2026 Decision Intelligence Benchmark — Special AI Report

The State of AI in CSR & Social Impact

How CSR and social impact leaders are adopting AI — or not — to achieve their business objectives

Table of Contents

This report is interactive. Use the controls to filter all charts by role level. Charts update instantly. Hover over bars to see exact values. Click All Roles to return to the overall view, which is the view reflected in the written narrative.

View by segment
Executive Summary

CSR Is the Least AI-Dependent Function in the Benchmark

CSR and social impact leaders are using AI — just not as much as the rest of the company.

Nearly all have some exposure, and about half report weekly or more frequent use of public generative AI. But on the measures that reflect depth of integration — daily adoption, automation outlook, dependence, proficiency — CSR sits below the cross-functional average. Respondents see AI as a useful assistant, not yet as infrastructure.

The data reveals a function that has not yet found AI use cases compelling enough to drive deeper usage. CSR teams describe their work as fundamentally human: relationship-driven, values-based, public facing, and dependent on trust. When asked about AI trust, data privacy emerged as a major driver, with trust rising for AI platforms under company governance.

Additionally, managers are outpacing senior leaders on AI use across all three AI types — the only function in the benchmark where that inversion holds consistently. That gap points to where AI is already taking root, and where the conversation about scaling it should start.

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

For the purposes of this report, and to better understand how leaders think about different kinds of AI, we grouped tools into three categories:

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, or organization-controlled AI tools built, licensed, or configured specifically for the company or its teams.
Part 1: AI Adoption

CSR Sets Modest AI Expectations — and Managers Use AI More Than Senior Leaders

CSR and social Impact leaders are not setting ambitious AI adoption targets, but an inversion of the trend observed in most other functions is notable and points of convergence also stand out.

Highlights from the data
  • Expected daily use is moderate across AI types: 50% of CSR leaders expect daily public GenAI use, 48% expect daily enterprise AI use, and 50% expect daily company-owned AI use — all meaningfully below the cross-functional average of 58% expected daily use.
  • Actual daily use trails expectations hardest on public GenAI: 20% report daily public GenAI use versus 50% expected — a 30-point gap. Enterprise AI shows a narrower 20-point gap (28% actual vs. 48% expected), and company-owned AI the widest: 8% actual vs. 50% expected.

Expectations of AI Use

CSR leaders hold modest expectations for AI adoption across all three categories. 50% expect their teams to use public generative AI daily, and a similar share expects the same of enterprise and company-owned AI. The consistency across types is notable — this is not a community that has high hopes for one kind of AI and skepticism toward others.

The seniority pattern cuts differently here than in most peer functions. Directors hold the highest daily use expectations for enterprise AI — 64% expect daily use, compared to 42% of managers and just 25% of senior leaders. CSR is the only function in the benchmark where managers outpace senior leaders on usage expectations across all three AI types.

Expected Daily Use by AI Type
Q: What is the frequency of use you expect from your team for each AI type?
(N=32) · “N/A” responses excluded; distribution normalized to valid responses per AI type.
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026

Actual AI Use

Actual use is significantly lower, particularly for public generative AI. Only 20% of CSR leaders report daily use of public tools — against a 36% cross-functional benchmark for actual daily use. Enterprise AI fares somewhat better at 28% daily use. Company-owned AI sits at 8% daily use, though with a much smaller response pool, this figure likely reflects the reality that most CSR teams have limited access to internally built AI environments.

The largest role-level divergence on actual use appears in enterprise AI. Managers report 45% daily enterprise AI use — notably higher than directors (18%) and senior leaders (0%). Managers are using the embedded tools in their operational workflows; directors and senior leaders may be working at a layer where enterprise AI features are less naturally integrated.

Actual Daily Use by AI Type
Q: What is the actual frequency of use from your team for each AI type?
(N=32) · “N/A” responses excluded; distribution normalized to valid responses per AI type.
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026

The Gap

The gap between expected and actual daily use is large for public GenAI (30 points) and even larger for company-owned AI (42 points). Enterprise AI shows a 20-point gap — still meaningful but narrower than most other functions.

