The gap between agentic AI ambition and agentic AI execution is now the defining strategic challenge for Global Business Services organisations. The ambition is not in question. Agentic AI has rapidly emerged as a top investment priority, cited by 65 percent of GBS respondents in the SSON research — far surpassing traditional automation. The execution is a different matter. Fobpossgbs
Fewer than 10 percent of companies have deployed agentic AI at functional scale. Just 6 percent of companies qualify as high performers where AI contributes meaningfully to EBIT — more than 5 percent — in a lasting way. Sixty-five percent want it. Six percent have materially benefited from it at enterprise level. Understanding that gap is more valuable than celebrating the ambition. SSON
What agentic AI actually is — and is not
The term is used with enough looseness in industry discussions to justify a brief clarification. Agentic AI systems differ from conventional AI applications in a fundamental way: they operate autonomously, executing multi-step workflows, making decisions, and adapting behaviour based on environmental feedback without continuous human intervention.
Previous waves of automation tackled parts of processes, leaving exceptions where humans had to step in. AI agents can reason, collaborate, and coordinate actions, allowing them to accomplish complex, multistep, nondeterministic processes that have so far depended on humans. Emagia
Bain's Technology Report 2025 crystallises the maturity model into four levels that are directly applicable to GBS planning. Level 1 is LLM-powered information retrieval — the copilot and knowledge assistant tier where most organisations currently operate. Level 2 is single-task agentic workflows. Level 3 is cross-system orchestration across silos. Level 4 is multi-agent constellations operating autonomously at enterprise scale. Most GBS organisations are between Level 1 and Level 2. The value that justifies the investment sits at Level 3 and above.
The three root causes of the execution gap
McKinsey's research is direct on why organisations are not closing this gap faster. Nearly two-thirds of respondents say their organisations have not yet begun scaling AI across the enterprise. The failure modes cluster into three categories. Punku
Fragmented processes. Agentic AI requires clean, standardised, well-documented processes to orchestrate across. An agent deployed on top of inconsistent country-by-country processes does not smooth those inconsistencies — it amplifies them. This is the most fundamental barrier and the one least often acknowledged in technology-led AI discussions. AI cannot compensate for fragmented data, unclear ownership, or broken processes. Attempting to deploy agentic AI without foundational process standardisation risks amplifying inefficiencies rather than resolving them. McKinsey & Company
Governance that cannot move at AI speed. In the age of agentic AI, organisations can no longer concern themselves only with AI systems saying the wrong thing — they must also contend with systems doing the wrong thing, such as taking unintended actions, misusing tools, or operating beyond appropriate guardrails. Traditional GBS governance — periodic reviews, paper-based controls, SLA reporting cycles — was not designed for autonomous systems operating continuously across multiple enterprise functions simultaneously. VentureBeat
Talent architecture misaligned with the new delivery model. Nearly a third of all companies struggle with AI-related talent and capability gaps, as well as problems integrating AI into existing systems. Nearly a quarter of top performers — compared with just 15 percent of other companies — cite change management as a core challenge to scaling agentic AI. The human-agent hybrid workforce requires different role designs, different performance management frameworks, and a different employee value proposition than the headcount-based models that GBS has historically operated. Bain & Company
The pragmatism principle
A purist view of architecture will not meet the moment. Given the current pace of AI innovation, companies should maintain an architectural North Star but sustain progress with fit-for-purpose, domain-specific, and human-in-the-loop builds for the foreseeable future. Libertify
This is the Bain position and it is worth sitting with. The organisations that are waiting for a clean, fully integrated, architecturally perfect agentic AI environment before deploying anything are watching their competitors compound gains with imperfect but operational implementations. The art is in selecting which processes to automate first — those with the highest volume, the clearest decision logic, and the most standardised data — and building operational experience and governance muscle while doing so.
The GBS-specific opportunity
Because shared services already manage standardised high-volume processes, they provide an ideal environment to test intelligent automation safely before scaling enterprise-wide. This shift elevates the role of GBS leadership. McKinsey & Company
This is the structural advantage that GBS organisations have over other enterprise functions attempting agentic AI deployment. The processes are documented. The volumes are sufficient for meaningful learning. The governance infrastructure, while imperfect, already exists. The organisations that exploit this advantage are building genuine AI deployment capability inside GBS — and then exporting that capability to the broader enterprise.
The ones that do not will find that AI deployment ends up distributed across business units, with no shared standards, no reusable infrastructure, and no organisational memory of what works. That is the fragmentation pattern repeating itself at a higher level of complexity. It is the GBS case in miniature — and it is being made in real time.
