A customer types "cancel my subscription." They don't want to leave. They want someone to notice the thing they paid for has been broken for a week. Any support tool can hand a chat to a human when the AI gets stuck. The harder case is when the AI does everything right, closes the chat as resolved, and still misses what a person would have caught.
The AI agent isn't flying blind. It has the context, the tickets, and the visitor details. What it doesn't have is everything that never made it into the system: the renewal being negotiated over email, the "minor" bug that three larger accounts encountered this week. Someone has to watch for that. At Text, someone now has a title: AI Supervisor.
PwC report: two tracks, one of them winning
PwC's 2026 AI Jobs Barometer examined over one billion job ads across 27 countries. The Technology, Media and Telecoms sector, where Text operates, sits at the sharp end of every trend the report identifies.
Nearly one in eight roles in TMT is now AI-related, the highest share across all sectors analysed. AI hiring in the sector grew 47.8% in 2024 and accelerated to 82.7% in 2025 — substantially outpacing overall job growth. The pace is not a hiring anomaly. It reflects a structural shift in what the work actually requires.
The report identifies two distinct trajectories. Jobs "professionalised" by AI, where AI handles routine execution and leaves the judgment-intensive work to people. This segment is growing twice as fast as jobs "democratised" by AI, with 42% higher wage growth since 2021. The split matters because it determines what kind of team a company is building, not just how many people are on it.
TMT records the strongest productivity growth of any sector analysed: 42% between 2018 and 2025, 13 percentage points above the next highest sector. The companies driving that number are not running leaner teams. The most AI-exposed companies in TMT see faster headcount growth (52% vs 36%) and higher wage growth (24% vs 17%) than the least AI-exposed ones.
There is a less comfortable finding buried in the same data. AI-exposed entry-level roles are seven times more likely than the least AI-exposed junior roles to require traditionally senior skills like leadership and strategic thinking. The repetitive tasks that once served as on-ramps — the tickets that taught new hires how to handle the next, harder ticket — are the first things AI automates. AI-exposed "seniorised" entry-level roles have grown 35% since 2019, while other entry-level roles have declined.
The AI Supervisor role at Text is a direct response to that problem. It is not a title PwC studied because Text launched it before the report was published, but it fits the pattern exactly. A professionalised role built around judgment and oversight, with a structured path in for people who would otherwise have no way to develop the harder skills the job now demands.
AI Supervisors: how Text built the role
Text answered that early, before there was a report to point at. When the reactive workload moved to AI Agents, the easy conversations went first — the ones that doubled as training. The default move at that point is the one PwC describes as the industry norm: thin the team, bank the savings, treat support as a cost to compress.
Text built two roles instead through the transformation of the roles of staff previously employed in customer service. Not by hiring engineers.
Why? Because they possess the greatest experience and product knowledge—irreplaceable assets. On the contrary, this knowledge can be further developed, and new skills can be built upon it.
The AI Supervisor watches the conversations the AI Agent flags and steps in before a customer hits a wall. The Supervisor has full visibility into every conversation between the customer and the AI agent, and can take over at any moment. It’s possible without the most popular friction on the customer’s end: a transfer request from the customer. Every takeover is tagged and logged, so each one becomes a data point: a case the system judged needed a person, and what that person did. That log feeds the next role.
“Introducing the AI Supervisor role has fundamentally changed the way we think about the relationship between humans and AI. Instead of positioning AI as a replacement for people, we've created a collaborative environment where both complement each other's strengths” - Marta Gołąbek, AI Supervisor at Text. “The way we work and approach support now is definitely different. The biggest advantage of this hybrid approach is that it combines the speed, availability, and consistency of AI with the judgment, empathy, and decision-making capabilities of human agents, especially in a role that's all about working with people. It creates a support experience that is more efficient, more proactive, and delivers better outcomes for customers than either could achieve on its own.”
The Automation Specialist takes those cases and teaches the AI agent to handle the next one like it. The Supervisor reads the signal live; the Specialist builds it into the system for good. The system escalates what it can't handle, a human resolves it, and the system learns from the outcome. The shift required rebuilding how the support team spends its time from the ground up. Before the transformation, the Automation team split its attention between live customer conversations and building processes to make the team more efficient — two things that constantly competed for the same hours. That tension is gone now. By replacing manual tasks and third-party tools with native workflows and AI-powered automations, Text created the operational foundation that made both roles viable.
"Our AI Agent now handles the majority of repetitive conversations, product onboarding, and sales skills, while automated workflows support both customers and agents behind the scenes. That allows our support specialists to work as AI Supervisors — monitoring conversations, stepping in when human expertise adds value, and focusing on the cases where they can make the biggest difference." - Justyna Gładysz-Kozak, Automation Specialist.
