{"id":34426,"date":"2026-05-15T16:34:00","date_gmt":"2026-05-15T14:34:00","guid":{"rendered":"https:\/\/askme.it\/insights\/public-sector-contact-centers-and-ai-what-works-what-doesnt\/"},"modified":"2026-03-26T12:22:46","modified_gmt":"2026-03-26T11:22:46","slug":"public-sector-contact-centers-and-ai-what-works-what-doesnt","status":"publish","type":"insights","link":"https:\/\/askme.it\/en\/insights\/public-sector-contact-centers-and-ai-what-works-what-doesnt\/","title":{"rendered":"Public sector contact centers and AI: what works, what doesn’t"},"content":{"rendered":"
\n

Since generative AI became widely accessible, governments around the world have been putting it to the test in their contact centers. The stated goal is always the same: reduce wait times, improve service quality, decrease call transfers between agents. The results, after two years of pilots, paint a more nuanced picture than initial expectations suggested.<\/p>\n

Gartner’s analysis of 14 GenAI use cases for government contact centers provides a reference map: each case is evaluated on two dimensions — expected value and implementation feasibility within 18 months — and classified into three operational categories.<\/p>\n

Cases that work: from virtual assistants to staff onboarding<\/h2>\n

The virtual assistant for the contact center is the use case with the strongest profile. These are AI systems accessible via voice or text, integrated into government portals or messaging platforms, capable of answering complex questions by drawing directly from content published by the agency. Unlike traditional chatbots, these systems allow follow-up questions and handle non-linear conversations. The impact on operational efficiency is significant: when properly implemented, they can absorb the largest share of incoming requests, reducing the load on human agents.<\/p>\n

AI chatbots for contact centers — a more evolved version of traditional script-based chatbots — also fall among the high-probability successes. Technical feasibility is good, with development building on existing platforms. The main obstacle remains internal: many administrations still struggle to ensure these systems respond appropriately in all contexts, and agent trust in the technology is not yet consolidated.<\/p>\n

Post-call automation — automatic transcription, content summarization, case management system updates — is another case with an excellent risk\/value ratio. It requires no direct interaction with citizens, reduces agents’ manual work, and integrates relatively easily into existing workflows.<\/p>\n

Contact center staff onboarding via AI — personalized training modules, call scenario simulations, automatic procedure updates — has a high feasibility profile and a direct impact on service quality over time.<\/p>\n

The multilingual contact center, which enables serving citizens in different languages without needing dedicated native-speaking agents, has gained significant ground. Generative AI handles translation and content adaptation in real time with growing quality, breaking down a historic barrier to public service access.<\/p>\n

Calculated risks: high value, complex feasibility<\/h2>\n

Contact center change modeling is an interesting case: AI systems that simulate the impact of changes to procedures, scripts, and responses before they go live, reducing the risk of unintended consequences. The value is clear — greater agility in responding to critical events and policy changes — but the technical complexity of handling edge cases is still significant.<\/p>\n

The virtual assistant for legislation — which helps agents navigate complex regulatory frameworks to correctly respond to citizen requests — has high potential but requires integration with up-to-date legal databases and a level of accuracy that current systems do not yet guarantee uniformly.<\/p>\n

Marginal gains: useful in specific contexts, not a universal priority<\/h2>\n

Call comparison — AI systems that analyze and compare the handling techniques of different agents — is technically feasible but has encountered significant internal resistance. Agents perceive this type of analysis as micro-surveillance of their individual performance. The operational value exists, but the impact on organizational culture must be managed carefully before proceeding with implementation.<\/p>\n

Natural language reporting — queries on contact center operational data without technical skills — offers limited benefits relative to the complexity of integrating with typically fragmented legacy systems.<\/p>\n

The unresolved issue: governance and response control<\/h2>\n

The factor most slowing large-scale adoption is not technical but organizational. Citizen-facing chatbots remain the most debated case among public administrations, with documented successes but slow overall progress. The main reason is concern about controlling generated responses: an AI system that responds inaccurately on tax, social security, or healthcare topics exposes the administration to legal and reputational consequences that are difficult to manage.<\/p>\n

The governments achieving the best results are those that have adopted a hybrid approach: AI for high-volume, low-complexity requests; human agents for sensitive cases; and clear escalation systems between the two levels. It’s not the most technologically ambitious solution, but it’s the one that produces measurable value in the short term while maintaining the control necessary in a public service context.<\/p>\n<\/section>\n","protected":false},"excerpt":{"rendered":"

Governments have been experimenting with generative AI in contact centers for two years. Results show where it’s worth investing and where risks outweigh expected benefits.<\/p>\n","protected":false},"featured_media":34428,"menu_order":0,"template":"","insights_category":[549],"insights_tags":[653,673,689,729,827],"class_list":["post-34426","insights","type-insights","status-publish","has-post-thumbnail","hentry","insights_category-ai-and-customer-service","insights_tags-chatbot-en","insights_tags-contact-center-en","insights_tags-customer-service-en","insights_tags-generative-ai","insights_tags-public-sector"],"acf":[],"_links":{"self":[{"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/insights\/34426","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/insights"}],"about":[{"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/types\/insights"}],"version-history":[{"count":1,"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/insights\/34426\/revisions"}],"predecessor-version":[{"id":34427,"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/insights\/34426\/revisions\/34427"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/media\/34428"}],"wp:attachment":[{"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/media?parent=34426"}],"wp:term":[{"taxonomy":"insights_category","embeddable":true,"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/insights_category?post=34426"},{"taxonomy":"insights_tags","embeddable":true,"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/insights_tags?post=34426"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}