Technology leaders predict generative AI focus will move from experimentation to production deployment, with emphasis on data quality and collaboration.
Major technology companies have spent two years implementing generative AI (Gen AI) across their operations, deploying AI models for tasks ranging from customer service to code development. These implementations have revealed challenges in data quality, security, and workforce adaptation that must be addressed as enterprises move from experimental deployments to production systems.
The rapid advancement of AI capabilities – particularly in large language models that power tools like ChatGPT – has driven widespread adoption across industries. However, enterprises face obstacles in achieving measurable returns on their AI investments. These include data quality issues, skills shortages and governance challenges that technology leaders expect to address in 2025.
Early adopters report that successful AI implementation requires fundamental changes to data management practices, security protocols and workforce development strategies. The focus is shifting from proving technical feasibility to demonstrating business value, with emphasis on practical applications that can scale across organisations.
ServiceNow, the digital workflow company, reports changing approaches to AI implementation across different departments and regions. “If 2023 was about learning and 2024 was about rapid experimentation, then 2025 is certainly the year of value realisation,” says Cathy Mauzaize, President of Europe, Middle East and Africa at ServiceNow.
Data quality emerges as key factor in AI success
Companies attempting to implement Gen AI face challenges with data quality and accuracy. Projects often struggle due to inadequate data preparation and training time. “Without good quality data, AI projects cannot succeed,” Cathy notes, emphasising a shift from proof-of-concept to proof-of-value approaches.
AI search technology, which can integrate structured and unstructured data from various sources like Workday and Microsoft Excel, offers potential solutions to data integration challenges. However, success requires clean data, governance frameworks and strategic implementation.
Transparency in data usage emerges as a priority. “To truly get value from AI, organisations need to bring together three things: the availability of clean, complete data; transparency around how that data is governed and used; and the right philosophy to get value from the work,” Cathy explains.
Zoom sees shift to AI-first user interfaces and workplace collaboration
Communications technology company Zoom predicts significant changes in how users interact with AI systems. Helen Hawthorn, Head of Solutions Engineering EMEA at Zoom, says “AI-first user interfaces will blend conversational AI with graphical user interfaces to create intuitive, personalised interactions that adapt to user behaviours and needs.”
The company also forecasts increased adoption of federated AI models, which combine multiple large language models (LLMs) – the technology behind Gen AI systems – to provide tailored solutions across platforms.
Helen adds that “AI agents will move beyond task automation to orchestrate actionable insights across entire organisations. These agents will focus on identifying inefficiencies, optimising workflows, and ensuring that critical actions are prioritised in real-time.”
The workplace transformation extends to hybrid working models. Steve Rafferty, Head of EMEA at Zoom, says “investment in reliable, high-performance technology and AI-driven collaboration tools will create more inclusive, flexible environments.”
Snowflake predicts mainstreaming of AI observability
Data cloud company Snowflake identifies AI observability – the ability to monitor AI system performance, accuracy and potential ethical issues – as a key trend. “I view AI observability as the missing puzzle piece to building explainability into the development process,” says Baris Gultekin, Head of AI at Snowflake.
The graphics processing unit (GPU) shortage, particularly acute in Europe, presents challenges for AI deployment. Regional data laws and security considerations complicate potential solutions like routing traffic to areas with greater GPU capacity.
Baris predicts that agentic systems – AI systems capable of autonomous decision-making – will emerge as a significant force. “2025 is when we will start seeing the hype of agentic systems start to bear fruit,” he says, “with the first set of high-value agentic use cases going into production – think handling customer service problems, identifying cyberthreats and project management.”
Cybersecurity concerns shape AI deployment
ServiceNow reports increased C-suite involvement in AI governance. “The deployment of AI introduces both operational risks and broader strategic, reputational and ethical concerns,” says Cathy, citing examples like algorithmic bias in recruitment tools.
Brad Jones, CISO and VP of Information Security at Snowflake, identifies new security opportunities and challenges. “As AI tools become more versatile and accurate, security assistants will become a significant part of the SOC [Security Operations Centre], easing the perennial manpower shortage.”
Concerns about data exposure through AI systems may be overstated. Brad notes that “people putting proprietary data into large language models to answer questions or help compose an email pose no greater risk than someone using Google or filling out a support form.”
However, new attack vectors are emerging. “I predict that we’ll start seeing patterns like attackers injecting themselves into different parts of the pipeline so that AI models provide incorrect answers, or even worse, reveal the information and data from which it was trained,” Brad warns.
Workplace transformation drives skill requirements
The integration of AI into workplaces requires new approaches to leadership and skills development. Steve Rafferty says “AI-native employees, those accustomed to integrating AI into daily life, will push businesses to fully embrace AI tools, reshaping workplace expectations.”
Developer roles face transformation as AI tools become more accessible. Jeff Hollan, Head of Applications and Developer Platform at Snowflake, notes that “as tools become increasingly powered by AI and easier to use, companies will no longer need highly-specialised talent to complete development projects.”
The shortage of AI-focused developers presents challenges. “There are simply not enough AI-centric developers to fulfill the increasing demand for deploying enterprise-grade AI applications,” Jeff explains. “Highly specialised AI developers also come with incredibly high salaries, limiting participation in the AI talent war to a small group of organisations.”
Data engineers will also face increasing demands. “As enterprises have started experimenting with AI, they’ve quickly realised that an AI system is only as valuable as the data that’s feeding into it,” Jeff adds. “Data engineering teams will be expected to either repurpose their existing tooling to work for AI-centric data pipelines and workflows, or to learn and implement new tooling.”
Management practices require adaptation to maintain productivity as AI deployment increases. “If a team becomes 15% more productive with AI, their manager also needs to become 15% more productive to keep up,” says Jeff. “It’s imperative that managers not only encourage AI adoption across their teams, but also ‘walk the walk’ themselves and keep up with team-wide productivity gains.”
