Workforce planning is the discipline most exposed to what AI is doing to work, and it is the one that has changed least. Most organisations still plan the workforce once a year, in a spreadsheet, as a headcount number attached to a budget line. That model assumes work is done by employees, that jobs are stable units, and that the only levers are hire, fire and freeze. None of those assumptions survives contact with a workforce in which contractors, specialist partners and AI agents now carry real execution-layer work.
The argument was put sharply on 1 September 2026 by two SAP executives, David Imbert and Lara Albert, in a VentureBeat piece whose title this article borrows: AI is redefining the workforce, and most planning models aren’t ready. Their evidence is three numbers from SAP’s own research. 62% of C-suite executives are dissatisfied with how well people data connects to business performance. 50% of organisations are planning for AI’s impact on productivity and capacity. Only 21% are planning for its impact on job design and organisational structure.
Those figures describe a gap between what leaders expect AI to change and what their workforce planning is set up to model. This article looks at where that gap comes from, what independent data from Stanford, Gartner, Microsoft and TalentNeuron says about the pace of the shift, and what a workforce planning model that can actually cope looks like. It builds on our earlier piece on AI workforce transformation and why redesigning work beats cutting headcount, and on the AI readiness assessment most businesses should run first.
Table of contents
- Why Workforce Planning Models Fail When AI Redefines the Workforce
- The Three Numbers Behind the Workforce Planning Gap
- What “Workforce” Now Means in Workforce Planning
- The Evidence That AI Is Redefining the Workforce Right Now
- Fragmented Systems: Why HR, Finance and Procurement Cannot See One Workforce
- From Annual Headcount to Continuous Workforce Planning
- Counting AI Agents in Workforce Planning
- The CFO and CHRO Partnership Workforce Planning Now Requires
- A Practical Workforce Planning Reset for UK Businesses
- Workforce Planning FAQ
- References
Why Workforce Planning Models Fail When AI Redefines the Workforce
The traditional workforce planning cycle was designed for a world in which the unit of work was a job, the unit of cost was a salary, and the horizon was a financial year. All three are now wrong often enough to matter.
The job is no longer the unit of work
AI does not automate jobs; it automates tasks, and the tasks inside a single role automate at very different rates. A claims handler, a paralegal and a junior developer each keep some tasks that need human judgement and lose others to software. A workforce planning model that counts roles cannot see that, so it either overstates the saving (by assuming the whole role goes) or misses it entirely (by assuming nothing changes until the role is cut).
TalentNeuron’s September 2026 study of seven large employers found that 34% of roles flagged for elimination still contained tasks requiring human judgement critical to strategy — exactly the error a role-level model makes.
The salary is no longer the unit of cost
Human capacity is a fixed cost that arrives in annual increments. AI capacity is a variable cost that scales with usage, sits on a cloud bill rather than a payroll, and can be switched on in a week. Contractor capacity sits in a third place again, usually in procurement’s vendor management system. When the three costs live in three ledgers, no one can answer the question workforce planning exists to answer: what is the cheapest, fastest and safest way to get this work done next quarter?
The year is no longer the horizon
David Green of Insight222 put it plainly in the TalentNeuron report: AI is “compressing planning cycles from years to quarters”. An annual workforce planning round that locks headcount in November cannot react to a model release in February that changes which tasks are automatable. The organisations that are coping treat workforce planning as a continuous operating discipline, revisited as often as the technology changes, rather than a budgeting event.
| Planning dimension | Traditional workforce planning | AI-era workforce planning |
|---|---|---|
| Unit of analysis | Job / role / FTE | Task, skill and capacity |
| Who counts as workforce | Employees on payroll | Employees, contractors, partners and AI agents |
| Cadence | Annual, tied to the budget | Continuous, revisited each quarter or on each capability change |
| Cost model | Fixed salary cost per head | Fixed, variable and vendor costs modelled together |
| Levers | Hire, freeze, reduce | Hire, reskill, redeploy, automate, contract — as one decision |
| Owner | HR, with finance approving the number | CFO and CHRO jointly, with procurement and IT at the table |
| Data | HRIS headcount export | Unified people, finance and vendor data |
The Three Numbers Behind the Workforce Planning Gap
SAP’s research, first quoted at its Sapphire conference in May 2026 and repeated in the VentureBeat article, gives the clearest single picture of the problem. It is worth reading the three figures together rather than separately.
