How Salesforce is building an AI-fluent workforce: a practical playbook for HR
On her first morning back from the holidays, Salesforce chief people officer Nathalie Scardino opened Slack and asked AI to catch her up. Approvals, team updates, the signal from the noise-summarized in seconds. That's now normal at Salesforce, and the numbers back it up: 85% of employees report confidence using AI, a 16% year-over-year lift across a 70,000+ workforce.
Salesforce calls this "customer zero"-using its own people and workflows to pressure-test how AI adoption actually sticks. The big lesson: tools don't drive outcomes. How you design work, govern decisions, and upskill managers does.
Tools aren't the strategy-governance and workflows are
"Technology alone is not a strategy," says Scardino. "The true differentiator is how organizations shape their governance and integrate AI into core workflows." Global AI investment may reach the trillions by 2030, but spend without workflow change is just noise.
Salesforce started by redesigning how work gets done, with Slack as the "front door" for agentic AI. The company then codified its approach into the AI Fluency Playbook, launched Jan. 8, to guide the human side of adoption-confidence, habit-building, and clear decision rights.
The three AI fluency metrics that matter
- Engagement: Employee sentiment and willingness to experiment with AI. Do people feel confident and curious-or hesitant?
- Activation: Habitual use in daily workflows. Not once-a-month novelty, but consistent, practical use.
- Expertise: The blend of human, business, and agentic skills to orchestrate work across the enterprise.
As Scardino put it, learning is the meta skill now. Teams that learn quickly, win quickly.
Critical thinking before digital delegation
AI can draft, summarize, suggest, and even act-but judgment still sets the bar. "Human judgment is even more important," says Scardino. Greg Shewmaker of rPotential adds that constants like judgment, agency, and critical thinking will keep mattering as everything else shifts.
Pierre Matchuet of Adecco is blunt: the obstacle isn't tech, it's organizational clarity. What can be delegated to an agent? Where must humans stay accountable? What does "good" look like when an agent acts on your behalf?
Managers are the adoption engine
Salesforce's data is clear: employees with managers who model AI use are 22 percentage points more engaged. Daily users report higher productivity, better focus, and greater job satisfaction. If your managers don't demo how to work with AI, your adoption stalls.
What success actually looks like
Usage is an interim metric. Business outcomes are the point. Shewmaker puts it simply: Are outcomes better than they were previously? That's the scoreboard.
Salesforce is also seeing talent agility rise alongside fluency-50% of new hires in the past year were internal moves. New roles are emerging too, like senior director of agentic talent management and performance systems.
Your AI fluency playbook (put to work in HR)
- Pick a front door: Standardize on one place where AI meets work (e.g., Slack, your HRIS, or your service hub). Make daily use frictionless.
- Set a delegation matrix: Document what AI can draft, suggest, act on, and what stays human-only. Include escalation rules and audit trails.
- Instrument the three metrics: Pulse sentiment, track weekly active AI users, and define skill checkpoints for expertise.
- Manager-first rollout: Train managers to role-model prompts, reviews, and decision gates. Give them demo scripts and weekly use cases.
- Redesign high-volume workflows: Start with recruiting comms, policy FAQs, learning support, and reporting. Ship small, measure fast.
- Build judgment muscles: Train on verification, bias spotting, and "chain-of-thought" reviews. Require human sign-off where risk is higher.
- Create team rituals: 10-minute weekly "What AI did for us" share-outs. Highlight wins, failures, and prompt libraries.
- Governance that breathes: Clear data access, model usage rules, and red-team tests. Update guardrails monthly based on findings.
- Link to outcomes: Tie AI use to cycle time, quality, and employee experience-not vanity usage stats.
- Fuel internal mobility: Map emerging agentic roles, post stretch assignments, and make transitions visible.
30-60-90 for HR leaders
- Days 1-30: Pick two workflows, a front door, and a manager cohort. Launch sentiment pulse and a basic usage dashboard. Publish your delegation matrix v1.
- Days 31-60: Expand to five workflows. Run weekly share-outs. Add judgment training and spot checks. Start reporting outcome deltas.
- Days 61-90: Formalize governance updates. Publish playbook v1.1. Launch internal mobility paths for agentic roles. Tie AI habits to performance conversations.
Skills to prioritize
- Human skills: Judgment, critical thinking, ethical reasoning, decision framing, and accountability.
- Agentic skills: Prompting, task decomposition, review/verify loops, data context setting, and workflow orchestration.
- Business fluency: Translating policies, compliance, and KPIs into repeatable agent tasks.
How to measure impact beyond usage
- Cycle time: Days-to-fill, ticket resolution, comp review turnaround, policy update speed.
- Quality: Error rates, escalations avoided, candidate experience scores, employee satisfaction with HR responses.
- Manager behavior: Percent of managers demoing AI weekly, coaching moments logged, team adoption spread.
- Talent agility: Internal mobility rate, time-to-productivity in new roles, skill attainment velocity.
Address the fears, don't sidestep them
Salesforce's team tackles concerns head-on: jobs will change, accountability stays human, and the company will support reskilling. The outcome if you get this right? As Matchuet puts it, people stop managing tasks and start managing outcomes. AI handles volume and coordination; humans focus on judgment, relationships, and exceptions.
Helpful resources
- NIST AI Risk Management Framework - a solid baseline for governance and decision rights.
- Complete AI Training - courses by job to upskill managers and HR teams on practical AI workflows.
As Scardino says, leaders can't afford to get this wrong. Start small, measure what matters, and make managers the multiplier.
Your membership also unlocks:
