AI service that lets the customer choose: how we implement it
We build service in which AI resolves what is simple and the customer can ask for a person at any moment. In our reading, giving that option is what makes the customer accept AI. Gartner predicts that 95% of service leaders will keep humans in the operation, and the CX Brain and AntennasBI study shows a score of 9.3 for human service and 7.0 for AI.
Why choice enters the design
In the CX Brain and AntennasBI study (Sep/2024, with 1,003 Brazilians), 56% of respondents felt negative emotions in AI service, and 51% said AI helps little or nothing. The study doesn’t state the score scale or the exact question, so we compare the two values only against each other.
What this data suggests for the design: whoever feels stuck with a bot leaves the conversation. Whoever can call a person tends to stay longer in the automated step. This last sentence is our design hypothesis, and we measure it in each implementation.
What we deliver
Topic map. We separate what AI resolves on its own, what it prepares and hands to a human, and what goes straight to a person.
Handoff rules. The customer asks for an agent and the conversation context goes along, without repeating information.
AI agent with clear limits. A reviewed knowledge base, answers AI can’t give and a record of each handoff to a human.
Indicators dashboard. Resolution rate, handoffs to humans, score by service type and reasons for abandonment.
What the path looks like
- Diagnosis of the channels and the most frequent topics.
- Design of the entry, choice, service and resolution flow.
- Building the agent and the integrations with CRM and WhatsApp.
- Pilot with part of the volume and tuning of the handoff rules.
- Operation with tracking of the indicators.
When it makes sense
For operations with a recurring volume of similar questions and channels like WhatsApp, chat and phone. If your service doesn’t have the topics mapped yet, we start there.
To understand the basis of this design, read what the Brazilian consumer thinks about AI service and how to close the score gap between human and AI. For those who want to go beyond service, see custom sales and service automation. If you want to talk about your case, talk to NorthStar.
Frequently asked
Can the customer talk to a person at any moment?
That’s the principle of the design. In each implementation we define how the request is made and how the context goes along.
Does AI resolve everything on its own?
No. It resolves what is mapped as simple, prepares what is intermediate and routes the rest to a person.
How long does implementation take?
It depends on the volume, the channels and the integrations. The timelines are defined in the diagnosis.
Does this reduce the service team?
The goal is to free the team for the cases that require judgment. Gartner predicts that 95% of service leaders will keep humans, and that is our design too.
How do we measure the result?
With resolution rate, handoffs to humans, score by service type and reasons for abandonment, tracked from the pilot on.
Sources
- Consumidor Moderno, on the CX Brain and AntennasBI study, Sep/2024. consumidormoderno.com.br
- Gartner, forecast on abandoning plans to cut human service, 10/06/2025. gartner.com
MARKETING · SALES · FINANCE · LOGISTICS
Intelligence for the whole team.
Scale for the business.
All your data in one layer: every team decides better and the company grows on numbers.
SCALE WITH DATA →