Rewiring the org for human-agent teams
Contents
About Microsoft
Microsoft, founded in 1975 by Bill Gates and Paul Allen, headquartered in Redmond, Washington, and led by chairman and chief executive Satya Nadella since 2014: is one of the world's most valuable companies (NASDAQ: MSFT). The Microsoft Cloud spans Azure, Microsoft 365, Dynamics 365, Microsoft Security, Microsoft Fabric and GitHub. Volge delivers across that estate through six alliance offerings.
What Volge delivers on Microsoft: six offerings
The offering this book covers: marked ◆: is Modern Workspace & E7, delivered on Microsoft 365 Copilot.
| Volge offering | What it covers | Anchored on |
|---|---|---|
| Modern Workspace & E7 ◆ | Secure modern work on the E7 bundle; human-agent collaboration | Microsoft 365 · Copilot |
| Applied AI | Building, governing and running enterprise AI agents and apps | Azure AI Foundry |
| Zero Trust Security | Identity, threat protection and data governance | Microsoft Entra · Defender · Purview |
| Business Applications | ERP, CRM and low-code, with in-process agents | Dynamics 365 · Power Platform |
| Digital Product Engineering | Autonomous software engineering and CI/CD | GitHub · GitHub Copilot |
| Cloud Infrastructure & Operations | Cloud build, run and cost governance (FinOps) | Microsoft Azure |
Sources: Microsoft corporate and product documentation, 2025–26.
This book: Modern Workspace & E7 on Microsoft 365 Copilot
Microsoft 365 Copilot is Microsoft's enterprise AI assistant embedded across the Microsoft 365 apps and grounded in an organisation's own work data through the Microsoft Graph, inside the tenant's security and compliance boundary.
In the field in 2026
- In-app Copilot across Word, Excel, PowerPoint, Outlook and Teams, grounded in your files, mail, chats and meetings
- Researcher and Analyst reasoning agents for multi-step research and advanced data analysis
- Custom agent building with Copilot Studio and the agent store, surfaced across Teams, SharePoint and Outlook
- Microsoft 365 Copilot Chat, secure, web-grounded chat with pay-as-you-go metered access to agents
Executive summary
Revenue-per-employee is decided by two numbers: the output a hybrid team produces, and the headcount it takes to produce it. Through 2026 the first number has become cheap to raise and the second has become difficult to hold. The agents are abundant; the discipline to reorganise work around them is not. That gap, and not the capability gap, is where the productivity return now sits.
Microsoft states the position without hedging in its 2026 Work Trend Index: "the constraint is no longer what people can do, it is how work is structured around them" (Microsoft, 2026). The supply of intelligence has stopped being the binding limit; the design of the work around it has taken its place.
The term "Frontier Firm" is Microsoft's. It was introduced in the 2025 Work Trend Index to describe a new kind of organisation built on three things: intelligence on tap, human-agent teams, and a new role for everyone, the agent boss (Microsoft, 2025). The 2026 edition moves the idea from a description of the workforce to a claim about the operating model: a Frontier Firm is one that rebuilds the operating model itself for the age of agents (Microsoft, 2026).
The supply side has already settled. Eighty-two per cent of leaders expected to use digital labour to expand workforce capacity within twelve to eighteen months, and 46 per cent reported already using agents to fully automate workstreams or business processes (Microsoft, 2025). When a capability becomes common, it stops conferring advantage. Microsoft draws the conclusion directly: "access to AI won't be the advantage for much longer" (Microsoft, 2026).
What separates the firms capturing the gain from the firms running pilots is structural, and three independent studies say the same thing. Microsoft finds that organisational factors, culture, manager support, and talent practices, explain more than twice the AI impact of individual factors such as mindset and behaviour, at 67 per cent against 32 per cent (Microsoft, 2026). PwC finds that nearly three-quarters of AI's economic value lands with one-fifth of organisations, the leaders that pursue growth and not productivity alone (PwC, 2026). BCG finds that the organisations using AI to reshape workflows end-to-end or to invent new business models have nearly doubled to 42 per cent, and that this population creates more value (BCG, 2026).
The central finding of this paper is the following. Revenue-per-employee gains accrue to enterprises that redesign the operating model around hybrid human-agent teams and that upskill their people to manage agents, not to enterprises that buy the most capable agents. The technology choice is largely closed by existing platform commitments. The operating-model choice is open, and it is the lever that decides whether a firm pulls ahead of its competitors or falls behind them.
The divide compounds. The distance between the leaders capturing value and the majority still piloting widens with every cycle. Microsoft observes that the firms rebuilding the operating model build "an organisation that learns faster than its competitors, compounds its own intelligence and gets harder to catch with every cycle" (Microsoft, 2026). A redesigned operating model is not a one-time efficiency. It is a compounding rate.
Five conclusions are advanced, each developed in the chapters that follow.
First, the operating model is the differentiator, not the technology. Organisational factors explain more than twice the AI impact of individual factors (Microsoft, 2026). The enterprises that capture revenue-per-employee gains restructure the work; they do not simply deploy more agents. This moves the agenda from procurement to organisation design.
Second, work is decomposed before it is staffed, and the human-agent ratio becomes the planning metric. Hybrid teams form around outcomes rather than functions (Microsoft, 2025). For each workstream the leadership team sets how many agents are required and how many humans are required to direct them. The ratio is task-specific, and setting it well is the core operating-model decision.
Third, every role moves up the value chain toward supervision. As agents take on tactical execution, the human contribution shifts to setting direction, defining standards, and judging outcomes (Microsoft, 2026). Forty-seven per cent of leaders named upskilling the existing workforce a top priority within twelve to eighteen months (Microsoft, 2025). The two human skills rated most important are quality control of AI output and critical thinking (Microsoft, 2026), both supervisory, neither technical.
