How organisations are adopting AI individually but failing to collaborate with it, and what the evidence says about closing the gap.
Preliminary Findings from the Global AI at Work Survey 2026
With supporting data from the CambrianEdge.ai Platform Behavioural Study
By CambrianEdge.ai Research Team
.png)
Every consequential technology eventually forces a conversation not about what the technology can do, but about what we want it to become. Artificial intelligence has arrived at that threshold.
The research presented in this report and our panel discussion today speaks to a challenge I have observed first-hand. My team and I have been among the early users of AI-native collaboration platforms, and the findings here mirror our own experience: the technology is ready. It is the organisational architecture around it that has not kept pace.
What strikes me most is the gap this study identifies. AI is moving faster than our ability to absorb it into how teams actually work. If we do not address this through infrastructure, through policy, through deliberate organisational design, we risk reducing a transformative technology to a collection of individual tools.
The conversations we must have now, in Parliament and in boardrooms alike, must ensure that as we shape the future of AI, intelligence remains human. The collaboration, the judgment, the standards. These are ours to build.
I am pleased to host this discussion at the House of Lords and commend this research to all who believe that getting AI right is not a technology problem. It is a human one.
Every major technology shift teaches the same lesson. The tools arrive first. Transformation follows, but only when organisations redesign the work itself, not just add new instruments to old workflows. Sixty-nine percent of businesses now use some form of AI. But over 80% report no meaningful impact on productivity.
This paradox, widespread adoption with limited results, is not a technology failure. It is an organisational design failure. And it has a name: the collaboration gap.
This report presents findings from CambrianEdge.ai's AI at Work: Global Survey, conducted in June 2026 across professionals in 14 countries, supported by behavioural data from 775 users on the AI-native platform, CambrianEdge.ai. Its central finding is direct: the bottleneck is not AI adoption. It is AI collaboration, the inability of organisations to move from individual experimentation to structured team workflows where humans and AI operate as a continuous system.
Among 136 survey respondents, 55% identify their biggest AI challenge as either solo use or no structured breakdown between human and AI work. 62% have no defined process for handing off AI-generated work to human review. 27% report having zero collaboration infrastructure.
BCG's 2026 survey of 300 global CMOs found that 96% claim AI is driving end-to-end transformation of their marketing function, yet 42% still use AI only to assist humans with discrete individual tasks. BCG calls this the "transformation illusion." We call it the collaboration gap.
Yet the data also reveals a clear path forward. Organisations with all five infrastructure layers report significant impact at three times the rate of those with none. Behavioural data from 775 platform users confirms: 98% engagement, with 44 organisations actively using AI across three or more business functions.
The message for policymakers and business leaders is straightforward: the AI tools work. The organisational infrastructure around them does not. Closing the collaboration gap is now the primary lever for unlocking AI’s economic potential.
A workforce actively experimenting, but not yet crossed to organisational integration. Just over half describe their AI maturity as integrating or rebuilding.
The survey reveals a workforce actively experimenting but not yet crossed to organisational integration. Just over half (51%) describe AI maturity as "Integrating" or "Rebuilding." 49% are still starting or experimenting.
136 respondents. Categories are self-reported. "Integrating" = using AI across multiple workflows with defined processes. "Rebuilding" = redesigning core workflows around AI.
Claude (72%) and ChatGPT (71%) lead adoption. This fragmentation is itself a finding: when teams use different tools without shared infrastructure, the collaboration problem compounds. Every handoff between tools is a handoff where context, quality standards, and institutional knowledge can be lost. Tool selection alone does not predict outcome quality, infrastructure does.
Outcome quality correlates with infrastructure depth, not tool selection. The same tools produce different results in different organisational contexts.
AI adoption is stalling at the collaboration layer. The biggest challenges are not about technology, they are about how people share AI work with each other.
The survey confirms what many of us have observed in the field: AI adoption is stalling at the collaboration layer. Organisations have invested in tools. They have not invested in the connective tissue that turns individual tool use into organisational capability.
The top two challenges, "no real breakdown between AI and human work" plus "used solo, not shared across team", account for 55% of all respondents.
say their biggest challenge is solo use or no structured breakdown between human and AI work. This is not a skills problem. It is an organisational design problem.
Structured handoffs, defined processes plus shared systems and workflows, produce 71% significant outcomes. Without them, only 38%.
Each method shows prevalence (% of all respondents) alongside % reporting significant outcomes. Defined handoff: 29 respondents. Shared systems: 22. Informal: 40. Still figuring out: 26. Solo: 19.