62% dissatisfied with people data
Nearly two-thirds of C-suite executives say they are dissatisfied with how well people data connects to business performance. That is not a complaint about HR reporting; it is a complaint that the board cannot see how a workforce decision — hiring 40 engineers, cutting a support tier, licensing an AI agent — translates into revenue, margin or delivery. Without that line of sight, workforce planning is a cost exercise rather than a strategic one.
50% plan for productivity, 21% plan for job design
Half of organisations model what AI will do to productivity and capacity. That is the easy half: it produces a number that can be dropped into a budget. Only a fifth model what AI does to the design of jobs and the shape of the organisation, which is the hard half and the one that determines whether the productivity ever arrives. The 29-point gap between those figures is the space in which most AI workforce plans fail — the capacity is assumed, but nobody has redesigned the work to release it.
The chart below plots the three SAP figures as stated; the fourth bar is the difference between the 50% and 21% figures.
Why the sponsor matters, and why it does not
The VentureBeat article is presented by SAP, and SAP sells the workforce planning software that would close the gap it describes — the May 2026 SuccessFactors release added a workforce planning capability inside SAP Enterprise Planning that connects Cloud ERP, Fieldglass and SuccessFactors data for “employees and contingent labour”. Readers should weigh that. But the three figures line up with independent evidence from Deloitte, Gartner and Stanford, which is why the rest of this article leans on those sources rather than on the vendor.
What "Workforce" Now Means in Workforce Planning
The definition of the workforce has been widening for a decade, and the planning model has not registered it. Imbert and Albert’s phrasing is exact: employees now work alongside contractors, specialised partners and AI systems that handle “real execution-layer tasks — not just support functions, but actual work”.
Four kinds of capacity, three systems of record
In most mid-sized and large organisations the four kinds of capacity live in different systems owned by different functions. Employees sit in the HR system. Contractors and statement-of-work partners sit in procurement’s vendor management platform. AI agents sit in IT’s cloud accounts, licensed per seat or metered per task. Nobody owns a view that shows all four against the work that needs doing, which is the view workforce planning needs.
The AI agent as a workforce line item
The newest category is the one that breaks the model most completely. An AI agent that triages support tickets, drafts contracts or reconciles invoices is doing work that would previously have been a role, a fraction of a role or a contractor day rate. Our own service page on AI employees and autonomous agents describes them as digital workforce for a reason.
Yet almost no workforce planning model has a row for them: they are not headcount, so HR does not count them; they are software, so finance books them as opex; they are automation, so IT owns the contract. The work they absorb is invisible to the plan.
What the leading organisations do differently
The organisations that are ahead treat the four kinds of capacity as interchangeable levers for getting a task done, and they model hiring, reskilling, automation and contractor engagement as one decision rather than four. That is the single change that turns workforce planning from a cost forecast into a strategy. It also changes who has to be in the room, which is the subject of the CFO–CHRO section below.
The Evidence That AI Is Redefining the Workforce Right Now
If the SAP figures describe the planning gap, four independent studies published in the last eight weeks describe the pace at which the ground is moving underneath it.
Stanford: early-career employment in AI-exposed jobs is 19% below trend
The Stanford Digital Economy Lab’s “Canaries in the Coal Mine” study, revised on 12 August 2026 with ADP payroll data running to June 2026, finds that employment for workers aged 22 to 25 in the most AI-exposed occupations is now 19% below where it would be had it kept pace with less-exposed peers. Experienced workers show no comparable gap.
