual insight delivered in a clean feed—thanks to Airdrop.
Step 2: Agentic workflows kick in
Behind the scenes, this research step triggered an agentic workflow modeled on what your best sales rep would do—competitive analysis, stakeholder mapping, and industry tailoring all rolled into one workflow.
Another AI agent picked up from there and generated a custom pitch, aligning talking points to Heineken’s known challenges, needs and the personas involved—like the CFO of the organisation.
Step 3: Language, tone, timing
The system drafted an email to the CFO—in Dutch. It picked up language preferences, struck the right tone, and even queued up a suggested follow-up cadence. All of it was editable, reviewable, and important a human-in-the-loop.
What used to take hours of jumping between systems and chasing context now happened in minutes—with strategic nuance and zero cut corners.
Under the hood: How it all works
None of this works without three core pieces under the hood:
1. Airdrop – Breaking down silos
This is where it starts. Airdrop pulls structured and unstructured data from your tools—CRM, support, product systems, internal docs—and pipes it into one central, searchable layer.
Airdrop is DevRev’s solution for migrating data. It allows customers to import existing data from external systems into DevRev, export data back to those external systems, and keep data synchronized between DevRev and the external systems. DevRev offers various snap-ins for integration with systems like Jira and Zendesk, and also enables developers to create their own connectors for other systems.
2. Knowledge Graph – Building context that matters
Once Airdrop brings in the data, the Knowledge Graph gives it shape and meaning. Relationships form between tickets, opportunities, customers, and features—so agents can understand the “why” behind the “what.”
Knowledge graph is an ontology that defines connections among product parts and people. It is used for various purposes, such as creating product enhancements from customer issues and generating enhancement release notes from linked documents.
3. Agentic Workflows – From insight to execution
Once context is set, AI agents and workflows jump in. They trigger actions, assign tickets, update fields, and launch sequences, mirroring your best humans and working alongside them.
A workflow is a series of triggers, actions, and conditions designed to achieve a specific goal or outcome. It is depicted as a flowchart or diagram that illustrates how various triggers and actions interconnect.
Workflows automate tedious tasks, providing precise control and efficiency, and can be tailored to an organization’s needs. They leverage AI to create adaptive, real-time workflows that handle complex tasks with minimal human intervention, enhancing productivity and efficiency.
No-code, all power: Build it yourself
One of the most exciting things is how accessible DevRev Platform is.
With DevRev’s no-code agent and workflow builders, any team can replicate this.
- Define an agent goal.
- Write clear instructions.
- Attach your own knowledge and plug into relevant workflows.
It’s like giving every team a senior analyst, researcher, and operator in one.
Want to route a customer request?
Build a rep-level pitch?
Surface the next best action for support?
You can do it—without writing a single line of code.
Why this matters
This demo wasn’t just about sales. In this session we just selected this particular function. It can be any across your organization.
It’s about what happens when your system knows how to talk to each other—when AI doesn’t just answer questions, but takes action with context and care.
It’s about bridging the gap between insight and impact.
And it’s about giving every team—from Sales to Support to Product—the power to execute faster, smarter, and with more consistency and personalization than ever before.
Let’s explore it together
Curious how DevRev could work in your organisation’s environment? We’d love to walk you through:
- How to train your own AI Agent using DevRev.
- How to build and test workflows in minutes.
- How to deploy this in a real-world setting—fast.
- How to build your own AI agents
- How to connect the tools you already use
- How to go from demo to deployment in days
Just reach out. Let’s make your workflows work for you.
