Three different entities, Microsoft, SDG Group, and Salesforce, agree that the shift is not merely technological: It affects how decisions are made, how cases are processed, and how the very role of public administration is understood in a country where regular AI use among the population already exceeds 40%, according to the Global AI Adoption in 2025 report.
The underlying diagnosis is shared: AI is being introduced into a public-sector ecosystem saturated with information, characterized by documentation-intensive procedures and a strong regulatory framework, while also facing growing pressure to shorten processing times, increase transparency, and deliver more personalized services. “Artificial intelligence offers public administration a highly significant opportunity to improve its internal operations while also enhancing the quality of services provided to citizens,” summarizes Javier Jimeno, Partner for Public Sector and Alliances at SDG Group. His comments highlight a key idea echoed throughout all three interviews: The value of AI lies in its ability to address specific public-sector needs more effectively, not in the mere adoption of fashionable technologies.
Along the same lines, Natalia Escobedo, Director of the Public Sector at Microsoft Spain, argues that AI is already “driving a profound modernization of public administrations in Spain,” with tangible benefits concentrated in three areas including efficiency, improved citizen services, and data-driven decision-making.
Meanwhile, Jesús Galindo, Vice President for Public Sector at Salesforce, emphasizes that combining AI with sector-specific data and digitized workflows “increases the resilience of public administrations in the face of growing challenges and strengthens public trust in government services.”
The first layer of impact is being felt in purely administrative processes. Today, public administrations handle large volumes of cases and files, involving heterogeneous documentation, sequential validation procedures, and strict legal deadlines.
According to Jimeno, AI helps “better manage the large volumes of information that administrations work with today, reducing administrative burdens and freeing up time for tasks that require professional judgment.” Generative AI tools are being deployed precisely in those stages where the work is largely mechanical: reading and classifying documents, extracting key information, drafting preliminary resolutions, reports, or routine communications.
Escobedo explains that projects with public-sector organizations always begin with a “baseline” assessment of the process without AI—how a case is currently handled, how long it takes, and how many hours it consumes—and compare it with a controlled pilot in which Copilot and other generative agents assist with tasks such as document retrieval, report preparation, and case management. This methodology makes it possible to translate productivity gains into highly specific indicators like hours saved per case, the volume of automated tasks, reductions in response times, and improvements in service levels.
These internal measurements are complemented by independent studies, such as Forrester’s Total Economic Impact analyses of Copilot for Microsoft 365, which point to a return on investment of up to 457% over three years and report that users feel more productive when carrying out writing and document-analysis tasks.
Salesforce employs a similar approach, combining operational-efficiency metrics—such as reductions in average resolution times, increases in the number of automated procedures, and decreases in manual tasks—with citizen-experience indicators, including first-contact resolution rates and response times. “Productivity gains are measured through improvements across multiple operational dimensions,” explains Galindo, who highlights the ability of AI agents to break down complex requests into simple actions and propose automated execution plans that relieve public servants of repetitive activities. In bureaucratic environments, he notes, it is essential to maintain traceability for every decision and ensure that models operate using official data and clearly defined contexts.
All three companies point to specific success stories within Spain’s public administrations that help bring this narrative down to earth.
SDG Group has supported a regional government administration in improving the management and processing of grants and public aid programs, an area that is typically high-volume and highly sensitive. In this project, a cognitive document-processing platform based on generative AI models was implemented. The platform is capable of reading and interpreting a wide variety of documents, applying advanced OCR, classifying documentation, extracting semantic information, and carrying out validations aligned with established business rules. The solution consolidates all the information required by technical staff into a single environment, enabling them to make decisions more efficiently while maintaining full traceability of every action taken.
At the same time, an agentic assistant was introduced to support administrative processing. Designed to handle internal queries, provide updates on the status of cases, and assist public employees in their day-to-day work, the tool serves as a practical aid rather than a replacement for staff. “It does not replace technical personnel under any circumstances,” explains Jimeno. “Instead, it acts as a first level of support that streamlines workflows, standardizes responses, and relieves teams of the most repetitive inquiries.”
The case illustrates a central theme running through SDG Group’s perspective: When AI is implemented with a clear purpose, high-quality data, and robust governance, “it becomes a genuine ally for public employees and a tangible accelerator of both efficiency and service quality.”