The effect runs through reduced hiring rather than layoffs, and it is concentrated in occupations where AI substitutes for human tasks; where AI complements experienced workers, employment is flat or rising. The authors call these early, descriptive indicators rather than causal estimates, but the gap has widened steadily since they first documented it in August 2025.
Gartner: a fifth of CHROs have already seen entry-level hiring stopped
A Gartner survey of 110 heads of HR in the fourth quarter of 2025, published on 27 July 2026, found that 21% of CHROs said at least one business leader in their organisation had stopped hiring for entry-level roles because of AI automation. The same research found that 95% of organisations had implemented AI in some form over the previous year but only one in five had achieved significant or transformational value from it.
Gartner’s Kaelyn Lowmaster warned that “organisations that respond by cutting their early career talent pipelines altogether risk creating significant workforce challenges down the road” — a workforce planning failure with a five-year lag.
Microsoft: leaders expect agents everywhere within 18 months
Microsoft’s 2025 Work Trend Index, a survey of 31,000 knowledge workers across 31 countries, found 81% of leaders expecting AI agents to be moderately or extensively integrated into their strategy within 12 to 18 months, 45% planning to hold headcount flat while using AI as digital labour, and 33% considering headcount reductions. It also found a mindset gap that workforce planning has to bridge: 67% of leaders were familiar with agents against 40% of employees. A plan that assumes the workforce understands what is coming is planning for a workforce that does not yet exist.
TalentNeuron: the planners are the growth market
The most telling data point may be TalentNeuron’s finding that, across Salesforce, Klarna, Wells Fargo, Google, Microsoft, Citi and BT Group, demand for strategic workforce planning skills rose 33% over two years, demand for people analytics roles rose 26%, and demand for learning and development specialists rose 42% — nearly doubling at Microsoft, Google and Citi. The companies most aggressive about AI are hiring workforce planners, not shedding them. Its chief executive David Wilkins summarised the report in one line: “AI is accelerating workforce transformation, but it is not replacing the need for workforce strategy.”
The bar widths above are each figure divided by the largest (42%), so the 33% bar is 79% of full width, 26% is 62% and 16% is 38%.
| Source | Published | Sample | Headline finding | What it means for workforce planning |
|---|---|---|---|---|
| SAP research (via VentureBeat) | May–Sep 2026 | C-suite executives | 62% dissatisfied with people data; 50% plan for productivity, 21% for job design | The planning model is not built to see job redesign |
| Stanford Digital Economy Lab | 12 Aug 2026 revision | ADP payroll data to June 2026 | Ages 22–25 in AI-exposed jobs 19% below trend | Entry-level pipelines are shrinking by hiring freeze, not layoff |
| Gartner | 27 Jul 2026 | 110 heads of HR, Q4 2025 | 21% saw entry-level hiring stopped; 95% adopted AI, 20% got real value | Adoption is running far ahead of measured value |
| Microsoft Work Trend Index | 2025 | 31,000 workers, 31 countries | 81% of leaders expect agents integrated within 18 months | Agents must be modelled as capacity now, not later |
| TalentNeuron | 1 Sep 2026 | Seven global enterprises | Workforce planning demand +33%; 34% of cut roles held critical judgement | Task-level analysis is replacing role-level cuts |
| Deloitte India | 29 Jul 2026 | 85+ organisations | 43% at advanced people analytics maturity; 70%+ still on static reporting | Most analytics cannot yet feed a continuous plan |
Fragmented Systems: Why HR, Finance and Procurement Cannot See One Workforce
The SAP authors’ first diagnosis is fragmentation, and it is the one most UK finance and HR leaders will recognise immediately. HR, finance and procurement each hold part of the workforce picture in a separate system built on separate assumptions, and executives are left unable to see how workforce decisions turn into business outcomes.
