AI in customer experience: how to measure impact before you scale

A reading of the SDAIA–PwC study on the future of customer experience: five impact dimensions, three agentic AI measures and the main implementation challenges.

The Saudi Data and AI Authority (SDAIA) has published a study prepared with PwC, "The Future of Customer Experience: From Generative to Agentic AI". The study argues that generative AI moves services from digitising procedures to intelligent, personalised service, and expects it to evolve into agentic AI that plans and executes multi-step tasks with minimal human intervention. It offers a framework for measuring impact rather than relying on impressions.

Five impact dimensions

Efficiency: lower costs and less time.
Experience: user satisfaction and response speed.
Inclusion: access for underserved groups.
Trust: fairness and public trust.
Adoption: scope and maturity of use.

Experience indicators cited in the study include CSAT, NPS, Customer Effort Score (CES) and average resolution time.

Three added measures for agentic AI

Autonomy risk management: controlled autonomy under human oversight and policy controls.
Goal alignment and drift detection: continuous evaluation of agent outputs against service objectives.
Decision accountability and explainability: traceable, auditable rationale that regulators and users can understand.

Numbers from the study's survey

The survey covered 35 entities nationally; 23 of them (66%) use virtual assistants or chatbots.
Participants expect generative AI to handle or improve 21% to 40% of customer interactions over the next three years.

What holds implementation back?

The most cited challenges:

Data quality or system integration (21 entities).
Budget or skills constraints (16).
Output accuracy and trust (15).
Privacy and regulation (13).
User adoption and change management (10).

The executive summary also highlights legacy-system integration, organisational readiness gaps and ethical concerns.

Our reading: five steps before scaling

Pick one use case with clear impact and set a baseline across the five dimensions before launch.
Check data quality and system integration before choosing the technology; it is the biggest barrier in the survey.
Put clear agent governance in place: autonomy level, human checkpoint and a decision log.
Link measurement to national digital transformation targets and digital service quality standards.
Don't overlook personal data protection, Arabic-first service and accessibility.

What isn't measured before launch can't be proven after it.

At Business Quality we help organisations assess readiness and design measurement and governance frameworks for AI use cases in services.

Source: SDAIA and PwC, "The Future of Customer Experience: From Generative to Agentic AI" (Arabic)