2026 Decision Intelligence Benchmark — Special AI Report

The State of AI in L&D

How learning and development leaders are adopting AI — or not — to achieve their business objectives

Table of Contents

This report is interactive. Use the controls below to filter all charts by role level — or click All Roles to return to the overall view. Charts update instantly; hover over any bar for exact values. The written analysis reflects the overall view.

View by segment
Executive Summary

The AI Intent Is There – Now, Building Capability Is the Work

L&D leaders are approaching AI with unusual trust, clear ambition, and a pragmatic sense that the technology is already changing the function. But the strongest through-line is not that AI has arrived fully formed. It is that leadership expectation is ahead of team adoption, trust is ahead of proficiency, and dependence is forming before governance and enablement have fully caught up.

That makes this an L&D story in the deepest sense. The function is not only being changed by AI; it is responsible for teaching the organization how to change with AI. The opportunity is to model what disciplined adoption looks like: clear use cases, responsible review, role-specific practice, and a practical path from belief to capability.

What this means for leaders
Different Perspectives on Different Types of AI

To better understand how leaders think about different kinds of AI, we grouped tools into three categories used throughout this report:

Public Generative AI
Broadly available tools such as ChatGPT, Claude, Gemini, and similar public-facing generative AI platforms.
Enterprise Platform AI
AI features embedded in vendor-supplied systems, such as CRM tools, productivity suites, or other enterprise software.
Company-Owned AI
Proprietary, internally governed AI tools built, licensed, or configured specifically for the company or its teams.
Part 1: AI Adoption

L&D Leaders Expect Daily AI Use – Adoption Has Not Caught Up

L&D leaders have ambitious targets for how often their teams should be using AI. Reviewing the data on expectations, actual use, and the gap between them reveals where the function currently stands.

Highlights from the data
  • Daily use is the expectation across all three AI categories: 65% expect daily public GenAI use, 62% expect daily enterprise AI use, and 64% expect daily company-owned AI use.
  • Actual daily use is meaningfully lower than expectations: 48% report daily public GenAI use, 29% report daily enterprise AI use, and 21% report daily company-owned AI use.
  • Ironically, the gap is widest for the tools organizations control most directly: Public GenAI use trails expectations by 17 points, enterprise AI by 33 points, and company-owned AI by a sizeable 43 points.

Expectations of AI Use

L&D leaders have ambitious targets for AI adoption across all three categories assessed in the benchmark. Public GenAI, enterprise platform AI, and company-owned AI all land in the same narrow band of expected daily use, with roughly two-thirds of leaders saying their teams should be using each type every day. This consistency of expectations across categories is meaningful. Leaders are not treating AI as a narrow use case or a single-tool experiment. They are instead imagining a standard operating expectation across all three types of AI tools.

The data further reveals that senior-most leaders (EVP/SVP/VP) hold the highest expectations for public GenAI: all of them expect daily use, compared with 56% of directors and 56% of managers. On enterprise AI the pattern shifts — directors expect daily use most often (67%), ahead of managers (57%) and senior leaders (62%). The ambition sits highest where organizational visibility is greatest.

Expected Daily Use by AI Type
Q: What is the frequency of use you expect from your team for each AI type?
(N=39) · "N/A" responses excluded; share normalized to valid responses per AI type.
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026

Actual AI Use

Actual use tells a different story. Public generative AI is the leveraged most, with 48% reporting daily use. Enterprise platform AI sits at 29% daily use, despite being embedded in systems that organizations already own. Meanwhile, company-owned AI is lower still, with only 21% reporting daily use (sometimes limited to specialty use cases).

The more governed, integrated, or organization-specific the AI environment becomes, the more actual use trails leadership expectation. Public tools are easier to access, and teams are using them both at work and at home. Enterprise and company-owned tools require rollout, training, workflow design, and confidence on the part of users that the tool is worth using.

Actual Daily Use by AI Type
Q: What is the actual frequency of use from your team for each AI type?
(N=39) · "N/A" responses excluded; share normalized to valid responses per AI type.
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026

The Gap

Gaps between expectations and actual usage appear consistently across all three types of AI, and, ironically, widen for the tools organizations have built themselves. Public GenAI shows a 17-point gap between expected and actual daily use. Enterprise platform AI shows a 33-point gap. Company-owned AI shows the most extreme divergence: 64% expected daily use versus 21% actual, a 43-point gap. Internally governed or company-specific AI may be where leaders see the greatest value, but it is also where teams appear furthest from the intended usage pattern.

The Gap: Expected vs. Actual Daily AI Use
Percentage-point gap between expected and actual daily use, by AI type
(N=39)
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026
Our Take

The root cause of the adoption gap is capability and culture, not tooling. L&D leaders have set a clear ambition for frequent AI use, especially in enterprise and company-owned environments, but those tools are not yet embedded in the way teams work. That gap should feel familiar: this community has seen before that access does not create adoption, and adoption does not create proficiency.

