Only 4% Have Full Customer-Service AI — in the Vendors' Own Survey
The survey, commissioned by the vendors boost.ai and Ciklum, shows that 41% of the 200 European decision-makers report running customer-service AI pilots with nothing yet in full production.
Only 4% Have Full Customer-Service AI — in the Vendors' Own Survey
The survey, commissioned by the vendors boost.ai and Ciklum, shows that 41% of the 200 European decision-makers report running customer-service AI pilots with nothing yet in full production.
On 10 September 2026, the companies boost.ai and Ciklum published the report “2026 State of AI in BFSI Contact Centers: From Pilot to Agentic CX”, based on a survey conducted by Coleman Parkes on the two companies' joint commission. According to the survey, 41% of the 200 decision-makers in European banking, finance and insurance say their organisation is running AI pilots in customer service without anything having reached full production. Only 4% report a fully “agentic”, integrated service model — what the report's subtitle calls “agentic CX”.
Both commissioners are themselves vendors of AI and contact-centre services to the same industry.
Pilots dominate, full scale is rare
Between the extremes lies a large group: 32% of respondents say they have AI live in only one channel. The press release uses the figure to underscore the distance between early experimentation and integrated, governed AI operations that deliver measurable business returns.
$11 per voice interaction — and two main obstacles
According to the survey, the average cost per voice-based customer interaction is 11 US dollars across the eight markets. The press release ties the cost level to pressure on contact centres, while also pointing to significant barriers to rolling out AI at scale. Complex integration with legacy IT systems is named the biggest obstacle, cited by 43% of respondents. Concerns about accuracy and hallucinations also weigh heavily, according to the press release: inaccurate AI responses are the leading concern at rollout, cited by 32% of organisations across all the markets.
Cost control today, revenue ambitions in 2027
61% say their organisation primarily uses AI for cost control. At the same time, half say their vision for 2027 is proactive, revenue-generating customer service across channels. The press release gives the share only as “half”, without an exact percentage.
The gap between today's cost focus and the 2027 ambitions, for that matter, concerns a market into which both commissioners themselves sell services.
200 decision-makers in eight countries
The fieldwork was carried out from 11 May to 1 June 2026 in eight European countries: Austria, Denmark, Finland, Germany, Norway, Sweden, Switzerland and the UK. Norway is included in the sample, but the results are presented only in aggregate — with no country-level figures or other market breakdown of the key numbers. The respondents come from banking (40%), insurance (40%) and finance (20%), and 42% say they have final decision-making authority over the customer-service AI strategy, while 58% are part of the decision-making team.
What vendor-funded research can show
In the press release, the report is described as “independent research”. That refers to the fact that Coleman Parkes carried out the fieldwork — not to an absence of commercial interests. Both commissioners themselves sell to precisely the sector being surveyed, and an industry picture in which many AI pilots never reach production bolsters both companies' sales argument.
All the figures are also based on self-reporting: respondents describe the status of their own organisations, and the data has been neither audited nor independently observed. AIMag has not had access to the report itself, the crosstabs or the questionnaire — the basis for this article is the press release announcing the study, distributed via PRNewswire on 10 September 2026 with a dateline of London and Sandnes, Norway.
Webinar and commercial follow-up
The findings are quickly followed up commercially: on 16 September 2026 at 15:00 CET (14:00 BST), the companies will host the webinar “Why AI Pilots Fail to Scale: Lessons from 200 BFSI Leaders” on the results.
The press release quotes Robert Easter, CRO at Ciklum: “The partnership is living proof that institutions can transition from disconnected experiments to secure, production-ready AI that drives efficiency, elevates agent performance, and delivers measurable value.” The statement concerns Ciklum's own partnership with boost.ai and must be read as the company's assessment of its own collaboration — not an independent assessment of the industry.
Source: Press release from boost.ai/Ciklum via PRNewswire, 10 September 2026 · Source ID: f66d8266-aabc-4d38-8ca0-a6b4f79b4bdf