Fourth, adoption is an organisational property, not an individual one. Culture, manager support, and talent practices account for 67 per cent of AI impact against 32 per cent for individual factors (Microsoft, 2026). A Transformation Paradox holds firms back: 65 per cent of AI users fear falling behind, yet 45 per cent say it feels safer to focus on current goals than to redesign work, and only 13 per cent are rewarded for reinvention (Microsoft, 2026).
Fifth, the work is changing faster than firms are changing the work. AI is reshaping jobs faster than organisations are redesigning them around AI (BCG, 2026). By 2030 an estimated 70 per cent of the skills used in most jobs will change (LinkedIn, 2026). The firms that close this gap deliberately, through a staged rebuild, convert capability into revenue-per-employee. The firms that wait for the gap to close on its own watch it widen.
This paper is written for IT managers and IT heads, working in partnership with human-resources and operations leaders. The Frontier Firm is an operating-model and workforce undertaking before it is a technology one. The IT function is its enabling partner: it supplies the agent platform, the identity and governance estate, and the data foundation on which hybrid teams depend. The chapters that follow set out the constraint that has moved, the structure of the hybrid team, the failure modes that stall adoption, and the sequence that converts agent capability into revenue-per-employee growth.
Chapter one: Revenue-per-employee, and the constraint that has moved
The takeaway
A Frontier Firm has rebuilt its operating model around human-agent teams; it has not merely adopted agents. The distinction decides revenue-per-employee. Organisational factors explain more than twice the productivity impact of individual or technological factors, which means the binding constraint has moved off the technology and onto the design of the work. The strategic question is no longer which agents to buy. It is how to reorganise the work so that humans and agents together produce more output per person.
2.1 The equation the rest of the paper serves
Start with the arithmetic, because every later argument resolves to it. Revenue-per-employee rises in exactly two ways: the same headcount produces more output, or output grows faster than headcount. Agents act on both terms: they raise output and they slow the headcount a given volume of work requires. Neither effect materialises on its own. Both depend on whether the work is redesigned so that humans are redeployed to higher-value tasks and agents absorb the tactical load.
This is the line that divides a local efficiency from a structural one. An agent bolted onto an unchanged process captures a saving inside that process and stops there. An operating model rebuilt around hybrid teams moves the saving across the firm and compounds it, a different rate of growth in output per person, sustained cycle over cycle.
2.2 The origin and definition of the term
Precision of language matters here, because the term is now used loosely. "Frontier Firm" is Microsoft's, introduced in the 2025 Work Trend Index and developed in the 2026 edition. It is defined and used as follows in this paper.
A Frontier Firm is an organisation built around three characteristics: intelligence on tap, human-agent teams, and a new role for everyone, the agent boss (Microsoft, 2025).
Intelligence on tap treats expertise as a utility, drawn on demand in the manner of electricity or cloud storage. The marginal cost of an additional unit of cognitive work falls toward the cost of compute.
Human-agent teams are working units in which people and AI agents collaborate against a shared outcome, with the mix set deliberately rather than by default.
The 2026 edition adds the operating-model dimension. A Frontier Firm is defined by "how deliberately leaders design work across functions, matching the level of human involvement to the outcome" (Microsoft, 2026). The weight shifts from the technology a firm holds to the way a firm organises around it.
2.3 Capability has stopped being scarce
A common input cannot be a durable source of advantage. The first reason to focus on the operating model is that the technology has become common.
Eighty-two per cent of leaders said they expected to use digital labour to expand their workforce within twelve to eighteen months (Microsoft, 2025). Eighty-one per cent expected agents to be moderately or extensively integrated into their company's AI strategy over the same period (Microsoft, 2025). Forty-six per cent reported already using agents to fully automate at least one workstream or business process (Microsoft, 2025).
Microsoft reads the trajectory plainly: "access to AI won't be the advantage for much longer. How the work is designed around it will be" (Microsoft, 2026). When every competitor can rent the same intelligence, the advantage migrates to the firm that organises labour around it most effectively.
By the numbers, the supply side, 2025–2026 82% of leaders expected to use digital labour to expand workforce capacity within 12–18 months (Microsoft, 2025). 81% expected agents to be moderately or extensively integrated into AI strategy within 12–18 months (Microsoft, 2025). 46% were already using agents to fully automate a workstream or business process (Microsoft, 2025).
2.4 Three studies, one variable
The claim that the operating model drives the gain rests on three independent bodies of research that converge on a single variable.
Microsoft's 2026 analysis isolates the drivers of AI impact. Organisational factors, culture, manager support, and talent practices, account for more than twice the AI impact of individual factors such as mindset and behaviour, at 67 per cent against 32 per cent (Microsoft, 2026). The firm draws the conclusion the data forces: "the real question isn't whether people have the right skills, it's whether the organisation is built to unlock them" (Microsoft, 2026).
PwC's 2026 AI performance study finds that nearly three-quarters of AI's economic value is captured by one-fifth of organisations, with the leaders focused on growth and not productivity alone (PwC, 2026). The value is concentrated, and the variable that concentrates it is how the firm uses the capability.
BCG's 2026 research finds that the share of organisations using AI to reshape workflows end-to-end or to invent new business models has nearly doubled to 42 per cent, and that this population continues to create more value and a better employee experience (BCG, 2026). The reshaping of the work, not the deployment of the tool, marks out the value creators.
Three different methods, three different samples, one answer: the return tracks the redesign, not the technology.
2.5 What the redesigners actually capture
The clearest measure of the prize comes from firms that rebuilt workflows rather than added tools.
Klarna integrated AI into 96 per cent of employee workflows and reported a 152 per cent increase in revenue per employee since the first quarter of 2023 (company disclosure, reported 2026). The gain follows from workflow redesign at scale, not from a single deployment.