Three functions, three definitions of a person
HR counts people by role, grade and location. Finance counts them by cost centre and fully-loaded cost. Procurement counts contractors by purchase order and day rate, and often does not count them as people at all. A workforce planning exercise that tries to reconcile the three usually ends in a spreadsheet maintained by one analyst, refreshed quarterly, and wrong by the time the board sees it. Deloitte’s July 2026 people analytics study found that more than 70% of organisations still rely on static reporting of exactly this kind, even though 80% say they are strengthening workforce planning and contingent talent strategies.
The orchestration gap
Deloitte’s US practice describes the same problem as an orchestration gap — the inability to connect tasks, skills, roles, capacity, productivity and cost into one dynamic view of how work is changing. Its figures are stark: 88% of organisations recognise that the gap matters, and only 7% report making good progress on closing it. That 81-point spread between recognising the problem and acting on it is the practical shape of “most planning models aren’t ready”.
What a unified data foundation has to hold
The fix is less about a new tool than about a shared data foundation, which is a data management and analytics problem before it is an HR one. The table below sets out what each function has to contribute to a workforce planning model that can see the whole workforce, and what each gets back.
| Function | Data it must contribute | Decisions it owns | What it gets back from shared workforce planning |
|---|---|---|---|
| HR (CHRO) | Roles, skills, tenure, attrition, reskilling capacity | Job design, career paths, redeployment | A business case for reskilling expressed in finance’s terms |
| Finance (CFO) | Fully-loaded cost per role, opex for AI licences, budget envelopes | Capital allocation across hire / automate / contract | A capacity forecast it can trust in the P&L |
| Procurement | Contractor and SOW spend, rates, vendor capacity | External sourcing and vendor terms | Visibility of which external spend is really workforce |
| IT / AI lead | Agent inventory, task volumes, usage-based costs, governance status | Which tasks are safe to automate and how they are supervised | A demand signal for automation tied to a business owner |
From Annual Headcount to Continuous Workforce Planning
The second shift the SAP authors describe is cadence. Leading organisations, they write, treat workforce planning as an ongoing discipline rather than annual budgeting, with finance, HR and procurement able to model scenarios that combine hiring, reskilling, automation and contractor engagement as interconnected levers.
What “continuous” means in practice
Continuous does not mean daily. It means the workforce planning model is a living scenario tool that the CFO and CHRO can rerun whenever an input changes — a new AI capability, a lost contract, a regulatory deadline — rather than a document that is finalised in the autumn and defended until the next autumn. In practice that is a quarterly review with the ability to run an ad-hoc scenario in days. Deloitte’s advice is to progress incrementally: build the foundation now and strengthen the planning muscles over time, rather than waiting for a perfect system.
Scenario modelling with shared assumptions
The value of continuous workforce planning is in the scenarios, and the scenarios are only useful if every function is using the same assumptions. If HR assumes a 20% productivity gain from an AI agent and finance has budgeted for 40%, the plan is fiction. A shared assumption register — how much of each task type is automatable, at what cost, with what supervision — is the unglamorous artefact that makes the rest work, and it belongs in the same place as the company’s AI strategy.
Measuring what the plan promised
The Gartner finding that 95% of organisations adopted AI while 20% saw real value is a measurement failure as much as an adoption one. A continuous workforce planning loop closes it by tracking, per task and per team, whether the capacity the model predicted actually arrived. Our analysis of Microsoft 365 Copilot cost against ROI for UK SMEs found the same thing at tool level: the licence is easy to count, the hours released are not, unless someone is counting.
Counting AI Agents in Workforce Planning
The hardest practical question is how to put an AI agent into a workforce planning model at all. RSM’s hybrid workforce planning guide, published on 31 August 2026, is the most concrete answer so far, and its approach is task decomposition.
