AI agents are ushering in rapid change to organizational structures and workflows. From new ways to automate work and crunch data to unlocking new opportunities for growth, AI is being deployed vertically and horizontally. It is forcing leaders to look at their businesses differently, to identify the ways that AI can have a real, tangible impact. This includes not only improving current processes and functions and making them smarter and more efficient, but also shifting the way they do business, engage with customers, and train employees. This shifting part, including which functions to start with, is not always obvious, but one area that is seeing its scope expand with the help of AI agents is customer service.
Customer service has traditionally been a function meant to get to a resolution as quickly and with as little friction as possible. Humans have traditionally been at the front lines, but automated systems and bots have largely replaced many of the interactions, with humans jumping in to avert an issue or handle an escalation. The case for using agents to make this process more autonomous is an obvious use case, and there are many emerging examples showing that resolution rates from agents are on par with those of human-centric models. Customer service platform Text, for example, has seen AI agent resolution rates average 74% across its customer base using these features, with some reaching 90% – all done autonomously.
AI Agents Can Put Service Teams On Offense
But there is a bigger opportunity here that isn’t so obvious. Customer service is often disconnected from other core parts of a business, such as product development, sales, and marketing. It, and the underlying data, sit in a silo.
That is changing with today’s AI agents, and in particular with two new features recently released from Text that expand upon what’s possible. The customer service function is ripe with powerful signals and insights that can be put to work proactively through the use of AI agents. It can provide context, predict intent, and act as a sales agent – autonomously, with human support as needed. Brand history, search behaviors, previous purchases, questions and prompts, and profile information can all be ingested in real-time to identify intent and new revenue opportunities. Customer service can now be a powerful profit driver for companies.
Selling Agents & Custom Skills
Last week, Text introduced two new features within its existing product lineup to enable companies to actively sell in a live chat environment, combinating automation and human involvement – selling agents and custom skills. The company began testing these features with an initial test group of 600 e-commerce companies back on March 1.
These new agents combine behavioral signals, customer history, and in-session context to identify moments when proactive assistance increases conversion probability, and they intelligently automate much of the traditional customer support work.
- Selling Agents – Text’s AI selling agent can be easily deployed and trained on a company’s product catalog, brand voice, and business rules, enabling it to identify intent, proactively engage customers, and convert conversations into purchases directly within chat, all in a single window. The system is built to operate in real time, using customer context, such as browsing behavior, purchase history, and prior interactions, to guide conversations and influence outcomes. It can recommend products, surface alternative options, qualify leads, and complete transactions without human intervention.
AI selling agents help companies scale their sales and business without expanding teams. Sales and customer service professionals will still be in the loop, but the agent can spot intent, read context, share offers, and even close sales without a human present – which could be after work hours or during busy periods. If a human is needed, the AI selling agent will understand and loop in a human, even before the customer requests it.
In practice, an AI agent can recommend a complementary product based on what’s in the cart, surface a relevant promotion when a customer hesitates on a product page, or qualify a lead before routing them to a specialist. These are actions traditionally reserved for trained sales professionals. Now they happen at scale, around the clock, without requiring additional headcount.
- Custom Skills – With custom skills, flexible frameworks can be used to define how the AI agent behaves in specific scenarios. This includes the ability to create structured workflows that guide AI actions based on customer intent, including resolving issues, collecting information, offering incentives, and triggering follow-ups. For example, an agent can detect user intent based on preset criteria such as page views, location, or time zone. Then, it triggers a set sequence of steps. Combined with customer context — name, location, browsing history, order count, past tickets — responses and interactions are hyper-personal and relevant to what the customer may want or need.
Custom skills can also do more than just support. Agents can offer discounts and ask for reviews after successful resolutions. They enable companies to scale without sacrificing quality. They package proven actions into reusable scenarios that are triggered at the right moment, and can loop in a customer service or sales agent as needed.
Early Results Point to a Bigger Shift
Although these features are new on the market as of last week, data from March and April from initial customer deployments confirm the approach is delivering positive, tangible results. From the 600-company test group, the following data points stood out from the two-month period:
- Chatting with AI agents improved overall order conversion rates by 266%.
- Chat sales attribution increased by 39% over the past month.
- Sales Operations rose by nearly 7% over the past month.
- AI agent adoption surged in April, with a 41% increase in the number of agents deployed.
- AI agent engagement grew even faster, increasing by 60%.
While the data is still early, the results suggest that live chat is beginning to function as more than a support channel for some businesses. These figures represent a fundamental shift in how businesses generate revenue online and with the support of AI agents. These are real deployments by real companies that are generating revenue from a channel that previously appeared only as a cost center.
Humans Still Have To Be In The Loop
None of this works without human expertise. The smartest implementations recognize that AI agents need to be trained on product catalogs, brand voice, business rules, and escalation criteria. That training doesn’t happen once — it requires ongoing refinement by people who deeply understand the business and who have experience working with customers.
Equally important is the handoff. The best AI systems are built to recognize when a conversation requires human judgment: a frustrated customer, a complex product question, a high-value prospect who needs a personal touch. At that moment, a human agent should be able to step in seamlessly – ideally before the customer asks. The agent sets the stage, and the human completes the ticket and/or closes the deal.
Text recently introduced an AI Supervisor training program for its customer service employees, equipping them with the skills to manage and work alongside their AI agent counterparts. These agent supervisors can monitor performance, refine workflows, audit outcomes, and identify gaps that require human intervention. The role requires analytical thinking, a strong understanding of customer psychology, and the ability to translate business strategy into AI behavior. Text has been able to upskill its entire 40-person customer service team to become AI supervisors, providing them with new skills and offering incentive pathways that were previously unavailable. Text is sharing this training program for free with its customers and will share it with the public on major e-learning platforms in the coming weeks.
More information about Text’s AI agent features is available at Text.com.