Set that against the macro figure and the contrast does the teaching. Companies reported an average productivity gain from AI usage of 1.8 per cent in 2025, a number that averages the whole population, including the majority that has not touched the work (Federal Reserve Bank of Atlanta, 2026). The distance between 1.8 per cent across everyone and the gains the redesigners report is the distance this paper is written to close.
| Indicator | Population | Figure | Source |
|---|---|---|---|
| Revenue per employee uplift | Klarna, workflow-redesigned | +152% since Q1 2023 | Company disclosure, 2026 |
| Average productivity gain from AI | All adopters | +1.8% in 2025 | Federal Reserve Bank of Atlanta, 2026 |
| Share of AI economic value captured | Top 20% of organisations | ~74% of value | PwC, 2026 |
| Organisations reshaping workflows end-to-end | All surveyed | 42%, nearly doubled | BCG, 2026 |
The figures are drawn from differing survey bases and are not strictly comparable. They are presented to illustrate the dispersion between redesigners and the wider population.
2.6 The implication for the leadership agenda
The supply of agent capability has become common, and a common input cannot sustain advantage. The variable that explains the dispersion in returns is the operating model, evidenced consistently across Microsoft, PwC, and BCG research. The leadership question moves with it. It is not "which agents", but "how is the work redesigned around them". The chapters that follow answer that question, beginning with the unit in which the redesign actually happens: the hybrid team.
Chapter two: The hybrid team: work decomposition, the human-agent ratio, and the agent boss
The takeaway
The hybrid team is the unit where the operating-model rebuild is executed, and it is built in a fixed order: decompose the work, set the human-agent ratio, then assign the human role. Decomposition comes first because a ratio cannot be set over work that has not been broken into tasks. The human-agent ratio is the planning metric that replaces headcount. The human role moves from doing the work to directing it, the agent boss, and the four collaboration patterns name the steps along that move.
3.1 Decompose the work before staffing it
The error most firms make is to ask "where can an agent help" before asking "what are the tasks". The Frontier Firm reverses the order. It breaks an outcome into discrete tasks, then assigns each task to a human, an agent, or a supervised pairing of the two. This is why the structure shifts from a fixed org chart to a fluid work chart.
From org chart to work chart
The work chart is Microsoft's term for the structure that replaces the org chart inside a Frontier Firm (Microsoft, 2025). The two answer different questions. An org chart answers "who reports to whom": it maps people to permanent reporting lines, optimises for control and accountability, and holds its shape for years. A work chart answers "what gets the outcome": it maps the right mix of people and agents to a specific goal, optimises for throughput, and re-forms as the goal changes.
Three properties define it. It is outcome-anchored: a unit assembles around a deliverable rather than a function. It is hybrid: humans and agents are both first-class nodes, not people in boxes with tools on the side. And it is transient: the unit forms when the outcome is set and dissolves when it is met, where an org-chart team persists regardless. The org chart does not disappear; payroll, line management, and duty of care still run on it. The work chart sits on top of it as the structure through which the work is actually done.
This is the operating-model consequence of decomposition. Once an outcome is cut into tasks and each task is assigned to a human, an agent, or a supervised pairing, the assignment is the work chart. Decomposition is the precondition for everything downstream. A rules-bound, high-volume, reversible task goes to a fleet of agents under light supervision; a high-judgement, high-stakes, hard-to-reverse task stays with a human who draws on an agent for support. The ratio and the role both fall out of how the work was cut. Cut it badly and no ratio rescues the team.
The consequence for the IT function is direct. A fluid work chart requires that agents be provisioned, identified, and governed as fast as teams form, which makes the agent platform and the identity estate the enabling infrastructure of the operating model. The structure cannot flex faster than the platform that supplies it.
3.2 The human-agent ratio as the planning metric
Once the work is decomposed, the second choice is quantitative. Microsoft introduces the human-agent ratio as the metric required to maximise the impact of hybrid teams (Microsoft, 2025), and in a Frontier Firm it does the planning work that headcount and span of control did in a traditional one.
The ratio answers two questions for each workstream: how many agents are needed for which roles and tasks, and how many humans are needed to guide them (Microsoft, 2025). Microsoft is explicit that getting the ratio right is critical and task-specific. There is no single correct ratio for an enterprise, and a ratio that suits one workstream will be wrong for the next.
The ratio is bounded by judgement and accountability, not by efficiency alone. Microsoft notes that leaders need to weigh when human and digital labour together outperform AI alone, when customers prefer a human touch, and when society expects a person to be accountable for the consequences (Microsoft, 2025). A ratio set purely to minimise cost will over-automate the work where accountability matters most. The ratio is an operating-model decision, not a cost-minimisation calculation.
The following table sets out an illustrative ratio framework. The figures are indicative and intended to show how the ratio varies with the characteristics of the work, not to prescribe values.
| Work characteristic | Indicative human-agent ratio | Human role | Rationale |
|---|---|---|---|
| High-volume, rules-bound, reversible | 1 human: many agents | Supervisor of a fleet; exception handler | Agents absorb the volume; the human handles exceptions |
| Moderate-volume, judgement-assisted | 1 human: few agents | Director; reviewer of outputs | The human sets intent and approves consequential steps |
| Low-volume, high-judgement, high-stakes | Many humans: 1 agent | Author; the agent assists | The human remains the actor; the agent augments |
| Customer-facing, trust-sensitive | Human-led, agent-supported | Relationship owner | A human touch is expected; the agent prepares and follows up |
The ratios are indicative. The principle, from Microsoft (2025), is that the ratio is task-specific and set deliberately by the leadership team.
3.3 The four collaboration patterns
The ratio is set; the third choice is the human role at the task level, and Microsoft's 2026 research gives it a precise vocabulary. The division of work between a person and an agent moves through four distinct patterns, first observed in software engineering and now appearing across other functions (Microsoft, 2026).
Author. The person produces the work and calls on AI for help as needed, a line of code, a sentence, a chart. The human remains the actor.
Editor. The person sets the intent, and AI creates the first draft for the person to edit and approve.
Director. The person creates a specification and hands off entire tasks for AI to execute in the background.
Orchestrator. The person designs a system in which multiple agents run in parallel across a workflow, flagging exceptions and escalations to the person.
