The conversation about AI and jobs tends to follow a familiar script. AI is coming. Roles will be eliminated. Workers should be afraid.
It’s the wrong frame entirely — at least for the delivery and program management functions that keep organizations running. The more likely outcome isn’t that AI eliminates these roles. It’s that AI eliminates the ceiling on how much they’re expected to handle. That’s a meaningfully different problem. And most organizations aren’t preparing for it.
The 10x math
Consider a company with 50 software engineers. AI-assisted development tools are already demonstrating significant productivity gains on coding tasks — GitHub’s Copilot research showed developers completing tasks 55% faster, and more recent agentic systems are demonstrating far greater leverage on full feature delivery. As the tooling matures, the working assumption inside many engineering organizations is a 10x productivity multiplier on the horizon.
When those gains arrive, the company doesn’t reduce its engineering team to 5 people. It expects output equivalent to 500. The headcount stays. The throughput expectations scale.
Now trace that math through the rest of the organization.
Every engineer-day of capacity needs a corresponding unit of requirements definition, design, testing, and launch readiness. If engineering capacity multiplies by 10, so does the demand on everyone supporting it. A team of 10 product managers supporting 50 engineers doesn’t shrink to 1 — it becomes a team of 10 expected to support the equivalent of 500. The same math applies across the delivery chain: product, design, QA, program management, and go-to-market.
This is the 10x problem. And it’s coming for every function surrounding engineering.
What changes for program managers
Program managers sit at the intersection of all of it — coordinating across engineering, product, design, and launch teams, managing dependencies, surfacing risk, and keeping complex programs on track. The volume and complexity of that work is about to scale dramatically.
The good news is that AI can absorb a meaningful share of the current PM workload. The tasks that are fundamentally about information processing and pattern matching are exactly what AI agents do well. For program managers, that list is longer than most people realize:
- Meeting notes and action items — AI can capture, summarize, and distribute decisions and next steps in real time
- Schedule generation and maintenance — AI can build and update project timelines based on inputs, dependencies, and team capacity
- Dependency mapping and cross-team impact analysis — AI can identify how changes in one workstream ripple across others
- Status report compilation — AI can aggregate data from multiple sources into coherent portfolio-level reporting
- Task completion follow-ups — AI can monitor open action items, send automated reminders, and flag items at risk of slipping
- Benefits identification and tracking — AI can surface expected outcomes from project briefs and track realization against them throughout delivery
- Cost modeling and tracking — AI can build spend projections, monitor actuals against budget, and flag variances before they become problems
- Risk register maintenance — AI can continuously scan project data, flag emerging risks, and update likelihood and impact assessments
- Resource allocation modeling — AI can analyze team capacity across programs and surface conflicts or gaps before they surface in a standup
- Stakeholder communication drafting — AI can generate routine updates, escalation notices, and executive summaries that PMs review and send
The cumulative time these tasks consume in a typical PM’s week is substantial. Automating them doesn’t just free up hours — it creates the headroom to operate at a fundamentally different level.
The work that remains distinctly human is different in character. Managing complex interpersonal dynamics. Sensing when a team is under-reporting risk. Reading between the lines in a stakeholder conversation. Building the trust that makes a difficult tradeoff decision land. Seeing around corners — synthesizing partial signals into a view of what’s coming before it shows up on a dashboard. These are judgment-intensive, relationship-intensive capabilities. AI can augment them. It can’t replace them.
The transition delivery teams need to make
The path forward isn’t to resist AI adoption or wait for a perfect enterprise rollout. It’s to aggressively test, iterate, and adopt the tools that handle the information-processing layer — then redirect the time and cognitive capacity that frees up toward the work only humans can lead.
In practice, that means:
- Audit your current workload. Separate tasks into “AI-automatable” and “human-essential” buckets. Be honest about how much of your day is information assembly versus judgment.
- Run real pilots. Test AI tools against actual work — meeting summarization, schedule generation, dependency tracking — not toy problems. See what holds up.
- Iterate fast. Adopt what works, discard what doesn’t. The organizations winning this transition aren’t running 18-month tool evaluations.
- Invest deliberately in the human-essential skills. Stakeholder management, risk intuition, cross-functional influence, and the ability to see around corners will become the primary value delivery functions bring. Develop them intentionally.
The organizations that make this transition fastest will find their delivery teams become force multipliers — able to support an order of magnitude more work without a proportional increase in headcount. The ones that don’t will find their delivery teams are the bottleneck in a world where engineering has already scaled.
The real question
The workers who should be worried about AI aren’t the ones whose jobs it might replace. They’re the ones whose organizations don’t move fast enough to retool before the demand curve arrives.
The 10x world is coming. The question isn’t whether your delivery team will survive it — it’s whether they’ll be ready for it.