One interpretation: CSR sees AI deployed broadly across the organization, but the function does not see emerging technology as critical to its future. The weekly-use reality for public tools (67% report weekly public GenAI use) suggests teams are engaged, just not on the frequency leaders imagined.

The Gap: Expected vs. Actual Daily Use
Percentage-point difference between expected and actual daily use, by AI type
(N=32)
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026
Our Take

CSR ranks 9th among 11 functions in expected daily AI use and last (11th) in actual daily adoption, placing the function below the cross-functional average on both measures. CSR leaders describe their work as inherently relational and trust-dependent, so they are not rushing to automate what they see as the human core of the function. The weekly-use signal indicates that this community is working with AI regularly, but it is far from integrated into operations. CSR leaders may want to spend some time identifying workflows that are ripe for automation and support from AI tools — and consult with managers on how they are already using them.

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

CSR Leaders Operate with Low AI Dependence and Are Pessimistic About Future Automation Opportunities

CSR leaders report the lowest AI dependence and among the lowest automation expectations of any function surveyed.

Highlights from the data
  • Lowest threat of disruption among the functions: 44% of CSR leaders say AI disappearing tomorrow would create no disruption to their work — the highest ‘business as usual’ response of any function in the benchmark. Only 19% anticipate moderate or major disruption.
  • Not much to automate: 81% of CSR leaders believe less than 20% of their function’s work could be automated by AI in the next 24 months — against a 63% cross-functional average for that lowest bracket. Only 6% expect 40% or more automation, compared to a 9% benchmark average.

Dependence on AI

The majority of CSR leaders say removing AI from their workflows tomorrow would cause minimal or no disruption — a reflection of what CSR teams report as the core of their work: stewarding relationships, securing executive alignment, managing external nonprofit partnerships, and translating program impact into business narratives. Those tasks sit in human judgment and organizational trust more than content generation or data processing.

Directors see 27% moderate or major disruption risk, appearing slightly more concerned than managers (13%) or senior leaders (20%). But no level reports significant dependence. CSR is the function where AI has not yet been deeply integrated into the operational layer and is not expected to any time soon.

Disruption if AI Disappeared Tomorrow
Q: If AI were suddenly unavailable tomorrow, how disrupted would your function be?
(N=32)
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026

AI Automation Outlook

CSR leaders see very limited automation potential in their function over the next 24 months. 81% believe under a fifth of their work is automatable — a more conservative read than nearly every peer function. Those who do see higher automation potential (the 6% expecting 40%+ automation) likely work in reporting and data management, where AI has more natural application.

Members note that they can see AI saving time on grant reports and volunteer communications, but when asked about their ‘real job,’ they default to the relational and strategic work that cannot be automated. The low automation outlook may partly reflect social desirability — leaders do not want to signal that their roles are replaceable — but it also reflects a genuine conviction about what CSR does that is not technology-driven.

Percent of CSR 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=32)
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026
Our Take

CSR sits below the cross-functional average on both AI dependence and automation outlook, which reflects the same underlying reality: this function has found meaningful places to use AI as a tool but has not yet built workflows where AI is load-bearing infrastructure. That distinction matters. Low dependence is not the same as low relevance. The CSR leaders who will be best positioned in two years are not those who automate the most — they are those who identify which parts of the reporting, measurement, and communications workflow AI can reliably own, and build governance around that ownership now, before dependence grows without structure.

Three actions for leaders
Part 3: Team Proficiency with AI

CSR Teams Are Learning to Use AI — But Advanced Proficiency Is Rare

Across all three AI types, CSR teams are almost entirely in the beginner-to-competent range. Managers rate their teams higher than senior leaders or directors do.