Microsoft, meanwhile, highlights its projects with several autonomous communities as evidence that generative AI has moved beyond the experimental stage. The autonomous community of Madrid, for example, has launched a generative AI training program for 170,000 public-sector employees using Microsoft 365 Copilot Chat, integrating these capabilities into tasks such as document drafting, information retrieval, and case preparation. Escobedo describes the initiative as a landmark in the modernization of regional government administration and positions it within a broader strategic partnership aimed at developing AI use cases for administrative processes and citizen services.
Andalusia is at a different stage of adoption, combining advanced analytics, predictive services, and generative AI on its Azure-based data governance platform. The objective is to anticipate citizens’ needs, optimize internal management, and equip public-sector employees with the skills to use these tools safely in documentation-intensive workflows. Galicia’s collaboration focuses on training 45,000 young people in AI-related skills as part of a broader strategy to strengthen the region’s technological capabilities and link AI adoption to economic growth and improved public services. Aragón, meanwhile, is advancing AI training initiatives within vocational education programs and social innovation projects associated with the arrival of new technological infrastructure, including data centers.
Salesforce’s examples are somewhat closer to the citizen-facing interface. The City Council of Barcelona has implemented the company’s technology solutions to improve service delivery within municipal social services, enabling more than 2,000 professionals to access system information more efficiently, manage resources more effectively, and improve communication with vulnerable citizens, while maintaining strong security and privacy safeguards. The company also highlights the active interest shown by the Community of Madrid in AI initiatives for public administration, including institutional visits to explore projects such as Agentforce and assess their potential to automate tasks and enhance the citizen experience.
Public receptiveness to these changes is a factor that Salesforce places at the center of the discussion.
A recent study by the company found that 95% of surveyed Spanish citizens would be willing to interact with AI agents in public administration, positioning Spain among the European countries most open to this type of solution. The finding points to an environment in which citizens not only accept technological mediation but actively demand it when they perceive that it makes their interactions with government services easier and more efficient.
Escobedo links this willingness to another key figure: Spain ranks among the six countries with the highest levels of active AI usage, with 41.8% of the population using AI on a regular basis, according to the report Global AI Adoption in 2025 — A Widening Digital Divide. “This level of maturity makes it easier for public administrations to integrate agents that automate repetitive tasks, improve interoperability between institutions, and free up time for higher-value functions that ultimately provide better service to citizens,” she argues.
Her observation highlights an important shift in context for public-sector IT leaders: the challenge is no longer introducing a technology that is unfamiliar to citizens, but rather aligning public services with a technology that has already become part of everyday life.
The impact of generative AI is not limited to internal productivity or the citizen experience. Looking ahead, all three spokespersons agree on the transformative potential of so-called agentic AI to reshape public services by 2026.
According to Escobedo, this period will bring “a qualitative leap forward,” with AI agents evolving from assisting with specific tasks to executing entire workflows—always under human supervision. These agents will help accelerate case processing, improve citizen services, and anticipate needs through real-time analysis.
Jimeno adds that such assistants will be able to “guide citizens through administrative procedures, answer routine inquiries, provide updates on the status of cases, and accompany individuals throughout their interactions with public administration.” In doing so, they will reduce friction and improve accessibility, while also serving as internal support tools that help organize workloads and facilitate access to institutional knowledge.
Galindo envisions a similar future, one in which by 2026 there will be “a far more mature adoption of agentic AI in public services, with agents capable of managing administrative procedures from start to finish, always under human supervision.” The condition for reaching that level of maturity, he argues, is that final decision-making authority remains with the administration and that recommendations remain fully traceable and auditable at all times, particularly in sensitive areas such as public grants, procurement, and the management of vulnerable populations.
The introduction of these autonomous capabilities inevitably raises questions about technological dependence and data sovereignty. While the three companies approach the issue from different angles, they share a common view: AI cannot be deployed in public administration without strong regulatory safeguards and effective control by government institutions.
For Microsoft, Escobedo identifies as a key priority ensuring that “public data remains under European control.” She points to the EU Data Boundary initiative, which guarantees that customer data is stored and processed exclusively within the European Union and in compliance with GDPR standards. Microsoft maintains that it is the first large-scale cloud provider to offer this level of data residency for the European public sector. The company further strengthens this framework through Microsoft Cloud for Sovereignty, which includes customer-managed encryption keys, immutable logs of technical access, and mechanisms designed to prevent international data transfers without authorization.
