Workers are lying less about AI—but they’ve developed a worse habit

WalkMe’s AI at Work Pulse Survey (n=2,037 U.S. adults) finds workers are less likely to fake AI competence: the share admitting they pretended to know AI fell from 45.2% (2025) to 28.3% (2026) and passing off AI-generated work dropped from 48.7% to 32.5%. However, only 24.6% say AI works on the first try and 50% report spending more time getting AI to do tasks than doing them manually—raising concerns that confidence is outpacing real productivity. WalkMe CEO Dan Adika argues employee “confidence” is not translating into P&L savings (he cites an implied $4–$5M/month savings gap in a 50,000-employee example) and warns that integration/security friction is likely causing “mega chaos” despite strong self-reported usage.

Analysis

This reads less like an AI demand acceleration signal and more like an ROI re-rating event. The market has been paying up for “AI attach” as if it were an immediate margin lever, but the bottleneck is implementation friction: permissions, workflow redesign, and governance slow conversion from demo usage to billable productivity. That argues for lower near-term multiple expansion across enterprise software names whose AI story depends on seat expansion or premium add-ons rather than control of the underlying workflow.

The relative winners are the platforms that already sit inside enterprise systems of record and identity/permissions. That favors MSFT and SAP structurally versus point-solution vendors, but only if they can prove that AI is embedded in actual workflows rather than sold as a standalone copilot. CRM is more exposed to disappointment because its AI monetization case is more dependent on discretionary budget approval and visible sales productivity, which are exactly the areas where CFOs will demand hard proof.

Over the next 1-3 months, the key catalyst is earnings guidance: investors should focus on AI attach rates, net retention, and any quantified productivity case studies, not user surveys. Over 6-18 months, the more important second-order effect is labor reorganization: if companies standardize agentic workflows, the economic value migrates from generic software licenses to control points around data, permissions, and orchestration. The contrarian point is that the consensus may still be underestimating replacement risk for junior knowledge work even as it overestimates near-term revenue conversion; that creates a choppy transition rather than a clean bullish AI spend cycle.

Key Decisions for Investors

• Stay underweight the AI narrative in enterprise software into the next earnings cycle; do not pay up for MSFT/CRM upside unless management quantifies AI revenue, retention, or margin contribution with numbers, not anecdotes.
• Relative value: long SAP vs short CRM over 1-3 months. SAP has the better moat if the real bottleneck is permissions and workflow integration; CRM is more vulnerable if buyers demand proof of sales-productivity ROI before expanding spend.
• Use any post-earnings strength in MSFT to trim exposure to the most AI-lifted parts of the software basket; the risk/reward is skewed toward multiple compression if AI attach rates are flat while expectations remain elevated.
• Watch-list alert: if a major enterprise vendor reports AI-driven savings greater than 5% of labor hours translated into budget cuts or headcount reduction, flip more constructive on the category; absent that, treat AI commentary as sentiment support, not fundamental upside.