Companies are adding AI agents to teams faster than they are adding people. That changes what a workforce plan has to cover. Someone has to decide where agents fit, what they cost, and how they change hiring. Right now that decision is happening without HR in most companies.
Karina Bernacki is VP of People at CoreWeave, the company building the infrastructure that runs the AI economy. She joins Tushar Makhija, founder and CEO of TeamOhana, to walk through how they do it and what to put in your own plan before next year's budget locks.
Attendees will learn:
- How AI agents change the math on headcount, capacity, and cost
- Why token spend belongs in the workforce plan, and how to read it as a planning signal
- The decisions People and Finance need to make together before budget season
- How to bring the CFO a plan for the human + AI mix before Finance builds one for you
Key Highlights
Every company is now an employer of two workforces, and only one of them has a plan. Salaries get budgeted a year ahead. AI spend shows up on the bill after the agents already did the work. Karina Bernacki, VP of People at CoreWeave, and Tushar Makhija, Founder and CEO of TeamOhana, get into what changes when AI becomes a second workforce: how to classify AI spend so it can be planned, why HR belongs in the room with Finance and IT before budget season, and how to build a hire-versus-deploy decision that leaders can actually defend. The conversation is practical, opinionated, and grounded in what CoreWeave is doing right now.
What You'll Learn
- Why AI spend belongs next to payroll on the workforce plan, not buried in software
- How to classify AI spend into personal enablement, copilots, and digital workers so each type gets budgeted correctly
- The workforce questions Finance and IT will not ask on their own, and why HR has to bring them
- How to structure a digital worker as a stack of agents that maps to a real role and cost comparison
- What to do about early-career hiring and succession when AI absorbs the work that used to build a bench
- The starting position to take into 2026 planning when no one has the right answer yet
Takeaway 1: Capacity Is the Right Conversation. Headcount Is the Easy One.
The old CEO question was "what's your headcount." The new one is "what capacity do we need to hit our goals, and where is it going to come from." That shift sounds subtle. It is not. It changes who gets to plan the workforce, what a role is worth, and how you defend a hire when an agent could do part of the job. Capacity conversations are harder because they force tradeoffs between humans, contractors, and agents inside a single envelope. Headcount conversations skip the tradeoff and produce a number no one can explain later.
Takeaway 2: Workforce Shape Happens by Default If HR Is Not in the Room
Finance will see the AI bill. Engineering knows what the tools deliver. If HR is not in the conversation, the hire-versus-deploy call becomes a procurement decision made by people who were never asked to think about it as a workforce question. The workforce still gets shaped. It just gets shaped by whoever is closest to the invoice. That is how companies end up with gaps in career pathing, broken spans of control, and no bench two years from now.
Takeaway 3: A Digital Worker Is a Stack of Agents That Maps to a Role
Karina's Ask HR example is the pattern. One agent handles intake in Slack. Another asks follow-up questions. Another routes to the right team. Another learns from repeat questions and flags gaps. Another drafts documentation updates. Collectively that is an IC2 role. Once you can see the stack that way, you can compare its cost against an FTE or contractor and make a real headcount decision. A single copilot that makes one person faster is a different animal, and you should budget it differently.
Takeaway 4: Take a Stake in the Ground Now, Even When the Numbers Are Wrong
The first year of planning AI spend as workforce spend will produce imperfect numbers. Adoption curves are uneven. Historical spend understates future need. The instinct to wait until the data is clean is the wrong one. Karina's approach at CoreWeave: give every API key an owner, classify the spend, set an envelope at roughly the 80th percentile of usage, and treat the first year as learning, not policing. The teams that get this right in 2026 will be the ones who started measuring in 2025, even when the measurement was rough.
"If HR is not in this conversation, and if it is only being driven by these bottom line numbers, this usage number, these spend numbers, the workforce shape is going to happen by default, not by intention. And that's where the negative outcomes can happen. Finance is going to see the AI bill. Engineering knows what it's delivering. If we are not involved, the hire versus deploy call is going to become a side effect of a budget conversation. It's going to be a procurement decision made by people who were never asked to think about it as a workforce question." — Karina Bernacki, VP of People at CoreWeave