"62% of organisations have no defined process for handing off AI work to human review. The bottleneck is not the tool, it is the transition."
Five infrastructure elements predict outcome quality with striking consistency. Each additional layer moves the needle. All five together, 100% significant impact.
Five elements make up the AI collaboration infrastructure: shared tools access, training programmes, prompt libraries, quality standards, and review processes. Each additional layer measurably increases the proportion reporting significant impact.
Consider the analogy: AI today is like electricity in the 1920s. Some organisations are still lighting candles. Others are redesigning entire cities. The infrastructure gradient in this data tells us exactly where that line falls. BCG's independent research with 300 global CMOs reaches the same conclusion from a different direction: "The differentiator is operating infrastructure", not tool selection, not budget, not stated ambition.
The correlation between infrastructure depth and significant outcomes is strong and statistically robust. 136 respondents.
This is not a marginal effect. Each additional infrastructure layer is associated with a measurable improvement in outcomes. Organisations with three or more layers report zero instances of "No change", meaning that at sufficient infrastructure depth, AI reliably delivers at least moderate impact. The candles are out. The electricity is working.
BCG's CMO survey found that the organisations pulling ahead are investing in data foundations, workflow orchestration, and talent development, while 42% remain stuck at the individual-task level. Our data quantifies this with statistical precision. Their data confirms it at enterprise scale.
% of respondents reporting each infrastructure element in place. "None of the above" = 27%. Multiple elements possible per respondent.
The most basic element, shared tools access, is present in fewer than half of organisations. Review processes, the most mature infrastructure layer, exist in fewer than one in four. For policymakers considering workforce AI readiness, this infrastructure gap is the actionable metric.
"27% of organisations have zero AI collaboration infrastructure. Not a single shared tool, training programme, or quality standard. They are building on sand."
AI-native collaboration platforms don't just improve outcomes, they change the nature of the challenge. The blocker shifts from "Should we use AI?" to "How fast can we move?"
Organisations that adopt AI-native platforms may differ systematically from those that do not. We report correlation, not causation, and state this caveat throughout. That said, the magnitude of the differences, and their statistical significance, warrants serious attention.
54 respondents (40%) use an AI-native collaboration platform. Across every metric, they outperform the rest by significant margins. The outcome gap, 80% vs 30%, is the largest single finding in the survey. Platform users are also twice as likely to have a defined handoff or shared system (52% vs 28%), and half as likely to report solo use as their top challenge (17% vs 29%).
Chi-square=29.52, p<0.0000001, Cramer's V=0.466. The effect size is large. This is not noise.
AI-native platform users: 54 (40% of respondents). Others: 82 (60%). Significant outcome = reported meaningful, measurable change in work quality or productivity.
Gap in “Significant outcome” between AI-native platform users (80%) and other respondents (30%).
For organisations on AI-native platforms, the primary blocker has already shifted from adoption questions to execution speed. Their peers are still debating whether to use AI at all.
Platform behavioural data from 775 users across 104 organisations, January through June 2026. When the infrastructure exists, people use it.
Survey data tells us what people say. Behavioural data tells us what they do. Between January and June 2026, we tracked usage patterns across 775 users in 104 organisations, spanning startups, enterprises, non-profits, law firms, educational institutions, and agencies across 39 locations in 15 countries. Three patterns emerge from this data that the survey alone could not reveal.
Pattern 1 — The engagement signal is strong. Behavioural data from the CambrianEdge.ai platform, January-June 2026, covers 775 active users across 104 organisations in 25 countries. 98% engagement rate after onboarding, meaning they returned to the platform after initial setup. When organisations provide AI collaboration infrastructure, people use it. This is not a trial-and-abandon pattern. 44 organisations are actively using AI across three or more business functions, cross-functional AI workflows of exactly the kind survey respondents identify as missing from their own organisations.
Pattern 2 — Use cases span the entire organisation. AI collaboration is not confined to a single function. Users are applying the platform across activities that would typically require coordination across multiple departments and tools.
User counts by primary function of AI usage. Multiple functions possible per organization. Top two functions account for 32% of all activity.
Pattern 3 — Multi-function collaboration is the leading indicator. 44 organisations are actively using the platform across three or more business functions. These are not individuals experimenting. These are teams building cross-functional AI workflows, the exact behaviour that survey respondents identify as missing from their own organisations. This is not limited to technology companies or startups:
775 users across 104 organisations. Startups & Entrepreneurs (26%) and Non-Profit Organisations (21%) are the two largest segments. Enterprise (5,000+) represents 4% of current users.
