T-Mobile: relational intelligence as the future of B2B customer care
A pilot that validated what comes after chatbots: digital personas that hold real conversations, integrate with enterprise systems, and are trusted by 91% of users on first contact, in one of Europe's most demanding B2B segments.
- 90%+
- 89%
- 91%
- 93%
T-Mobile Czech Republic
T-Mobile Czech Republic is a leading telecommunications provider and a subsidiary of Deutsche Telekom. They serve millions of customers across consumer and enterprise segments, with a strategic focus on positioning themselves as a leader in technological innovation.
T-Mobile sought a fundamentally new approach to B2B customer care, one that goes beyond scripted chatbots and IVR menus to deliver genuinely personalised, context-rich interactions at scale.
B2B customer care has hit its ceiling.
Business customers demand expertise, speed, and personalisation. Traditional channels (call centres, self-service portals, scripted chatbots) each solve one dimension at the cost of the others. T-Mobile needed to validate whether a new kind of interaction could serve their most demanding B2B clients.
Complexity meets scale
Roaming alone involves zone-specific pricing, package recommendations, activation workflows, and account-level permissions, multiplied across hundreds of enterprise accounts.
Conservative audience
B2B clients, from financial institutions to manufacturing, are the most demanding and risk-averse audience. Any new channel must earn trust immediately.
Beyond information delivery
The goal was not another FAQ bot. It was to validate whether a relational digital persona could proactively guide B2B customers to optimal solutions through natural conversation.
Digital personas that hold real conversations.
Promethist deployed two digital personas, a male and female assistant with distinct communication styles, directly into a native mobile app, connected to T-Mobile's internal systems. Users authenticate with their T-Mobile credentials, and the agent immediately knows their account context: active services, tariff configuration, and permission levels.
The pilot focused on roaming: a deliberately complex domain requiring zone-specific pricing, package recommendations, and real-time service activation. If the concept works here, it works anywhere.
What the agent does in a single conversation
- Identifies the destination zone and checks current data availability
- Informs the customer about their active services and any restrictions
- Asks about travel duration and data requirements to recommend the optimal package
- Offers personalised package recommendations with pricing
- Activates services and records changes to CRM, all within the conversation
- Hands off to a human operator with full context when needed
- Generates conversation reports for the operator team
Not a chatbot. A relational digital persona.
Natural conversational flow
Not a scripted chatbot. The assistant leads genuinely personalised conversations, adapts to context, and proactively guides customers toward optimal solutions.
System integration
Connected to T-Mobile's internal systems in real-time, reading tariff configurations, account settings, and active services to deliver precise, context-aware answers.
Multiple personas
Male and female digital assistants with distinct communication styles, letting users choose who they prefer to interact with, increasing engagement and comfort.
Voice-first interaction
Full voice conversation via native mobile app: users speak naturally and the assistant responds conversationally, making the experience feel human.
Real-time service activation
Not just informational: the assistant can activate roaming packages, record changes to CRM, and execute service modifications within the conversation.
Proactive guidance
Goes beyond answering questions: proactively identifies opportunities, recommends optimal packages based on travel plans, and surfaces relevant information unprompted.
A real conversation, not a scripted flow.
Here's how the digital persona handles a real customer scenario, identifying needs, proactively surfacing issues, and guiding toward resolution in natural language.
How much will it cost to call from the UAE?
Calling from the UAE costs 57.02 CZK per minute. I notice you don't have roaming activated. If you activate it, you'll be able to call and use data abroad. Shall I remove this restriction?
The agent doesn't just answer. It identifies the underlying issue and proactively offers resolution.
I'll be there two weeks.
Enjoy! For a two-week stay in the UAE, I'd recommend the Data World C 2 GB package, valid for 30 days at 1,279 CZK. If you need more data, there are larger packages available. Want me to activate this one?
Personalised recommendation based on travel duration, then offers to execute the change immediately.
Strong validation across every dimension measured.
The pilot ran in the B2B segment during summer 2025, with methodology co-designed with research agency Ipsos to ensure rigorous data quality and interpretive validity. Results were consistent across usefulness, trust, interaction quality, and willingness to adopt.
90%+
Perceived as useful
Respondents consider the digital persona a relevant and valuable extension of T-Mobile's services
91%
Responses rated trustworthy
Users found the assistant's answers credible and reliable, even on their first interaction
89%
Would use it again
Would return to the assistant if it covered a broader range of topics: a clear signal of product-market fit
90%
Smooth communication
Rated the conversational experience as easy or very easy: no friction, no learning curve
84%
Valued persona choice
Found the option to choose between multiple digital assistants with different styles attractive
93%
Ready for voice
Can envision communicating with the assistant by voice in the future, confirming multimodal demand
Objective accuracy: 93% correct.
Beyond subjective surveys, an objective analysis of conversation transcripts was conducted throughout the pilot. In a random sample of 45 roaming recommendation scenarios, 42 conversations were fully accurate and compliant with the business brief.
The remaining cases were either edge cases intentionally excluded from scope (at T-Mobile's request) or minor tone issues, not factual errors.
94%
Rated response quality as excellent or good
91%
Found responses trustworthy
80%+
Satisfied with visual design
73%
Would use it as a standalone app
Relationship builds business, even with AI.
The digital assistants were deliberately designed with extended conversational capabilities beyond roaming queries. The hypothesis: stronger relational skills lead to more engagement, which leads to more business conversations. The data confirmed this.
Small talk leads to business
Users who initiated casual conversation with the assistant naturally transitioned to resolving business topics, building relationship first, then transacting.
Testing builds trust
Users who initially tested the assistant's limits (trying to 'catch it out') subsequently became power users. One went on to have 10+ conversations and 120 conversational turns.
Travel tips drive engagement
The assistant provided contextual travel recommendations alongside roaming information. One user had 6 conversations with 44 turns total, returning specifically because of this capability.
From pilot to permanent channel.
The pilot validated the core hypothesis: digital personas can deliver relevant, personalised, and trustworthy support even in the most demanding B2B environments. Users expressed clear demand for broader topic coverage, role-based personalisation, and multimodal interaction, confirming the path to production deployment.
Broader service coverage
Users requested tariff management, number administration, billing support, and device setup. The platform is ready: expanding topic coverage is configuration, not development.
Role-based personalisation
Administrators need fleet management. End users need personal service answers. The agent adapts its capabilities and knowledge to the role of whoever is speaking.
Continuous learning
Production deployment enables continuous data collection and development tracking over time, moving from snapshot validation to ongoing optimisation.
Multimodal expansion
93% of users are ready for voice. The infrastructure supports voice, text, and visual interaction simultaneously, meeting users in whichever mode fits their context.
Bring relational intelligence to your customer experience.
From pilot to production, we build digital personas that earn trust, resolve real requests, and transform how enterprises interact with their customers.