The pattern is the operational form of the ratio. Author implies many humans to one agent; Orchestrator implies one human to many. Span of control widens as the work moves from left to right: a single person supervises a handful of agents under Director and a fleet under Orchestrator. Microsoft is clear that the goal is not to push every task to Orchestrator, but to match each workstream to the right pattern (Microsoft, 2026). The mistake is treating the rightmost pattern as the destination for all work rather than the correct fit for some of it.
Figure 1: The four collaboration patterns, from Author to Orchestrator (Microsoft, 2026). Human involvement in tactical execution declines from left to right; the human-agent ratio inverts and span of control widens across the same axis.
3.4 The agent boss is the supervisory layer
The four patterns converge on one new role. Microsoft terms it the agent boss: a person who builds, delegates to, and manages agents to amplify their impact (Microsoft, 2025). Read it as the supervisory layer of the hybrid team: the point of accountability for work that agents execute.
The role is not reserved for specialists. Microsoft's position is that the agent boss is "a new role for everyone", and that within five years teams will routinely train and manage AI agents as part of their responsibilities (Microsoft, 2025). The 2026 research confirms the direction: as agent use rises, human involvement does not disappear but changes shape, with tactical execution declining and the need to set direction, define standards, and judge outcomes rising (Microsoft, 2026).
The shift already shows up in how managers spend their time. BCG finds that 47 per cent of respondents report spending more time managing and directing AI than doing the work themselves, and that 72 per cent say AI has already considerably changed the skills expected in their roles (BCG, 2026). Sixty-five per cent of managers and leaders believe agents will take over at least half of their job within three years (BCG, 2026).
The organisational answer is starting to form. More than a quarter of managers, 28 per cent, were considering hiring AI workforce managers to run human-agent teams, and 32 per cent planned to bring on AI agent specialists within twelve to eighteen months (Microsoft, 2025). A new supervisory tier is being staffed before most firms have named it.
By the numbers, the changing human role, 2026 47% of respondents spend more time managing and directing AI than doing the work themselves (BCG, 2026). 72% say AI has already considerably changed the skills expected in their roles (BCG, 2026). 65% of managers believe agents will take over at least half of their job within three years (BCG, 2026).
3.5 The team topology
A mature hybrid team has a definite shape. The human sits at the centre as the agent boss, accountable for the outcome. The agent boss directs a set of agents, each scoped to a task drawn from the decomposition, and supervises their work. The agents reach tools and data through the platform, and they may hand off to one another where the handoff is governed. Exceptions route back to the human, which is the load the supervisory layer is sized to carry.
Figure 2: A hybrid team topology. The human owns the outcome and supervises a set of scoped agents that act on tools and data, escalating exceptions to the human (Volge synthesis, drawing on Microsoft 2025–2026 collaboration patterns).
3.6 The implication for the leadership agenda
The hybrid team is designed in order: decompose the work, set the human-agent ratio, match each task to a collaboration pattern, and assign the agent boss who supervises the result. None of the four is a technology decision. Each is an operating-model decision, supported by the IT platform but owned by the business and human-resources leadership. The capability that lets a person hold the agent-boss role, and the reasons firms struggle to build it, is the subject of the next chapter.
Chapter three: Why adoption stalls: the upskilling gap and the transformation paradox
The takeaway
Adoption stalls for two reasons, and neither is the technology. The workforce lacks the supervisory skills the agent-boss role demands, quality control of AI output and critical thinking, and the incentive system rewards delivering current goals rather than redesigning the work. The first is a capability-building problem; the second is the Transformation Paradox. Both are organisational, which is why culture, manager support, and talent practices explain more than twice the impact of individual mindset. A firm that trains without fixing incentives, or fixes incentives without training, releases neither.
4.1 The pace that forces the issue
The case for upskilling rests first on a pace of change that is now well evidenced and faster than the systems meant to absorb it.
By 2030 an estimated 70 per cent of the skills used in most jobs will change, with AI the principal catalyst (LinkedIn, 2026). Over the nearer term, an estimated 50 to 55 per cent of jobs in the United States will be reshaped by AI within two to three years (LinkedIn, 2026). BCG's framing is that AI will reshape more jobs than it replaces, with most roles remaining but changing substantially, and 10 to 15 per cent of United States jobs potentially eliminated over five years or longer (BCG, 2026).
The strain is already visible on the floor. One-third of workers experienced fifteen or more major changes in the past year, and only 27 per cent believe their organisations manage change well (LinkedIn, 2026). The change is outrunning the capacity to absorb it: the gap an upskilling pipeline exists to close.
4.2 The scarce skill is supervision, not construction
The most consequential finding for upskilling is that the priority skills govern agents; they do not build them.
Asked which human skills matter most as AI takes on more work, AI users named two above the rest: quality control of AI output, at 50 per cent, and critical thinking, defined as analysing information objectively and reaching a reasoned judgement, at 46 per cent (Microsoft, 2026). These are the skills of the agent boss, the ability to direct an agent, review its work, and stand accountable for the result.
This reframes the upskilling target. The scarce capability is not building an agent, which the platform increasingly supplies. It is judging an agent's output, setting the standard it must meet, and owning the consequence. The pipeline therefore trains judgement, supervision, and standard-setting, not toolmaking. A firm that builds a curriculum around prompt syntax has trained the wrong layer.
Technical demand is rising too, but it is the smaller part of the agenda. LinkedIn data shows rising demand for prompt engineering, large-language-model skills, and machine-learning operations, alongside rising demand for onboarding design, mentoring, and cross-functional leadership as organisations reskill internally (LinkedIn, 2026).
By the numbers, the skills agenda, 2026 50% of AI users rate quality control of AI output as the most important human skill (Microsoft, 2026). 46% rate critical thinking among the most important (Microsoft, 2026). 70% of the skills used in most jobs will change by 2030 (LinkedIn, 2026). 47% of leaders named upskilling the existing workforce a top priority within 12–18 months (Microsoft, 2025).
4.3 The role-evolution path the pipeline targets
Roles evolve along one path, and making it explicit lets the upskilling pipeline aim at a defined target rather than a vague aspiration. Each role travels the same route, from performing tactical work to directing and governing the agents that perform it.