Split the work into three kinds of task
RSM divides work into human-led tasks (judgement, relationships, ethics and sensitive employment decisions), agent-led tasks (structured, high-volume, rules-based activities with defined inputs and outputs — benefits enrolment, payroll calculations, compliance data aggregation, screening-criteria matching) and hybrid tasks, where AI accelerates a decision a human still makes. Erzsébet Malzenicky, Experian’s global head of workforce strategy, makes the same point in the TalentNeuron report: “You can’t design a workforce blending human and automated capability without task-level analysis.”
| Task type | Definition | Examples | How it enters the workforce planning model |
|---|---|---|---|
| Human-led | Requires judgement, relationships, ethics or accountability | Performance conversations, negotiations, dismissals, strategy | Headcount and skills, planned as roles |
| Agent-led | Structured, high-volume, rules-based, defined inputs and outputs | Payroll calculation, benefits enrolment, compliance aggregation, candidate screening | Agent capacity with a cost centre, usage cost and a named owner |
| Hybrid | AI accelerates a human decision without replacing it | Drafting, summarising, first-pass analysis, forecasting | Human roles with a productivity multiplier and a supervision load |
Give every agent an owner and a cost centre
RSM’s governance rules read like a checklist for the workforce planning model itself: the business unit that benefits from an agent carries its cost rather than IT by default; AI capacity is budgeted as a variable cost, not a fixed one; output expectations are defined before deployment; and decision rights and escalation paths are documented for when an agent exceeds its scope. Without them, the guide warns, organisations get “AI agent sprawl, accumulating without clear accountability”. That is the same conclusion we reached looking at why identity and permissions are not enough to govern agent behaviour.
Plan the supervision, not just the automation
Every agent-led task creates a human task of checking it. The Microsoft Work Trend Index frames this as getting the human–agent ratio right, so that teams neither under-use agents nor burn out overseeing them. A workforce planning model that books the saving from automation without booking the supervision cost overstates the gain — one reason so many of Gartner’s 95% never reached the 20%. Where the agent uses natural language processing to read documents, for instance, the error modes are specific enough that the reviewer needs training, which is itself a workforce planning line.
The CFO and CHRO Partnership Workforce Planning Now Requires
The third of the SAP authors’ arguments is about ownership. CFOs and CHROs, they write, are being pulled into shared accountability for workforce strategy, because neither can address modern labour challenges independently.
Why finance cannot do it alone
Finance owns the number but not the work. A CFO can model the cost of 200 people, a 15% AI productivity assumption and a contractor budget, but cannot say which tasks those people do, which of them an agent could do safely, or what the retraining path looks like for the people whose tasks go. Modelled without that, the AI line in the budget is a guess dressed as a saving.
Why HR cannot do it alone
HR owns the work but not the money or the technology. A CHRO can map tasks and skills, but cannot commit capital to automation, cannot see the procurement ledger where a third of the real workforce sits, and rarely controls the AI platform decisions IT is making. The Gartner data shows what happens when business leaders act without HR: entry-level hiring stops, and the pipeline the company will need in five years disappears from the plan.
What shared accountability looks like
In practice the partnership needs three things. A single workforce planning model both executives sign, with procurement and IT feeding it. A shared set of metrics that expresses people decisions in business terms — cost per unit of work, time to capacity, value delivered per task — rather than headcount alone.
And a governance cadence, which Imbert and Albert identify as the real difficulty: the challenge is not the technology but leadership alignment on metrics, planning cadence and shared governance. Bill Gates, in his 26 August 2026 essay warning that “there is no plan” for the upheaval AI will cause, was talking about governments; the same is true of most boards.
A Practical Workforce Planning Reset for UK Businesses
None of this requires a large enterprise platform to start. A mid-sized UK firm can rebuild its workforce planning around AI in a quarter using the sequence below, which draws on RSM’s seven steps and Deloitte’s foundation-first advice.
Step 1: audit the work at task level
Take the ten highest-volume processes in the business and break each into tasks, classifying every task as human-led, agent-led or hybrid. Record the hours, the people, the contractors and any software already doing part of it. This is the exercise Caterpillar ran before automating mining equipment, and it is the one most AI pilots skip.