"98% engagement. 44 organisations using AI across three or more business functions. When the infrastructure exists, people use it."
18% of organisations have already reversed course on AI initiatives. This is not a failure rate, it is an early warning for the broader ecosystem.
18% of survey respondents have rolled back or abandoned at least one AI initiative. Among those who have not reversed course, outcomes vary widely: 32% kept everything with confidence, 35% have mixed results but are continuing, and 32% say it is too early to tell.
For the majority of organisations, the verdict on AI is still out, and the next 12 to 18 months will determine whether their current approach becomes a sustained capability or an expensive experiment that gets quietly shelved.
The rollback signal should concern policymakers as much as business leaders. Public investment in AI workforce readiness, digital skills, and innovation infrastructure risks being undermined if the organisational layer, the collaboration infrastructure, is not addressed alongside the technology itself.
The evidence points to concrete actions for policymakers and business leaders. The challenge is not adoption. It is design.
Workforce AI readiness is not about training people to use ChatGPT. It is about building organisational capability, shared access, quality standards, review processes, and structured handoffs, that turns individual AI literacy into collective AI fluency. The five infrastructure elements in this study provide a concrete, measurable framework for assessing readiness at the organisational level.
You do not have an AI adoption problem. You have an AI collaboration problem. And the data shows that solving it produces measurable, statistically significant improvements in outcomes. Three moves follow from the evidence:
The Global AI at Work Survey will continue through 2026, targeting 300+ respondents across more countries and industries. Platform behavioural data will be supplemented with token-level usage analytics to understand not just whether AI is used, but how deeply it is integrated into work itself.
BCG's survey of 300 CMOs and ours of 136 professionals independently identify the same bottleneck. BCG calls it the "transformation illusion." We call it the collaboration gap. The data converges: high-level strategic commitment to AI is not sufficient. The organisational layer, how AI work is shared, reviewed, and iterated, is where transformation actually happens or fails.
The collaboration gap is not permanent. It is a phase. Every major technology shift follows the same arc: tools arrive, chaos follows, then architecture emerges. We are now in the architecture phase. Organisations that understand this, and invest accordingly, will compound advantages that are very difficult to replicate once the window closes.
"The path from AI literacy to AI fluency is not a training programme. It is an environment."
First, the infrastructure must be built for collaboration from the ground up, not individual productivity bolted onto existing workflows, but shared systems where teams create, review, and deploy AI-generated work together, with structured handoffs, quality standards, and institutional context built into the platform.
Second, it must meet users where they are and move them forward. Our behavioural data shows 98% engagement across 775 users, including segments with no prior AI experience. The path from AI literacy to AI fluency is not a training programme. It is an environment, one where new users learn by doing within structured workflows, where quality standards are embedded rather than taught, and where every interaction builds organisational capability, not just individual skill.
CambrianEdge.ai was designed to address both of these requirements. It is an AI-native collaboration platform built for teams, where shared workspaces, structured handoffs, prompt libraries, and quality review processes are not add-ons but architecture. It adapts to users across the fluency spectrum, enabling organisations to onboard new users into productive AI collaboration from day one while continuously deepening capability for advanced teams.
The organisations that build the connective tissue first, the shared systems, the quality standards, the structured handoffs, the training programmes, will not merely use AI more effectively. They will compound their advantage over time. Because once the collaboration gap closes, what follows is not incremental improvement. It is a fundamentally different way of working.
The question is no longer whether AI will transform the work. It is whether organisations can transform themselves fast enough to meet it.
Global AI at Work Survey: 136 respondents, 14 countries, conducted June 14–22, 2026.
Beta Cohort Profile: 531 users across 152 organisations in 25 countries. CambrianEdge.ai beta programme, June 2025–June 2026.
CambrianEdge.ai Platform Behavioural Data: 775 user records across 104 organisations. Usage tracking January–June 2026.
CambrianEdge.ai • Gutenberg • Cambridge Central Asia Forum • US-India Strategic Partnership Forum • Stanford SEED • Digimentors • Brand Communion
* This is an ongoing global survey. The data presented here reflects 136 respondents collected between June 14-22, 2026, and will be updated as the sample grows toward 300+ respondents across more countries and industries. Findings should be treated as indicative patterns rather than definitive population estimates; approximately 67% of current respondents are based in India. For full methodology, raw data access, or research partnership inquiries, contact support@cambrianedge.ai.
© 2026 CambrianEdge.ai. All rights reserved. This report may be cited with attribution.