Figure 3: The role-evolution path. Tactical execution declines along the path; direction, standard-setting, and governance rise (Volge synthesis, drawing on Microsoft 2025–2026 and BCG 2026).
The table below shows how representative roles change, to make the upskilling target concrete.
| Role today | Role in the Frontier Firm | Principal new capability |
|---|---|---|
| Service-desk analyst | Supervisor of a service-agent fleet | Exception handling; quality control of agent resolutions |
| Financial analyst | Director of reconciliation and close agents | Standard-setting; audit and judgement over agent output |
| Software engineer | Orchestrator of coding and testing agents | System design; review of generated code |
| Marketing manager | Agent boss for content and campaign agents | Intent specification; brand and quality control |
| IT operations engineer | Designer and governor of the agent estate | Identity, observability, and reliability of agents |
4.4 Adoption is an organisational property
The finding IT and human-resources leaders most need to internalise is that adoption is set by the organisation, not the individual.
Microsoft's 2026 research shows that organisational factors, culture, manager support, and talent practices, account for more than twice the AI impact of individual factors such as mindset and behaviour, at 67 per cent against 32 per cent (Microsoft, 2026). The firm describes three components of an AI-ready environment: a culture that treats AI as a strategic advantage and encourages experimentation, managers who model and incentivise AI use, and talent practices that build skills and create space to apply them (Microsoft, 2026).
The consequence is unforgiving for the training budget. An upskilling programme delivered without those surrounding conditions underperforms. Training that managers do not reinforce, incentives do not reward, and culture does not support will not convert into adoption. The investment belongs in the environment, not only in the curriculum, and a firm that spends on the curriculum alone has mis-allocated against a 67-per-cent finding.
4.5 The transformation paradox
The principal obstacle to adoption is a tension Microsoft terms the Transformation Paradox, and it is governed by incentives rather than capability.
Sixty-five per cent of AI users fear falling behind if they do not use AI to adapt quickly, yet 45 per cent say it feels safer to focus on current goals than to redesign work with AI (Microsoft, 2026). Only 13 per cent of workers say they are rewarded for the reinvention of work with AI even when results fall short (Microsoft, 2026). The same fear that should drive adoption also freezes it.
The reading for leadership is exact. Employees will not redesign their work when the incentive structure rewards only the delivery of current goals, because the redesign carries short-term risk and the reward system does not price that risk. The 13 per cent figure is the binding constraint, and it is released by changing what the organisation rewards, not by adding another tool or another course.
This is the same gap BCG identifies from the outside: AI is reshaping jobs faster than companies are reshaping work (BCG, 2026). It is not a capability gap. It is an organisational and incentive gap, and it closes through deliberate change management, not through further technology.
4.6 The implication for the leadership agenda
Upskilling is necessary but not sufficient. The skills that matter most are the supervisory skills of the agent boss, quality control and critical thinking, and they have to be built across the workforce, not confined to specialists. They will not convert into revenue-per-employee unless the organisation also supplies the culture, the manager support, the talent practices, and, decisively, the incentives that reward the redesign of work. Two levers, pulled together; pull one alone and the rebuild stalls. The sequence that pulls them in the right order is the subject of the final chapter.
Chapter four: Sequencing the rebuild: blueprint, roadmap, and a maturity model
The takeaway
Sequence is the decision that distinguishes a rebuild that holds from one that collapses back to pilots. The Frontier Firm operating model reduces to six dimensions the leadership team has to redesign deliberately. The transformation that delivers them builds the platform and the organisational foundations first, then migrates work along the collaboration patterns only as confidence and controls mature. A maturity model locates the enterprise and defines the next increment. The cadence is conservative on autonomy and aggressive on the foundations: autonomy granted ahead of the controls is the most common way these programmes fail.
5.1 The operating-model blueprint, in six dimensions
A Frontier Firm differs from a traditional one across six dimensions. The table sets out the contrast and serves as the central diagnostic of this chapter.
| Dimension | Traditional firm | Frontier Firm |
|---|---|---|
| Unit of organisation | Fixed org chart; functional reporting lines | Fluid work chart; teams formed around outcomes |
| Unit of labour | Human headcount | Hybrid human-agent teams; human-agent ratio per task |
| Human role | Performs the work | Directs, reviews, and governs agents (agent boss) |
| Capability model | Hire and train for execution skills | Upskill for quality control, critical thinking, and orchestration |
| Performance metric | Output per employee | Revenue-per-employee; outcomes per hybrid team |
| Decision on work | Allocate tasks to people | Match each workstream to a collaboration pattern |
The blueprint synthesises Microsoft (2025, 2026), BCG (2026), and PwC (2026). The unifying principle is that each dimension is redesigned deliberately, not allowed to emerge by default.
The shift in organisation structure is shown below.
Figure 4: Traditional org chart against the Frontier Firm work chart. Reporting lines give way to outcome owners directing hybrid teams (Volge synthesis, drawing on Microsoft 2025).
5.2 Who owns the rebuild
The IT function enables the operating model; it does not own it, and confusing the two is itself a failure mode. The operating model cannot flex faster than the platform that supplies and governs the agents, so IT's responsibilities can be stated precisely.
IT supplies four foundations. It provisions the agent platform on which hybrid teams are assembled. It supplies the identity estate that gives every agent a first-class, governable identity. It supplies the observability and governance layer that allows agents to be supervised and revoked. It supplies the data foundation on which agents act.
Ownership is the decisive point. The operating-model redesign belongs to the business and human-resources leadership, because they own the outcome, the workforce, and the human-agent ratio. IT enables it. A rebuild led by IT alone reproduces the failure in which agents are governed as software rather than organised as labour: the controls exist but no one has decomposed the work or set the ratio. A rebuild attempted without IT lacks the platform to execute. The two are designed together or they fail apart.
