Consulting Guide: Talent Strategy

Planning talent when the roles keep moving

A talent strategy for AI-enabled operations has a problem an ordinary workforce plan does not. The role categories the plan depends on are still being defined. A team that writes a three-year headcount plan around "data analyst" or "customer support agent" in 2026 is planning against categories that will not mean the same thing in 2028. Automation absorbs part of the work, augments another part, and spins out new work that has no clean title yet. Plan against today's org chart and you lock in roles that taper; plan against a guess at the future and you hire for capabilities the business cannot yet use.

The discipline this guide teaches is to plan against the durability of the work rather than the label on the role. A mid-market operations group of 400 people typically splits into roughly 45 percent work that is durable regardless of AI adoption, 35 percent work that tapers as automation lands, and 20 percent emerging work whose demand is real but whose title is unsettled. The talent strategy sizes each band, decides build versus buy versus borrow for each, and ties the answer to a hiring and reskilling cadence the client runs after the engagement closes. The output is not a spreadsheet of open requisitions. It is a plan the operating team can defend when the categories move again.

The stakes are concrete. A 400-person org that misreads the bands by ten points in either direction is committing roughly $3.8 million a year of loaded cost to the wrong capabilities. Overweight the tapering band and you carry processing headcount the automation will strand. Overweight the emerging band and you buy expensive capacity a quarter or two before the business can use it. The banding discipline keeps that error small, and it is why this guide opens with decomposition rather than with a hiring target handed down from finance.

Sort the work into three bands, then size and source each

Start by decomposing the target organization into work packages rather than roles, tag each package as durable, tapering, or emerging, and estimate its FTE demand two years out. The worked example below is a 400-person operations org with a fully loaded average cost of $95,000. It shows how the bands change the sourcing decision and where the reskilling budget goes.

Read the table as a sourcing argument. The tapering band frees 55 FTE of demand; redeploying 40 of them into emerging work at $6,200 a path costs $248,000 against roughly $560,000 to buy the same 40 heads externally at a blended $14,000 hire cost plus ramp. The reskilling case is not soft. It is cheaper per capability moved and it retains institutional context that external hires spend two quarters rebuilding. The plan commits to redeploy-first for every emerging role where an internal candidate can reach proficiency inside two quarters, and reserves external buying for capabilities the business genuinely does not hold.

Four phases, three decisions, four artifacts

• Phase one, weeks one to two: produce the scope and evidence pack. Decompose the org into work packages, tag each durable, tapering, or emerging, and lock the framing decision in week one, naming the question the strategy answers and the boundaries it respects. Acceptance criterion: every FTE is assigned to a band.
• Phase two, weeks three to five: produce the recommendation memo. Size two-year demand per band, run the build-buy-borrow economics with real loaded costs, and make the evidence decision in week three, committing to the skills-gap data the recommendation will rely on before analysis starts.
• Phase three, week six: produce the operating cadence document. Design the hiring and reskilling cadence, quarterly capability reviews, and the redeployment pipeline. Make the operating-cadence decision in week six so the cadence is designed alongside the plan, not retrofitted at handoff.
• Phase four, weeks seven to eight: produce the handoff package. Hand the operating team a talent plan, a redeployment playbook per tapering role, and a scorecard that tracks capability coverage rather than seats filled.
• Throughout: hold each phase to its acceptance criteria. A phase does not advance until its named artifact meets them, which is what stops the engagement drifting into overlapping conversations that produce material without producing commitment.

Where talent strategies for AI operations fail

• Planning against role titles instead of work packages. Titles carry assumptions that will not survive. Fix: decompose to work packages and tag durability before you count heads.
• Treating the tapering band as pure attrition. Writing off 55 FTE ignores the redeployment economics, which favor reskilling at roughly half the cost of external hiring. Fix: run redeploy-first and reserve buying for genuine capability gaps.
• Buying emerging capabilities the business cannot yet use. Hiring eight prompt engineers before there is automation for them to own burns $200,000 of ramp for idle capacity. Fix: sequence emerging hires to land one quarter ahead of demand, not two years.
• Shipping a headcount spreadsheet with no operating cadence. A plan without a quarterly capability review ages out the quarter it ships. Fix: the week-six cadence decision makes the review a standing instrument the client runs.
• Measuring seats filled instead of capability coverage. Filled requisitions can coexist with unmet capability if the roles were the wrong ones. Fix: the handoff scorecard tracks coverage of each emerging capability against two-year demand.

Before you present the strategy

• Every FTE in scope is tagged durable, tapering, or emerging, with two-year demand sized per band.
• The build-buy-borrow mix for each emerging role carries real loaded costs and a redeploy-first test.
• The redeployment pipeline names candidate roles, target roles, and the two-quarter proficiency path for each.
• The operating cadence document specifies a quarterly capability review with a named owner and a standing agenda.
• The handoff scorecard measures capability coverage against demand, not requisitions filled.