Workforce scenario planning: modeling the impact of AI on roles

The question that started it

Priya Narasimhan, CHRO at a mid-market financial services firm, walked into the quarterly planning meeting with a question that had been circling the executive team for weeks: “Our CTO says we can automate 30% of customer service with AI. What does that actually mean for our people?”

It is the kind of question that sounds simple but is not. Thirty percent of what, exactly? Which tasks? Which roles are affected? Does 30% of tasks mean 30% fewer people, or does it mean the same people doing different work? And what happens to service quality, employee engagement, and institutional knowledge in each version of the answer?

This is where most organizations get stuck. They have a headline number from the technology team and a vague sense that “something needs to change,” but no structured way to model the options. The result is either paralysis or a blunt headcount reduction that ignores the complexity underneath.

Why the headline number is misleading

“Thirty percent automation” is a number that sounds precise but means almost nothing without decomposition. It could mean 30% of call volume is automatable. It could mean 30% of the tasks within each role can be handled by AI. It could mean 30% of roles become redundant. Each interpretation leads to a completely different workforce strategy.

This ambiguity is not unique to Priya’s company. Nearly every organization facing AI-driven transformation encounters the same problem: a technology assessment produces a single number, and the workforce planning team is expected to turn it into a people strategy. Without scenario modeling, the gap between the technology assessment and the people decision gets filled by assumptions, politics, and gut feel.

The scenario: customer service transformation

The customer service organization has 340 employees across three tiers. Tier 1 handles basic inquiries (password resets, account lookups, FAQ responses). Tier 2 handles complex issues that require judgment (disputes, exceptions, escalations). Tier 3 handles specialized cases that require deep product knowledge and regulatory awareness.

The CTO has identified that AI can handle most Tier 1 volume and assist with roughly half of Tier 2 cases. Tier 3 work is largely untouched by current AI capabilities.

Priya does not want a single recommendation. She wants to see the options.

How the agent approaches it

The Workforce Planning Agent starts with task decomposition, not role elimination. This distinction matters. A role is a bundle of tasks. Automating some tasks within a role does not necessarily eliminate the role. It changes what the person in that role spends their time doing.

The agent pulls the current role architecture for customer service: task inventories, time allocation data, skill profiles, performance distributions, and tenure data. It cross-references this with the AI capability assessment from the technology team.

Then it builds three scenarios, each with different assumptions about speed, scope, and strategy.

Scenario A: efficiency-focused

This is the model most executives instinctively reach for. Automate what you can, reduce headcount proportionally, and pocket the savings.

The agent models it out:

• Tier 1 workforce reduced from 160 to 45 (AI handles 85% of volume, humans handle exceptions and quality oversight)
• Tier 2 workforce reduced from 120 to 90 (AI assists but humans remain primary)
• Tier 3 unchanged at 60
• Net reduction: 145 positions
• Annual savings: approximately $11.2M in fully loaded compensation
• Implementation timeline: 12-18 months

But the agent does not stop at the cost line. It models the second-order effects:

• Institutional knowledge loss from departing Tier 1 staff, 40% of whom have 5+ years of tenure
• Engagement impact on remaining staff who watched colleagues get laid off
• Recruitment cost if AI capabilities plateau and you need to rehire
• Brand and employer reputation risk

The net present value looks attractive in year one. By year three, the model shows significant risk from knowledge erosion and reduced organizational adaptability.

Scenario B: redeployment-focused

Same AI implementation, different workforce strategy. Instead of reducing headcount, redeploy the freed capacity toward work that has been understaffed or not staffed at all.

The agent identifies redeployment opportunities by scanning open roles, strategic priorities, and skill adjacencies:

• 85 former Tier 1 staff retrained for proactive customer outreach (a function that does not currently exist)
• 30 former Tier 1 staff moved into AI training and quality assurance roles
• 20 Tier 2 staff promoted to Tier 3 (backfilling anticipated attrition and supporting growth)
• 10 staff moved into customer success roles in the B2B division, which has been understaffed for two quarters
• Net reduction: 0 positions
• Annual cost impact: approximately $2.1M in retraining and transition costs
• New revenue potential from proactive outreach team: $4-7M (modeled from industry benchmarks)

The second-order effects look different here:

• Institutional knowledge retained and redirected
• Engagement scores likely to increase (career growth narrative vs. layoff narrative)
• Organizational capability expanded into areas that were previously aspirational
• Employer brand strengthened, which impacts recruiting costs across all functions

Scenario C: phased hybrid

The agent generates a third option that blends elements of both, phased over 24 months:

• Phase 1 (months 1-8): Deploy AI for Tier 1, freeze Tier 1 hiring, begin reskilling programs
• Phase 2 (months 9-16): Natural attrition reduces Tier 1 by approximately 50 positions. Retrain remaining staff into new roles. Launch proactive outreach pilot.
• Phase 3 (months 17-24): Evaluate AI performance at Tier 2. Scale redeployment based on results. Adjust plan based on actual vs. modeled outcomes.
• Net reduction: approximately 50 positions (through attrition, not layoffs)
• Annual savings by year 2: approximately $3.8M
• New capability investment: proactive outreach, AI oversight, expanded Tier 3

This scenario trades speed for lower risk. It preserves optionality. If AI performs better than expected, more capacity gets redeployed to higher-value work. If AI underperforms, the organization has not eliminated the expertise it needs.