Highlights from the data
  • Competent, not advanced, is where CSR teams land: Across public GenAI, 91% of teams rate as competent or lower. Enterprise AI shows 91% competent or lower while Company-Owned AI is 100% either beginner or competent.
  • Senior leaders see the most beginner-concentration: On public GenAI, 80% of senior leaders rate their teams as beginners — compared to 43% of directors and 20% of managers. Senior leaders may be looking at team capability from the height of strategic ambition; what they see as beginner, managers are calling competent.

Proficiency Across AI Types

Public generative AI is where CSR teams have built the most capability: 9% reach advanced proficiency — still low by benchmark standards but ahead of enterprise AI (8%). Company-owned AI shows no advanced users at all among the CSR leaders who responded with a view into those tools. These figures reflect a function that has not yet had the deliberate practice or guidance around the potential of emerging technology to reshape their operations.

The company-owned AI proficiency gap is significant. With 46% of responses falling into beginner and none reaching advanced, it suggests AI is being adopted slowly and without strong enablement support on CSR teams.

Team Proficiency by AI Type
Q: How would you rate your team’s overall proficiency in using each AI type?
(N=32) · “N/A” responses excluded; rows normalized to valid responses per AI type.
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026

Senior Leaders See the Proficiency Gap Most Starkly

The role-level cut reveals an unusual pattern. Senior leaders are far more likely to rate their teams as beginners on public GenAI than managers are — 80% of senior leaders versus 11% of managers call their teams beginners.

Our Take

CSR ranks 10th among 11 functions on team proficiency — near the bottom of the benchmark — with 56% of teams rated competent or advanced on average across AI types. That position reflects the weekly-engagement reality: teams are touching these tools, developing familiarity, but not yet reaching the advanced tier that would enable workflow redesign. The senior leader ‘beginner’ read on their own teams is worth investigating. It may signal a genuine capability gap, or it may signal a mismatch in expectations. Either way, the function benefits from a shared definition of AI competence — what it looks like in CSR-specific practice, not just in general AI fluency.

Three actions for leaders
Part 4: AI Trust & Mistrust

CSR Leaders Hold Moderate Trust in AI

Trust in AI tracks with governance in CSR: lower for public tools, higher for enterprise and company-owned. But across all three types, a substantial share of leaders sits at neutral — neither trusting nor actively mistrustful.

Highlights from the data
  • 28% of CSR leaders express moderate or significant trust in public GenAI — placing CSR 6th among 11 functions on this metric, near the 29% cross-functional average.
  • Trust rises sharply for enterprise and company-owned AI: 47% express moderate or significant trust in enterprise AI, and 44% in company-owned AI — with zero reported significant mistrust in company-owned tools. Governance drives trust.
  • Data privacy leads all mistrust drivers: 72% of CSR leaders cite data privacy or security concerns for public GenAI. Inaccurate or hallucinated outputs (66%), policy misalignment (59%), and bias (59%) follow closely.

Trust Across AI Types

Public generative AI generates the most skepticism in CSR. More than a third of leaders express moderate or significant mistrust, and only 28% express moderate or significant trust. The 34% who sit at neutral represent a persuadable middle — a potential adoption lever.

Enterprise and company-owned AI earn more trust, with the trust-mistrust distribution shifting substantially once AI sits behind an organizational firewall. 47% of CSR leaders express positive trust in enterprise tools, and company-owned AI earns 44% trust with no significant mistrust at all. That pattern — higher trust the closer AI is to the organization — is a trend we see across our benchmark.

Trust by AI Type
Q: For each type of AI, what is your level of trust with it?
(N=32) · “N/A” responses excluded; rows normalized to valid responses per AI type.
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026

Drivers of Mistrust

Data privacy is the top mistrust driver for public GenAI in CSR at 72% — high in absolute terms but calibrated to the function’s exposure. CSR work involves external partners, grantees, community beneficiaries, and volunteers whose information cannot be freely entered into public AI environments. That constraint is practical.

Inaccurate outputs (66% public GenAI) and bias (59%) follow. Bias concern is particularly relevant for a function whose work is explicitly focused on equity, community representation, and social impact: a biased AI output in a grant recommendation or volunteer matching context carries consequences that go beyond efficiency. This may also explain the heightened concern around policy misalignment (59% for public GenAI).

