What layer has AI reached in your customer relationships?
The question is no longer whether you use AI in customer service. 18 questions about your actual state show where your customer experience sits on a 5-level ladder — from Ad-hoc to Autonomous — which pillar is holding it back, and what is worth doing first.
5 minutes · 18 questions · no sign-up needed to see your result
How it works
Answer 18 questions
Each question offers 4 descriptions of an actual state — from manual to predictive. Pick the one closest to how your company serves customers today.
See your level and bottleneck
Your overall level on the 5A model, a score per pillar, and the pillar that is limiting the whole experience.
Get a roadmap
Concrete recommendations for every pillar and the 3 things to do in the next 90 days.
Company-wide AI maturity ≠ AI maturity in front of your customers
A company can be highly mature about AI in internal operations and still serve customers the way it did ten years ago — and vice versa. This assessment measures only what customers can touch: interaction channels, depth of understanding, proactivity, the CX team, experience governance and measurable value.
Want to measure AI capability across the whole organisation? Take the company-wide assessmentThe 6 pillars we assess
- Interaction & Channels
Where does AI meet your customers for you, how much does it carry, and does it step back at the right moment?
- Customer Understanding
Does whoever answers — human or machine — know who the customer is, and is the voice of the customer heard at scale?
- Personalisation & Proactivity
Are you waiting for customers to complain, or reaching them before they have to?
- People + AI in the CX Team
Is AI lifting your CX team, or just running alongside it?
- Experience Governance
When AI stands in front of a customer, who is holding the brakes?
- CX Value
Is AI in CX an experimental cost, or a number in your business reporting?
5 maturity levels — the 5A model
- Level 1Ad hocAd-hoc — humans handle everything
Every customer interaction goes through a person. A few staff quietly use AI tools to draft replies, but the company has no deliberate direction. Response speed depends on who is on shift — a message at 2 a.m. waits until morning.
- Level 2EstablishingAssisted — AI helps in isolated spots
There is a scripted chatbot on one channel, or an auto-reply. The AI does not know who the customer is, freezes on anything off-script, and connects to no data — outcomes depend on pre-written scripts rather than organisational capability.
- Level 3PerformingAutomated — AI carries real load
AI handles a significant share of common requests on the main channels, hands over to humans with context, and the numbers are measured monthly. But each channel and team is still an island — a customer who switches channels starts over.
- Level 4OptimizingAnticipatory — proactive for each customer
AI sees a complete customer profile and uses it to personalise every reply. The company reaches out on early signals instead of waiting for complaints, and the voice-of-customer feedback loop closes automatically.
- Level 5EmbeddedAutonomous — CX that predicts and improves itself
AI orchestrates journeys end-to-end, intercepts problems before customers report them, and learns from every interaction. Humans focus on exceptions and relationship moments. Cost-to-serve and value are managed like a P&L.
Sidotech: from scripted chatbot to the Performing level
Sidotech's 6-person support team used to answer hundreds of technical messages by hand every day; their old scripted chatbot could not cope with a catalogue of hundreds of products. After deploying an AI Agent with Filum.ai, nearly 90% of conversations on Zalo and Facebook are handled end-to-end by AI — and the team now focuses on the cases that genuinely need a human.
How to answer
Answer based on what consistently happens today, not on plans or intentions. A chatbot you once piloted and switched off does not count as the current state.
Not sure? Pick “Not sure”. It is a valid and useful answer — we count it separately instead of scoring it low. The fact that nobody can see part of the CX picture is itself a finding.
Your result renders right in your browser. You only leave your details if you want the detailed action roadmap.
