ActivTrak CEO: What 120,620 workers reveal about AI maturity

ActivTrak’s Productivity Lab (tracked 120,620 employees across 1,009 orgs from Q4 2025–Q2 2026) finds productivity/health metrics rise as AI use moves to regular task-level adoption, with healthy utilization peaking at 75%. However, when AI is embedded into day-to-day workflows, healthy utilization falls ~5 percentage points to levels statistically similar to low/no AI users, suggesting that “more AI/more maturity” may not maximize outcomes. Only 2% reach workflow integration (Stage 3) versus 27% using AI as research assistance (Stage 1) and 14% for task execution (Stage 2), reinforcing a cautious, right-stage rollout approach to avoid runaway costs and misaligned processes.

Analysis

This is less a broad “AI demand is good” signal than a budgeting signal: enterprises appear to get the highest ROI from workflow-adjacent assistance, not from pushing every seat into expensive, deeply automated usage. That favors software layers that sit inside existing processes and can show measurable time saved, while pressuring vendors that depend on rising token intensity, premium model routing, or end-to-end automation narratives.

The second-order effect is procurement discipline. If CFOs start capping AI use at the point where productivity plateaus, the revenue mix shifts from consumption-heavy spend to lower-cost, governed deployments. That is mildly negative for frontier-model/API economics and for RPA/agentic automation names whose bull case assumes broad workflow replacement; it is more constructive for MSFT and NOW-style embedded workflow platforms where AI is one feature, not the product thesis.

Near term, the market may underreact because usage stickiness sounds bullish, but the more important variable is monetization quality over the next 1-3 earnings cycles. Over 6-18 months, the message is that “AI attach rate” will matter less than “AI dollars per verified output,” which could compress multiples for companies selling aspiration and expand them for companies selling measurable labor substitution. A falsifier is evidence that enterprise customers are expanding AI spend without margin leakage or that premium-model usage keeps rising despite governance pressure.

Contrarian take: the consensus is missing that moderate adoption can be the equilibrium, not a waypoint. If that proves right, the biggest winners are not the loudest AI pure-plays but the boring workflow integrators that can capture productivity without forcing a platform rewrite.

Key Decisions for Investors

• Long MSFT / short PATH, 3-6 months: favor embedded copilots and task assistance over full automation; target 1.5-2.0x downside/upside asymmetry if procurement shifts toward ROI discipline.
• Add on pullbacks to NOW over 1-3 months: workflow software with measurable labor savings should hold budget better than standalone AI tooling; risk is only if AI attach rates fail to show up in billings.
• Trim exposure to premium AI consumption names on rallies and hedge with QQQ put spreads into the next two earnings seasons: the thesis is slower token growth and tighter model routing, not a collapse in AI adoption.
• Watch list, not trade: TGT and other large enterprises only become buyable on AI if they can show opex leverage in SG&A or store labor; absent that, the article argues for restraint rather than acceleration.
• Do not chase pure-play automation exposure until management proves workflow redesign, not just license expansion; a reversal would be a material rise in AI-led productivity metrics plus margin expansion in upcoming guidance.