This architecture is complemented by specific commitments in Europe, including a European Security Program and a European Digital Resilience Commitment incorporated into contracts with EU member states and the European Commission. According to Escobedo, this framework legally obliges Microsoft to ensure the continuity of its cloud services within Europe, challenge any orders seeking to disrupt those services when there are legal grounds to do so, and activate contingency mechanisms with European partners to maintain operational continuity. She notes that these commitments align with Spain’s National Security Framework (ENS), the objectives of the Spain Digital 2026 strategy, and the inclusion of security services within the CCN-CERT CPSTIC catalogue, in addition to the appointment of a dedicated Chief Information Security Officer (CISO) for Europe.
Salesforce argues that its platforms have been designed “from the outset according to a responsible and secure AI approach” and that they comply with both European regulations—including the AI Act—and national strategies such as Spain Digital 2026. Galindo emphasizes that data sovereignty is ensured through full administrative control, end-to-end traceability, and deployment options tailored to the specific requirements of the public sector. To mitigate the risk of vendor lock-in, the company promotes open standards, interoperability, public-sector workforce training, and an operating model in which government institutions retain control and confidence throughout the entire lifecycle of their systems.
SDG Group approaches the issue of dependency from the perspective of governance and long-term project design. Jimeno warns that “the real challenge lies not only in adopting these technologies, but in how they are governed and evolved over time.” He stresses that the sustainable impact of agentic AI depends on its seamless integration into each administration’s technological ecosystem, its alignment with existing processes, and the preservation of established control mechanisms. Rather than advocating isolated deployments, he supports an approach in which AI becomes a flexible extension of existing systems—capable of adapting to regulatory and organizational changes without creating rigidity or excessive dependence on any single provider.
Transparency and explainability are another central element of the discussion, particularly when AI is involved in sensitive administrative decisions.
Microsoft structures its approach around a Responsible AI framework based on the principles of fairness, transparency, and human accountability. In practice, Escobedo explains, this translates into systems that incorporate controls to identify which data has been used, how it is processed, and which factors influence the outcomes, ensuring that “any AI-assisted decision can be audited and traced.” Tools such as Microsoft Purview and Content Safety capabilities help monitor model behavior, detect anomalies, identify potential biases, and strengthen the security and quality of AI-generated responses.
Salesforce advocates a similar model, with mechanisms for explainability, auditing, and human oversight, as well as systems that allow users to understand why a particular recommendation or response has been generated. Galindo emphasizes that biases are continuously monitored and that clear boundaries are established to ensure that “the final decision always rests with the administration, especially in sensitive processes.” This emphasis on human oversight is a recurring theme across all three perspectives and draws a clear distinction from narratives centered on full automation: in the public sector, AI is viewed as an assistant, not the ultimate decision-maker.
Ultimately, integration with existing infrastructure is the decisive factor in determining whether AI remains confined to pilot projects or becomes a foundational layer of government operations. SDG Group argues that AI “cannot be approached as an isolated layer or as a replacement for existing systems, but rather as a natural extension of each administration’s technological and organizational ecosystem.” Before introducing AI solutions, Jimeno explains, the company first develops a deep understanding of how each institution operates and how its processes function in practice. AI is then introduced gradually, connected to existing systems, and designed to leverage available information without creating duplication or disrupting established control mechanisms. According to Jimeno, this approach makes it easier for employees to view AI as a support tool rather than as an external or disruptive element.
Microsoft addresses integration through a cloud architecture designed to coexist with existing systems while meeting sovereignty requirements, supported by initiatives such as EU Data Boundary and Microsoft Cloud for Sovereignty. The Cloud Spain Central region, certified under the highest level of Spain’s National Security Framework (ENS High), enables public administrations to connect their data repositories, open-data portals, and platforms funded through the Next Generation EU program with AI services without having to completely redesign their existing infrastructure. This allows organizations to build on previous investments in data governance and digital modernization while adopting new AI capabilities.
Salesforce, for its part, emphasizes the importance of an open architecture and technologies such as MuleSoft, which make it easier to connect with sovereign data portals and legacy systems without having to replace infrastructure that is already functioning effectively. “Our solutions integrate easily with existing infrastructures thanks to an open architecture,” Galindo explains, stressing that this interoperability is essential if AI initiatives are to move beyond the pilot stage and become embedded in the day-to-day operations of public administration.
Taken together, these three experts paint a picture of rapid—but not uncritical—AI adoption across Spain’s public administrations. The priorities they highlight—measurable efficiency gains, improved public services, data-driven decision-making, data sovereignty, transparency, and human oversight—define an ambitious agenda for technology and innovation leaders. The challenge is to build a more proactive, citizen-centric administration while preserving the legal safeguards and control mechanisms that are fundamental to public service.
Original article in Spanish here.