Three actions for leaders
Part 2: AI Dependence & Automation Outlook

L&D Is Already Reliant on AI — and Expects that Reliance to Deepen

L&D is meaningfully dependent on AI today — and the function has a stronger view than most leadership peers of how far that dependency will grow.

Highlights from the data
  • Senior-most L&D leaders report higher departmental dependence on AI: 44% of leaders VP and above say they would see moderate disruption if AI went away — compared with 21% of directors and 31% of managers.
  • L&D leaders generally anticipate more work becoming AI-powered: 15% of leaders overall (and 22% at the VP-and-above level) expect 40% or more of L&D’s work to be automated over the next 24 months — well above the rates we see in other functions we surveyed.

Dependence on AI

AI is not yet fully mature inside L&D workflows, but it is already present enough that most leaders would sorely feel its absence if removed.

Senior leaders report the deepest dependence: 44% of VP-and-above leaders say AI disappearing would create "moderate" disruption, compared with 21% of directors and 31% of managers. That may reflect senior leaders’ visibility into strategy, planning, content scale, and productivity expectations — the places where AI is beginning to reshape how work gets organized, and with an understanding of future realities for the business to which others may not have access.

Disruption if AI Disappeared Tomorrow
Q: If AI were suddenly unavailable tomorrow, how disrupted would your function be?
(N=39)
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026

AI Automation Outlook

As dependent as they are on AI in their workflows, L&D leaders are not predicting wholesale replacement. Most expect AI to automate less than 40% of the function’s work over the next 24 months. That said, a smaller cohort is more bullish: 15% of L&D leaders (and 22% at the VP-and-above level) expect 40% or more of the work to become AI-powered, above the cross-functional benchmark average.

L&D leaders are neither dismissive nor breathless — but are pragmatic in their outlook. They see AI changing the work — especially content production, curation, personalization, synthesis, and administrative lift — while still recognizing the judgment-heavy nature of learning strategy, facilitation, stakeholder influence, and culture change.

Percent of L&D Work Automatable by AI in the Next 24 Months
Q: Approximately what percentage of your function’s work could be automated by AI over the next 24 months?
(N=39) · "N/A" responses excluded; distribution normalized to valid responses.
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026
Our Take

L&D leaders are not saying AI can run the function, but they are also not treating it as optional. The practical implication is not to slow down — it’s to get deliberate. The function is already dependent enough that unmanaged AI loss would hurt; that’s an argument for governance and enablement now, not later.

Three actions for leaders
Part 3: Team Proficiency with AI

L&D Teams Are Capable but Not Yet Expert — and the Gap Varies by Role and AI Type

Across all three AI types, most L&D leaders rate their teams as (baseline) competent. Expert-tier proficiency is rare. And the gaps between senior leaders and managers on specific AI types point to uneven deployment and access across the organization.

Highlights from the data
  • Public GenAI has the most advanced users: 19% of teams reach advanced proficiency on public GenAI — higher than enterprise AI (8%) or company-owned AI (16%).
  • Advanced proficiency is rare and concentrated in public tools: 19% of teams reach advanced on public GenAI, versus 16% on company-owned and 8% on enterprise AI.

Proficiency Across AI Types

Across all three AI types, the L&D proficiency rating is broadly "competent" — teams are able to use these tools but have not yet developed the expertise to redesign work around them. Public generative AI shows the strongest "advanced" tier at nearly 20%, reflecting longer exposure and more self-directed practice. Company-owned AI and enterprise AI trail significantly, with advanced proficiency at 16% and 8% respectively.

Team Proficiency by AI Type
Q: How would you rate your team’s overall proficiency in using each AI type?
Beginner / No experience
Competent
Advanced
Expert
(N=39) · "N/A" responses excluded; rows normalized to valid responses per AI type.
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026

The role-level cuts reveal a more uneven picture. Directors rate their teams’ enterprise AI proficiency lower than either senior leaders or managers — 50% describe teams as “beginner,” compared to 22% of senior leaders and 21% of managers. Directors, often overseeing the rollout work of these tools, see more work ahead to achieve proficiency than their counterparts do.

While strong, L&D’s proficiency base may not yet be enough to support the daily-use expectations leaders are setting. Competent teams can experiment. Advanced teams can redesign. The gap between those two states is where most L&D organizations currently sit.

Our Take

Teams can say they trust a tool and use it occasionally — but competence scores show what the daily impact actually is. The beginner concentration in enterprise AI in particular points to a specific training gap that adoption mandates alone won’t close. L&D is the function best positioned to design the instructional solution to its own proficiency problem.

Three actions for leaders
Part 4: AI Trust & Mistrust

L&D Teams Trust AI More than Any Function Surveyed

L&D ranks #1 among all functions on AI trust. The mistrust drivers that do exist — output accuracy, data privacy, policy alignment — concentrate on public tools and ease considerably for enterprise and company-owned AI.