5.3 The transformation roadmap
The roadmap is staged deliberately, and the staging is the argument. The evidence of chapter three is that adoption is an organisational property and that the binding constraint is incentive and culture, not capability. So the roadmap builds the organisational and platform foundations first and grants autonomy along the collaboration patterns only as those foundations mature.
Figure 5: The transformation roadmap. The cadence is conservative on autonomy and aggressive on the organisational and platform foundations (Volge synthesis).
Stage one, the first 90 days: establish the foundations. Stand up the agent platform, the identity estate, and the data foundation under IT ownership. In parallel, and this is the discipline most programmes skip, address the organisational foundations: name outcome owners, define the agent-boss role, and rewrite the incentive structure so that the redesign of work is rewarded, releasing the 13 per cent reward gap directly (Microsoft, 2026). The exit test for the stage is a platform ready to provision agents and an organisation ready to reward their adoption.
Stage two, months three to nine: prove hybrid teams in one or two workstreams. Choose one or two workstreams suited to hybrid teaming: high-volume, rules-bound, and verifiable. Decompose the work, set the human-agent ratio deliberately, assign agent bosses, and upskill them in quality control and critical thinking. The exit test is a demonstrated, measured gain in revenue-per-employee for those workstreams, not a proliferation of pilots that prove nothing in particular.
Stage three, months nine to eighteen: scale and reshape the work. For the proven workstreams, migrate along the collaboration patterns from Editor toward Director and Orchestrator as the evidence justifies, never ahead of it. Extend to adjacent workstreams. Convert the org chart into a work chart for the transformed functions. The exit test is a portfolio of hybrid teams operating under a deliberate operating model.
Stage four, beyond eighteen months: the Frontier operating model. Generalise the work-chart structure, the human-agent ratio discipline, and the agent-boss role across functions. The destination is the durable advantage Microsoft describes: an organisation that learns faster than its competitors and compounds its own intelligence with each cycle (Microsoft, 2026).
5.4 A Frontier Firm maturity model
The model below lets leadership locate the enterprise and define the next increment. It is the diagnostic counterpart to the roadmap, and it integrates the operating-model dimensions of section 5.1.
| Level | Description | Structure | Human role | Typical outcome |
|---|---|---|---|---|
| 0: Traditional | No agents in production; fixed org chart | Functional hierarchy | Performs all work | Baseline productivity |
| 1: Augmented | Copilots assist individuals | Org chart unchanged | AI-assisted worker (Editor) | Modest, individual productivity gains |
| 2: Hybrid teams | Agents in 1–2 workstreams; ratio set | Hybrid teams within a function | Agent boss (Director) | First measured revenue-per-employee gain |
| 3: Reshaped functions | Work charts replace org charts in key functions | Outcome-based teams | Orchestrator of agent fleets | Function-level productivity step-change |
| 4: Frontier Firm | Work-chart operating model across the firm | Fluid, outcome-driven across functions | Workforce designer; sets ratios and standards | Compounding, durable advantage |
Most enterprises sit between Level 0 and Level 1 as of mid-2026. The population that captures the bulk of AI value, the top 20 per cent identified by PwC (2026), sits at Level 3 or above. Level 4 is emergent.
5.5 The implication for the leadership agenda
The blueprint, the roadmap, and the maturity model are one system, and the sequence binds them. The six-dimension blueprint defines the target. The staged roadmap builds the foundations before it grants the autonomy. The maturity model fixes the enterprise's position and the next increment. The firm that builds the platform and organisational foundations first, decomposes the work, sets the human-agent ratio deliberately, upskills its people to govern agents, and rewards the redesign of work is the firm that converts agent capability into revenue-per-employee growth. The firm that grants autonomy ahead of the controls is the one that ends the year with pilots and no gain.
Chapter five: Implications for the IT leadership agenda
The Frontier Firm is an operating-model and workforce undertaking, and the evidence assembled in this paper supports a small number of clear conclusions for IT leaders and their partners in human resources and operations.
The supply of agent capability has become common, and a common input cannot sustain advantage. The variable that explains the dispersion in returns is the operating model, evidenced consistently by Microsoft, PwC, and BCG. Organisational factors account for more than twice the AI impact of individual or technological factors (Microsoft, 2026). The redesign of work, not the deployment of tools, is the lever that moves revenue-per-employee.
The IT function's job is to enable the transformation, not to own it. IT supplies the agent platform, the identity estate, the governance layer, and the data foundation on which hybrid teams depend. The operating-model redesign is owned jointly with human-resources and operations leadership, because they own the workforce, the human-agent ratio, and the incentives that govern adoption. The two are designed together.
Six questions are recommended for the IT leadership agenda, to be addressed in partnership with human-resources and operations leaders.
First, is the operating-model redesign owned jointly by IT, human resources, and operations, or is it being run as a technology procurement.
Second, has the work been decomposed and the human-agent ratio set deliberately for each priority workstream, or is the ratio emerging by default.
Third, is the workforce being upskilled in the supervisory skills of the agent boss, quality control of AI output and critical thinking, and not only in the technical skills of building agents.
Fourth, do the incentives reward the redesign of work, releasing directly the finding that only 13 per cent of workers are rewarded for reinvention (Microsoft, 2026).
Fifth, do the agent platform and identity estate flex as fast as hybrid teams form, so that the work chart is not throttled by the platform.
Sixth, is autonomy granted only after the organisational and platform foundations are in place, rather than ahead of them.
Volge's role is in the execution. The platform decision is, for most enterprises, already shaped by their existing commitments. The disciplines that convert that capability into revenue-per-employee, work decomposition, the human-agent ratio, the upskilling pipeline, and the incentive redesign, are where transformations succeed or stall, and they are where Volge concentrates its work. Volge looks forward to discussing the application of this framework to specific operating contexts in detail.