Step 2: build the single view
Pull the HR, finance and procurement data for those processes into one model — a governed spreadsheet is fine at first — with fully-loaded cost per task, not per head. Add a row for every AI agent or automation already running, with its owner, cost centre and usage cost. This is the moment the 62% figure becomes visible in your own numbers.
Step 3: agree the assumptions
Write down, jointly between the CFO and CHRO, the automation assumption for each task type, the supervision load it creates and the reskilling path for affected people. Publish it internally. Disagreement here is cheap; disagreement after the budget is set is not.
Step 4: run three scenarios
Model a hire-led, an automate-led and a blended scenario for the next four quarters. The blended one usually wins, because it uses reskilling and redeployment — Deloitte’s core levers — to keep the judgement-heavy 34% of tasks that TalentNeuron found inside “eliminated” roles.
Step 5: protect the entry-level pipeline deliberately
Decide, as a policy rather than a default, how many early-career roles the business will keep and what those roles now do. Gartner’s advice is to redefine them so that new hires contribute to higher-value work earlier, not to remove them. Stanford’s 19% gap is what happens when nobody decides.
Step 6: govern it as a standing process
Put workforce planning on the quarterly executive agenda alongside the financial forecast, with the assumption register reviewed each time. Treat it as part of strategic IT planning as much as HR planning, because the capacity decisions are increasingly technology decisions. If you lack the people to run it, that is a legitimate use of AI-aware staff augmentation rather than a reason to wait.
Step 7: measure the value, not the adoption
Report each quarter on whether the capacity the model promised arrived, task by task, and feed the answer back into the assumptions. That single loop is the difference between Gartner’s 95% who adopted and the 20% who benefited.
Workforce Planning FAQ
What is workforce planning?
Workforce planning is the process of working out what work an organisation needs done, what capacity — people, contractors and increasingly AI systems — it needs to do it, and how to close the gap through hiring, reskilling, redeployment, automation or contracting. Strategic workforce planning looks three to five years out; operational workforce planning covers the next few quarters.
Why are most workforce planning models not ready for AI?
Because they count jobs and salaries once a year, while AI changes tasks and costs continuously. SAP’s research found only 21% of organisations plan for AI’s effect on job design, and Deloitte found only 7% making good progress on connecting tasks, skills, capacity and cost into one view.
Should AI agents be counted in workforce planning?
Yes. RSM’s guidance is to revise workforce planning to include AI agent capacity alongside human FTEs, with each agent assigned an owner, a cost centre, output expectations and escalation rules. An agent that does execution-layer work is workforce, whatever ledger it sits in.
Who should own workforce planning — HR or finance?
Both. The SAP authors argue the CFO and CHRO now share accountability for workforce strategy, with procurement and IT contributing data. A model owned by one function will systematically miss what the other sees.
Is AI actually reducing jobs yet?
Selectively. Stanford’s August 2026 data shows a 19% employment gap for 22-to-25-year-olds in AI-exposed occupations, driven by reduced hiring rather than layoffs, while experienced workers show no comparable gap. Gartner found 21% of CHROs had seen entry-level hiring stopped because of AI, and TalentNeuron found the same AI-leading employers increasing demand for workforce planning and people analytics roles.
How often should workforce planning be revisited?
Quarterly at minimum, with the ability to rerun a scenario within days when an input changes. Insight222’s David Green describes AI as compressing planning cycles “from years to quarters”, and an annual cycle cannot keep up.
References
AI is redefining the workforce — and most planning models aren’t ready
SAP SuccessFactors Innovations Usher in a New Era of Autonomous HCM
Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence
Gartner: AI is already cutting some entry-level hiring
TalentNeuron Research Finds AI Is Driving Workforce Redesign, Not Solely Workforce Reduction
2025: The Year the Frontier Firm Is Born
43 percent of organisations reach advanced people analytics maturity
AI-Enabled Strategic Workforce Planning
Planning a hybrid workforce of people and AI agents
Bill Gates warns ‘there is no plan’ for the ‘upheaval’ AI will cause
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.