What makes this different from a spreadsheet

Any workforce planning team could build a spreadsheet with three scenarios. What the agent adds is not the math. It is the connected intelligence.

The agent pulls real skill profiles, not averages. It knows that 23 of the Tier 1 staff have customer success experience from previous roles, which makes them strong candidates for the B2B redeployment in Scenario B. A spreadsheet model would miss this entirely.

The agent models attrition probability at the individual level. It knows that 18 Tier 1 staff are within two years of retirement, which changes the math on Scenario C significantly. Natural attrition in that tier is higher than the baseline assumption.

The agent connects to learning catalog data. It knows that the reskilling path from Tier 1 to proactive outreach requires 6 weeks of training, not 12, because the existing staff already have 70% of the required skills. The retraining cost in Scenario B drops by $800K when you use actual skill gap data instead of assumptions.

And the agent tracks outcomes over time. Six months after implementation, it can compare actual results against the modeled scenario and recommend adjustments. This is not a one-time planning exercise. It is a continuous feedback loop.

Modeling the transition itself

Scenarios are not just about endpoints. They are about the path between where you are and where you want to be. The agent models the transition period explicitly, because the transition is where most workforce transformations fail.

For Scenario B, the agent modeled a 16-week reskilling period for the Tier 1 staff moving into proactive outreach roles. During those 16 weeks, customer service capacity drops by approximately 25%. The agent flagged this and recommended a mitigation plan: stagger the transitions in three cohorts of 28 people each, so that capacity never drops below 85% of current levels. It also identified that the AI system would need to absorb more Tier 1 volume during the transition, which accelerated the deployment timeline for the technology team.

For Scenario C, the transition model was simpler because it relied heavily on natural attrition, which is gradual by definition. But the agent flagged a risk: if attrition came disproportionately from the highest performers (which historical data suggested was likely), the remaining Tier 1 workforce would be less capable during the transition period. It recommended retention bonuses for the top quartile of Tier 1 staff during the 24-month phased period.

These transition details are invisible in a static scenario model. They only emerge when you model the journey, not just the destination.

The decision Priya actually made

Priya presented all three scenarios to the executive team. The CFO initially favored Scenario A. The CHRO and COO favored Scenario B. The CEO asked about Scenario C.

They chose a modified version of Scenario C, with one change: they accelerated the proactive outreach pilot from Phase 2 to Phase 1 because the revenue opportunity was compelling enough to justify the earlier investment.

The agent updated the model in real time during the meeting, showing the adjusted cost curve and capability timeline. No one had to go back and rebuild the spreadsheet.

The role of task-level data

The reason scenario planning works at this level of precision is task decomposition. Most workforce planning operates at the role level: how many customer service representatives do we need? But a role is a container for tasks, and AI does not automate roles. It automates tasks.

A Tier 1 customer service representative spends roughly 40% of their time on lookups and FAQ responses (highly automatable), 25% on basic troubleshooting (partially automatable), 20% on documentation and follow-up (automatable with supervision), and 15% on empathetic customer interaction during complaints and escalations (not automatable). Automating the 40% does not eliminate the role. It changes what the role looks like.

This is the insight that transforms the planning conversation. When Priya’s team looked at the task-level data, they realized that many Tier 1 employees were already doing Tier 2 work informally, handling judgment calls that technically should have been escalated but were resolved faster by experienced representatives who knew the answer. Those employees were not “Tier 1 workers.” They were mis-leveled. The AI transformation was also an opportunity to fix the role architecture.

Without task-level data, none of this would have been visible. The conversation would have been “how many people do we cut?” instead of “how do we redesign the work?”

Why this matters beyond one decision

Every organization is going to face this question. Not just for customer service, but for finance, legal, marketing, operations, and every other function where AI changes the task composition of roles.

The organizations that treat this as a headcount math problem will make blunt decisions and lose institutional capability. The organizations that model scenarios at the task level, connect them to real skill and attrition data, and track outcomes over time will make better decisions and build stronger workforces.

Workforce scenario planning is not a nice-to-have for the AI era. It is the core planning discipline. And it requires the kind of connected intelligence that no spreadsheet or static dashboard can provide.