The practical result is a redefinition of what support work actually is. "The biggest change is that support doesn’t have to react to every complicated chat. What we do is combining AI, automation, and human expertise to continuously improve the customer experience, and not the process itself" - says Justyna Gładysz-Kozak.
What this looks like in practice
In May 2026, Text handled 104,000 total chats. 97,000 of them were resolved by an AI agent end-to-end. Of the remaining conversations, 4,800 legitimate chats were escalated to AI Supervisors — either by transfer or proactive takeover.
The more telling number is the takeover rate. Takeovers — a human joining a conversation without any trigger from the customer — already account for over 10% of escalated chats, based on May data. The expectation is that this share grows every month, and that proactive takeover eventually makes the explicit transfer request largely redundant.
Of the 104,000 total chats, 28,000 (27%) were actively monitored by AI Supervisors. Text is currently working to bring that number down by focusing auto-supervision on legitimate chats even more, rather than the full volume, which includes spam and confused sessions. "First results are very promising, with drops reaching ~60% week over week, but we're still testing this one," says Kacper Wiącek, Head of Customer Success at Text.
Customers notice the difference. Two recent reviews illustrate what proactive supervision looks like from the other side:
"I really appreciate the quick turnaround and the clear communication. This AI support is maybe one of the best I've used"
— Customer service manager, media outlet.
"Tomek was amazing! Seriously, he went above and beyond what I would expect a rep to do when providing me with answers to one-off questions. He took the time to research and find info for me that was actually useful, instead of giving me some generic answer. Shout out to Tomek for being a stand-up guy, and helping me learn more about the Text.app."
— Customer, marketing industry from the US.
Put your service on offense
Stepping into a chat is the first step toward putting service on offense. Text customers, especially in e-commerce, are recording sales growth as a direct result of AI-human collaboration in the chat window. Text-powered chats generate $9.7M in monthly revenue for U.S. ecommerce customers. E-commerce is not the only industry whose customer service metric is measured in dollars — Text measures it the same way.
A concrete example from May: a customer contacted Text support via chat to renew their annual subscription. An AI Supervisor stepped in to accelerate and facilitate the process. The conversation lasted a matter of minutes. The customer renewed for 70 user licenses, got a fast answer to their billing question, and described Text's support as premium. That deal closed inside a chat window.
These aren't old roles with new names
Neither of the two new roles, AI Supervisor, Automation Specialist, is the old support role renamed. Each asks for something the reactive job never did: judge what the AI surfaced, decide what it needs, step in, escalate, or retrain. That's the professionalised track PwC is describing.
"When AI took over most of the daily workload, we didn't want to lose people with the knowledge and experience about our product. They were the ones most capable of training and scaling our agent. This is precisely the element that the media talks about — the gained time and space to focus on operationally important tasks. The AI Supervisor is what Text built to keep that layer of judgment and customer care in the system." - Kacper Wiącek.
The part that the PwC's report doesn't show
PwC's headline numbers are easy to wave around. The honest version of Text's story is quieter. The model is still a pilot, the roles are still being sized, and some assumptions are still being tested. The goal is to catch AI mistakes before customers notice them.
What Text has built is a working mechanism. The way into AI Supervision isn't a pile of easy chats. It's structured exposure to the cases the agent routes up, with the takeover log as the curriculum. The hard conversations became the training, because the easy ones were left.
Text creates the future of Customer Service
The companies pulling ahead use AI to concentrate expertise, not dilute it. The harder problem is the entry-level one: the path people used to climb is disappearing, and someone has to rebuild it. The job titles will keep appearing whether or not anyone plans for them. At Text, that path now starts with a role called AI Supervisor.
The default response to AI handling more of the work is a smaller team. Text made a different bet. AI alone doesn't catch what a person would have caught, and it doesn't close a 70-license deal in a chat window. What it does is a model where every escalation feeds back into making the system sharper — and where the people working alongside AI are developing skills, not losing the conditions that used to build them.
PwC's data shows which companies are growing faster, paying better, and pulling ahead on productivity. They're the ones using AI to build capability. The operational question is whether the people working alongside it are developing skills the market will keep rewarding. Text's answer is structural: a role that didn't exist last year, a curriculum built from real escalations, and a support operation that measures itself in revenue. That's already running.
About Text Inc.
Text Inc. is built on the foundation of LiveChat, the category-defining platform that pioneered modern live chat and brings over two decades of experience in conversational customer experience, now extended through AI-driven capabilities. The company serves more than 35,000 customers worldwide, including Unilever, Atos, Wembley Stadium, MIT, Stanford, and Coop Travel. Text Inc. is headquartered in Boston, Massachusetts, and is part of the Text Group (WSE: TXT, TXT.WA).