Appendices
Appendix A: Operating-model comparison: traditional firm against Frontier Firm
| Dimension | Traditional firm | Frontier Firm | Source basis |
|---|---|---|---|
| Organisation unit | Fixed org chart; functional lines | Fluid work chart; outcome-based teams | Microsoft, 2025 |
| Labour unit | Human headcount | Hybrid human-agent teams | Microsoft, 2025 |
| Planning metric | Headcount and span of control | Human-agent ratio, set per task | Microsoft, 2025 |
| Human role | Performs tactical work | Directs, reviews, governs agents (agent boss) | Microsoft, 2025–2026 |
| Collaboration model | Person executes; tools assist | Author, Editor, Director, Orchestrator, matched per workstream | Microsoft, 2026 |
| Capability priority | Execution skills | Quality control, critical thinking, orchestration | Microsoft, 2026 |
| Adoption driver | Individual skill and effort | Culture, manager support, talent practices (67% of impact) | Microsoft, 2026 |
| Incentive design | Rewards delivery of current goals | Rewards reinvention of work with AI | Microsoft, 2026 |
| Performance metric | Output per employee | Revenue-per-employee; outcomes per hybrid team | PwC, 2026; BCG, 2026 |
| IT function role | Supplies applications | Supplies agent platform, identity, governance, data | Volge synthesis |
Appendix B: Workforce-readiness checklist
The following is a minimum readiness set for an enterprise beginning the Frontier Firm transition, drawn from Microsoft (2025–2026), BCG (2026), and LinkedIn (2026).
Operating-model design
- Outcome owners are named for each priority workstream, in the business rather than in IT alone.
- The work is decomposed into tasks before the team is staffed.
- The human-agent ratio is set deliberately for each workstream, and it is task-specific.
- Each workstream is matched to a collaboration pattern (Author, Editor, Director, Orchestrator).
- Work charts are defined for the functions being transformed, replacing the fixed org chart.
Capability and upskilling
- The agent-boss role is defined and assigned, and it is understood as a role for the broad workforce.
- Upskilling prioritises quality control of AI output and critical thinking over technical build skills.
- Role-evolution paths are mapped for the roles most affected.
- A first wave of agent bosses is trained before the first hybrid team goes live.
Adoption and change
- Managers model and incentivise AI use, consistent with the finding that manager support is a primary driver.
- Incentives reward the reinvention of work with AI, releasing the 13 per cent reward gap (Microsoft, 2026).
- A culture of experimentation is established, with space created to apply new skills.
- The Transformation Paradox is addressed explicitly: employees are given permission and reward to redesign work, not only to deliver current goals.
Platform and IT foundations
- The agent platform provisions and governs agents as quickly as hybrid teams form.
- Every agent has a first-class, governable identity.
- Observability and revocation are in place before agents act on live systems.
- The data foundation that agents act upon is accessible, governed, and reliable.
Appendix C: Glossary
- Agent boss. A person who builds, delegates to, and manages agents to amplify their impact; in Microsoft's framing, a new role for everyone (Microsoft, 2025).
- Collaboration patterns. The four patterns of human-agent work, Author, Editor, Director, Orchestrator, that define how a person and an agent divide a task (Microsoft, 2026).
- Frontier Firm. Microsoft's term for an organisation built around intelligence on tap, human-agent teams, and the agent-boss role, and defined by how deliberately it designs work across functions (Microsoft, 2025–2026).
- Human-agent ratio. The deliberate, task-specific ratio of agents to the humans who direct them within a workstream (Microsoft, 2025).
- Hybrid team. A working unit of people and agents collaborating against a shared outcome.
- Intelligence on tap. The treatment of expertise as an on-demand utility, accessed in the manner of electricity or cloud storage (Microsoft, 2025).
- Orchestrator. The collaboration pattern in which a person designs a system of multiple agents running in parallel, handling exceptions and escalations (Microsoft, 2026).
- Revenue-per-employee. Total revenue divided by headcount; the headline measure of the productivity gain a Frontier Firm seeks.
- Transformation Paradox. The tension between the pull to perform and the push to transform, in which employees fear falling behind yet find it safer to pursue current goals (Microsoft, 2026).
- Work chart. A fluid, outcome-based structure that draws the right mix of humans and agents around a goal, replacing the fixed org chart (Microsoft, 2025).
Appendix D: Sources
This paper draws on the following, accessed June 2026. Figures are attributed in-text to their primary source. Where a statistic appears across multiple secondary outlets, the most authoritative primary attribution is given.
- Microsoft: 2025 Work Trend Index Annual Report: 2025, the year the Frontier Firm is born (April 2025). Source of the Frontier Firm concept, intelligence on tap, human-agent teams, the agent-boss role, the human-agent ratio, the work chart, and the 82 per cent, 81 per cent, 46 per cent, 47 per cent, 28 per cent, and 32 per cent figures.
- Microsoft: 2026 Work Trend Index Annual Report: agents, human agency, and the opportunity for every organisation, and the accompanying Official Microsoft Blog, How Frontier Firms are rebuilding the operating model for the age of AI (May 2026). Source of the four collaboration patterns, the Transformation Paradox, the 67 per cent against 32 per cent organisational-factor finding, the 49 per cent, 58 per cent, 80 per cent, 50 per cent, 46 per cent, 65 per cent, 45 per cent, and 13 per cent figures, and the methodology of trillions of M365 signals and 20,000 workers across 10 countries.
- Boston Consulting Group: AI Will Reshape More Jobs Than It Replaces and AI Is Reshaping Jobs Faster Than Companies Are Reshaping Work (2026). Source of the 42 per cent reshaping figure, the 47 per cent, 72 per cent, and 65 per cent role-change figures, and the 10 to 15 per cent job-elimination estimate.
- PwC: 2026 AI performance study (2026). Source of the finding that the top 20 per cent of organisations capture roughly 74 per cent of AI's economic value, focused on growth rather than productivity alone.
- LinkedIn: Work Change Report and Skills on the Rise 2026, and the World Economic Forum analysis of LinkedIn data (2026). Source of the 70 per cent skills-change figure, the 50 to 55 per cent jobs-reshaped figure, the change-management figures, and the in-demand-skills data.
