DIANE GUTIW
AI INFRASTRUCTURE
Despite rising investment in AI, a critical implementation gap persists. According to CGI research, many organisations struggle to move AI projects from proofof-concept to production, stalled by policy and process gaps, legacy systems, persistent talent shortages to implement and maintain AI solutions as well as limitations in existing data foundations.
AI Magazine speaks with Diane Gutiw, PhD, Global AI Research Lead at CGI, to analyse this friction. Drawing on insights from CGI’ s 2025 Voice of Our Clients research, she diagnoses why enterprises are stuck and details the necessity of building robust, responsible AI frameworks, not as brakes, but as the essential engine for achieving scalable, competitive advantage.
Q. WHAT STOPS MOST COMPANIES FROM MOVING AI PROJECTS FROM PROOF-OF-CONCEPT TO PRODUCTION?
When we look to our 2025 Voice of Our Clients research – covering 1,813 discussions with 1,477 CXOs across government, banking, retail, manufacturing, energy, insurance, healthcare and communications – we see a significant momentum when it comes to AI: 36 % of the organisations engaged in the discussions are now implementing traditional AI and 26 % are implementing Gen AI.
Yet there’ s still a critical implementation gap for many organisations due to three barriers that constantly emerge.
DIANE GUTIW
TITLE: VICE-PRESIDENT, GLOBAL AI RESEARCH LEAD
COMPANY: CGI
INDUSTRY: IT AND BUSINESS CONSULTING
Diane Gutiw, PhD, is Vice-President and Global AI Research Lead at CGI, spearheading applied AI thought leadership and responsible AI standards globally, with a doctorate in Information Technology Management.
First, legacy systems and technology constraints affect 46 % of organisations. These enterprises are trapped in siloed systems that are not integrated, creating data fragmentation that undermines AI deployment whether you’ re running a bank, hospital or government agency.
While you can develop pockets of AI solutions or models you cannot scale AI and achieve real value with a patchwork infrastructure as, like human reasoning, the more information you have available the better decision you can make – gaps in information result in incomplete AI outputs and insights.
Second, is organisational challenges. Companies are missing the operating models, policy frameworks and governance structures required to manage AI risk and validate secure AI solutions.
202 January 2026