Highlights from the data
  • L&D leads all functions on public GenAI trust: 51% of L&D leaders express moderate or significant trust in public GenAI like ChatGPT, Claude, etc. — the highest of any function in this benchmark.
  • Enterprise and company-owned AI earn even stronger trust: Enterprise AI shows 69% moderate or significant trust, and company-owned AI shows 62% with zero reported mistrust across the cohort.
  • Accuracy and data privacy lead the concerns: For public GenAI, 77% cite data privacy and 67% cite inaccurate or hallucinated outputs — the top two drivers. Both ease sharply for enterprise and company-owned tools.

Trust Across AI Types

The trust picture is uniquely favorable for L&D. Public generative AI — the type that produces the most skepticism across other departments surveyed — earns a majority trust response here. Enterprise and company-owned AI perform even better, reinforcing the idea that L&D leaders are open to AI when it feels closer to the organization’s systems and governance.

Distinguished somewhat from the overall group, senior-most L&D leaders are the most skeptical. Among senior leaders (EVP/SVP/VP), 22% report significant mistrust of public GenAI — compared with 0% among directors and 6% among managers. That level of concern is practical: senior leaders carry more exposure to AI governance risk and more personal accountability for what goes wrong when public tools surface sensitive organizational data.

Trust by AI Type
Q: For each type of AI, what is your level of trust with it?
Significant mistrust
Moderate mistrust
Neutral
Moderate trust
Significant trust
(N=39) · "N/A" responses excluded; rows normalized to valid responses per AI type.
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026

Drivers of Mistrust

L&D teams are focused on the risks that most directly affect whether AI can be used responsibly in learning work: data privacy and security, inaccurate outputs, and bias. These concerns concentrate on public tools and ease considerably for enterprise and company-owned AI. Zooming in on senior leaders, we see mistrust concentrated around public GenAI and its potential conflicts with internal policies and data security risks.

For L&D, those concerns are not abstract. Learning content carries organizational values, compliance expectations, role-specific guidance, and employee experience implications. A flawed output can be more than inconvenient; it can become bad guidance at scale.

Drivers of AI Mistrust — Public GenAI
Q: When you have trust concerns with Public GenAI, what are the primary reasons? (select all that apply)
(N=39) · Multi-select; denominator is total respondents. Reflects drivers cited for Public GenAI specifically.
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026
Our Take

Trust and mistrust in this data are not opposites. L&D leaders can have moderate trust in a tool and simultaneously cite near-universal concerns about its output accuracy. That is a realistic picture of what responsible L&D AI use requires. Moving AI behind the firewall reduces some risks but does not eliminate the need for human review, quality standards, and governance. The function’s job is to design workflows that honor both sides of that equation.

Three actions for leaders
Part 5: How L&D Compares Across Functions

L&D Stands Out in the Benchmark in Its Advanced Embrace of AI

Compared with other functions, L&D stands out on trust — #1 of all functions on public GenAI trust. Where it trails is capability: advanced proficiency sits in the teens across all three AI types, and actual daily use runs well behind the function’s own expectations. We score every function on the same six AI vectors, each measured the same way, then plot L&D’s distance from the cross-functional average — bars to the right mean L&D leads its peers.

Highlights from the data
  • L&D ranks first on public GenAI trust: 51% express moderate or significant trust — the highest of any function and nearly double the benchmark median.
  • L&D also ranks high on AI dependence: 34% of L&D leaders report moderate or major disruption if AI disappeared — above benchmark average across peer functions.
  • L&D ranks high on automation potential: 15% of leaders see 40%+ of the function’s work becoming AI-powered in 24 months, above the 10% benchmark average.
Where L&D Stands Out on AI
L&D’s distance from the cross-functional average on six AI vectors, aggregated across all three AI types
L&D ahead of average
L&D behind average
The benchmark average is the mean of the other ten functions surveyed: Social Media, Healthcare Social Media, ESG & Sustainability, CSR & Social Impact, Talent Marketing, Data Privacy, DEI, Data Strategy, Supply Chain, and Employee Experience. Each vector is the percent of valid responses on a standard benchmark question, averaged across the three AI types where applicable. Hover any bar for L&D’s score, the benchmark average, and L&D’s rank.
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026 (N=39 L&D; 11 functions, cells n≥5).

Cross-Functional Position

Most peer functions show a more cautious AI posture. They may use AI heavily, but they often pair that usage with skepticism, especially around public tools. L&D is different. The function shows a broader trust base across public, enterprise, and company-owned AI — and that trust advantage is reinforced by a higher-than-average sense of current dependency and a more expansive view of what AI can do for the function over the next two years.

That position is an advantage, but it also creates pressure. Trust and dependence can accelerate adoption, but they can also get ahead of competence and governance. L&D’s benchmark position suggests a function ready to move — provided leaders invest in the skills, standards, and workflow changes required to make that movement productive.

Our Take

L&D has bought into AI more fully than most peer functions — on trust, on dependency, and on automation outlook. That is a meaningful signal for a community responsible for helping organizations learn, adapt, and build capability. But trust and dependence are only the starting line. The next phase is whether L&D can turn that posture into disciplined use: enough structure to manage risk, enough practice to build fluency, and enough specificity to make AI part of the work rather than an idea hovering above it.