Latest News | GBI Impact

Agentic AI is here. The leading edge is capturing most of the value

Written by Andy Williams | Nov 18, 2026, 11:30:59 PM

OVERVIEW

The agentic enterprise has gone from vision to reality, and the pace of the transformation has been striking.

For this year’s State of AI in the Enterprise report, Box surveyed 1,640 IT decision-makers across the United States, United Kingdom, France and Japan, who self-selected into four AI-maturity tiers: early stage, developing, advanced and leading edge.

In a year, the share of organisations describing themselves as advanced or leading-edge has risen from 8% to 64%, while the share calling themselves early-stage (or not started) has collapsed from 53% to 9%.

83% of organisations surveyed are running AI agents and 19% are doing so autonomously at scale. 80% report notable ROI — at least a 10% improvement — from their AI work to date.

“80% report notable ROI — at least a 10% improvement — from their AI work to date.”

At every maturity level, respondents agree on the same three things: agents need access to company content, governance is key and, with the speed things are moving, no one wants to get locked into any single AI tool or technology.

Organisations on the leading edge see markedly better results. Half of leading-edge companies say their ROI has been “significant” — an improvement above 25%. Only one in nine early-stage companies report the same.

So, what are the leading-edge companies doing right?

They are not the ones who understand AI differently. They are the ones who have built the infrastructure for the agentic enterprise — the operating muscle around agents: teams to deploy them, governance to control them, content to ground them and architecture flexible enough to keep upgrading as the AI stack changes.

1. The leading edge uses AI for new work, not just cheaper work

General-purpose assistance (44%) and helpdesk automation (42%) are still the most common use cases, and the leaders pursue those efficiencies too. What sets them apart is a second agenda the rest have not reached. 48% of leading-edge organisations — against 23% of early-stage ones — name doing existing work at far greater scale than was previously possible as a primary objective.

41%, against just 21% at the early stage, name doing entirely new types of work that were not feasible before: a salesperson reviewing a contract without waiting on legal, or a one-off internal tool that used to go unbuilt because no developer was free to build it.

2. AI is reshaping workforces, and might actually grow them

58% of organisations expect headcount to rise over the next three years, climbing to 79% among the most mature. Only 9% say AI agents are primarily eliminating roles today. The roles being added fastest — agent operators, governance professionals and workflow specialists — barely existed two years ago.

3. Context is this year’s AI bottleneck

96% of organisations say agents need access to company-specific content, but only 36% have connected agents to trusted content across many use cases. The 2026 challenge is not models; it is making enterprise knowledge accessible, usable and trustworthy for the agents that depend on it.

4. AI governance is essential, but is not yet widely implemented

Nearly half (49%) of organisations have already had an AI-related data exposure incident. In that context, the speed at which governance frameworks have been adopted is no surprise. The share reporting established or advanced frameworks rose from 24% in 2025 to 73% this year. The gap is in instrumentation. Only 39% have comprehensive visibility across sanctioned and unsanctioned use, only 34% have formal standards for how agents access company data, and 27% still call their governance ad hoc or developing.

5. AI leaders are betting on platform flexibility

68% of respondents are concerned about being locked into a single AI provider. Since our 2025 survey, the average number of AI tools officially adopted by respondents has increased to 3.3.

79% now say it is important or critical that agents operate “headlessly” — connecting directly to systems, APIs and data sources without a human interface. The leading edge is building so it does not have to bet on which tools win.

“The organisations seeing the strongest returns aren’t simply deploying more AI; they’re operationalising it differently.”

They integrate AI into multi-step workflows, connect systems to enterprise knowledge, formalise governance and deploy agents for new categories of work rather than standalone productivity gains for individuals. AI is becoming a core operational capability at the leading edge — a capability that generates measurable business impact today and is beginning to reshape the nature of work.

Content, governance and flexibility are the three foundations of the agentic enterprise, and are places where the gap between the leading edge and everyone else is wide. The rest of this report takes each in turn.

To read the full State of AI in the Enterprise report from Box, visit https://www.box.com/state-of-ai