Drivers of AI Mistrust
Q: When you have trust concerns, what are the primary reasons? (select all that apply)
(N=32) · Multi-select; denominator is total respondents.
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026
Our Take

Trust and caution coexist in this data in a way that reflects the mission of CSR and how profit or efficiency aren’t always the most important factors of this operation. CSR leaders who cite data privacy as their top AI concern are not anti-technology — they are protecting their function and the company’s reputation. The 34% ‘neutral’ block on public AI trust represents a group waiting for governance answers that do not yet exist at most organizations. Moving those leaders forward requires clear organizational policy on what data can enter which AI environments and visible demonstration that AI can serve CSR’s goals without compromising its values.

Three actions for leaders
Part 5: How CSR/Social Impact Compares Across Functions

CSR Uses AI Selectively, Not Extensively

The CSR and social impact function is relatively low on adoption, dependence, and automation expectations — and, by a smaller margin, on proficiency as well. Teams are learning the tools and using them weekly; they just haven’t embedded them into daily workflows yet, or closed the capability gap with peer functions.

Highlights from the data
  • CSR trails the benchmark on every AI vector measured: Expected usage ranks 9th, AI adoption last (11th), AI dependence 8th, AI trust 7th, automation outlook 6th, and team proficiency 10th among 11 functions.
  • Proficiency is not the exception it looks like at first glance: 56% of CSR teams rate as competent or advanced on average across AI types — 5 points below the 61% cross-functional average, and one of the function’s weaker relative rankings, not a near-average one.

Cross-Functional Position

CSR’s benchmark position is consistent across all six vectors: every one sits below the cross-functional average, with automation outlook and trust the closest to the benchmark and actual adoption the furthest. Expected daily use (49%) is 9 points below the cross-functional average of 58%. Actual daily adoption (19%) is 17 points below the 36% average — the largest absolute gap in the benchmark, and the lowest of any function surveyed. Dependence (19% moderate or major) is 8 points below the 27% average. Automation outlook (6% expecting 40%+ automation) is 3 points below the 9% average.

These gaps tell a coherent story. CSR is not failing to use AI — members describe regular tool use, particularly public generative AI, for research, drafting, and summarization. The function has simply not reached the stage where AI is woven into how work gets done at the daily level, and leaders are not yet expecting it to be. That is different from resistance, and it calls for a different response than mandates or urgency campaigns.

Proficiency follows the same pattern as the other vectors, rather than breaking from it. At 56% competent-or-advanced, CSR sits 5 points below the 61% cross-functional average — a real gap, though a smaller one than on adoption. Teams that are regularly engaging with AI, even weekly, accumulate familiarity faster than teams with no access or leadership discouragement. CSR’s weekly engagement is laying groundwork for capability, even though it hasn’t yet closed the gap with peer functions.

Where CSR/Social Impact Stands Out on AI
CSR’s distance from the cross-functional average on six AI vectors, aggregated across all three AI types
Ahead of average
Behind average
The benchmark average is the mean of the other ten functions surveyed: Social Media, Healthcare Social Media, ESG & Sustainability, L&D, 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 CSR’s score, the benchmark average, and CSR’s rank.
Source: Assemble’s Decision Intelligence Benchmark — Q1 2026 (N=32 CSR/Social Impact; 11 functions, cells n≥5).
Our Take

CSR’s benchmark position reflects a function in early integration rather than one disengaged from AI. The weekly-use pattern, the early signs of proficiency building, and the strong intent to invest in impact measurement create the conditions for a meaningful shift in the next 12 to 18 months — but only if that shift is directed at specific, mission-relevant use cases rather than general AI adoption targets. The function does not need to become the most AI-dependent in the benchmark to realize AI’s value. It needs two or three workflows where AI demonstrably improves the quality, speed, or scale of CSR’s impact — and the organizational confidence that comes from getting those right.

Three actions for leaders