- Federal Reserve Bank of Atlanta: How Might AI Change the Workplace? Evidence from Corporate Executives (March 2026). Source of the 1.8 per cent average productivity-gain figure.
- Klarna, company disclosure on revenue-per-employee, as reported in 2026.
Addendum: How Volge's forward-deployed engineers shall help: Modern Work
The Volge engagement team
Volge delivers through one accountable team that runs from the boardroom to the production branch. Four roles carry the engagement, and they hold their remit from the first workshop to operational handover rather than handing off between phases.
| Role | What they own |
|---|---|
| Principal | The client relationship and the account. Sets the engagement's business outcomes, the governance cadence, and the value tracking, and stays the single point of accountability throughout. |
| Architect | The target architecture and the transformation roadmap. Sequences the work, sets the standards, and makes the platform, build-versus-buy, and control-design calls. |
| Principals | The senior engineering leads. Each owns a workstream, makes the hard technical calls, sets the quality bar, and mentors the embedded engineers. |
| Forward-deployed engineers | The embedded build. They sit inside the client's team, build and ship in the client's own environment, and own the last mile to production. |
The Principal holds the outcome, the Architect holds the design, the Principals hold the workstreams, and the forward-deployed engineers hold the keyboard, one team, accountable from strategy to shipped software.
This paper has held that revenue-per-employee gains accrue to the enterprises that redesign the operating model around hybrid human-agent teams, convert the org chart into a work chart, and upskill their people to direct agents, rather than to the enterprises that buy the most capable agents. That redesign is real-world enablement engineering, and it is executed on a workplace platform, for most enterprises, Microsoft 365. Volge delivers it through forward-deployed engineers.
What a forward-deployed engineer is
A forward-deployed engineer (FDE) is a senior Volge engineer embedded inside the client's own team and the client's own environment. The FDE does not hand over a design and withdraw. They build the agent platform and the adoption scaffolding in the client's Microsoft 365 tenant, alongside the client's people, and they stay through to a working hybrid-team operating model: the point at which humans and agents are teaming against real outcomes and revenue-per-employee is being measured rather than assumed.
The distinction the FDE model presses is the one this paper has pressed throughout. A licence is an input; behaviour change is the output. The agent-boss role, the human-agent ratio, and the collaboration patterns are operating-model decisions, and they are realised only when the underlying estate is governed, the workforce is enabled, and adoption is measured. The FDE is the person who owns that realisation inside the client's estate, not from a slide.
How the FDE helps the client
The FDE owns the real-world enablement engineering that hybrid teams depend on. Four pieces of that work sit with the FDE directly.
- The agent platform and the identity and governance estate. The FDE provisions the platform on which hybrid teams are assembled and gives every agent a first-class, governable identity, so the work chart can flex as fast as teams form without outrunning the controls.
- Copilot enablement. The FDE remediates the estate so Copilot draws only on data the user is entitled to see, then runs the per-role enablement that turns a dormant licence into a habit.
- The work-chart tooling. The FDE builds the tooling through which outcomes are decomposed, ratios are set, and humans and agents are assigned to tasks: the structure through which the work is actually done.
- Measurement of revenue-per-employee. The FDE instruments the operating model so the gain is measured at the workstream level, against the baseline, rather than asserted.
What Volge delivers
The capability set the FDE brings maps directly to the disciplines this paper has identified as the lever, drawn from Volge's Microsoft 365 and Copilot activation practice.
| Capability | Client outcome |
|---|---|
| Copilot readiness and oversharing remediation | The estate is made safe for AI before it is switched on: over-permissioned sites and Teams are tightened, broad and inherited access is removed, sensitivity labels are applied, and access reviews hold the estate in place, so an agent draws only on data the user is entitled to see. |
| Copilot adoption programming | High-value use cases are identified per role, people are trained in them, champions are built, and usage and impact are measured, so the licence is converted into a habit rather than left dormant. |
| Operating-model and role redesign | Ownership is assigned across governance, collaboration, adoption, security, and licensing, and a delivery sequence is run from governed foundation to activated AI, so autonomy is granted only after the controls are in place. |
| Change management and upskilling | Behaviour is moved through a structured journey, awareness, desire, knowledge, ability, and reinforcement, and enablement is built persona by persona, so resistance is treated as information and adoption survives the dip that follows every launch. |
| Governance and compliance | The tenant is baselined, information governance is established, and Purview, data-loss-prevention, and acceptable-use controls keep workplace AI inside the boundary, auditable, and aligned to regulatory transparency duties. |
| Employee experience | Microsoft Viva is delivered inside the flow of work, listening loops capture sentiment on a cadence, and the digital-experience layer makes the workplace feel like one connected place, so the human side of adoption is attended to. |
The adoption and transformation sequence
Volge's sequence is deliberate: the organisational and platform foundations are built first, and AI capability is activated on top only as confidence and controls mature. The order mirrors the staged rebuild advanced in this paper.
Figure A1: Volge's Modern Work adoption and transformation sequence. The cadence is conservative on autonomy and aggressive on the foundations, consistent with the staged rebuild of chapter four.
How the model helps firm and client together
The FDE relationship is built to compound, not to end at handover.
A feedback loop. Because the FDE works inside the client's estate against the client's real outcomes, what is learned in one engagement, which workstreams take to hybrid teaming, which oversharing patterns block Copilot, which ratios hold, flows back into Volge's practice and forward into the client's next workstream.
Reusable adoption accelerators. Each engagement hardens a readiness assessment, a remediation pattern, a persona use-case library, and an adoption-maturity diagnostic, so the path from a dormant estate to a measured, adopted, AI-enabled workplace shortens with every cycle. The client inherits accelerators proven elsewhere; the firm inherits the client's results.
An outcomes-based relationship. The FDE is judged on the adoption funnel, enablement, activation, habit, and impact, and on measured revenue-per-employee, not on licences deployed or hours billed. The measurement shift is the same one the paper presses: outcome over output.
The Volge difference
Four things separate the FDE model, and each answers a failure mode named earlier in this paper.
Operating-model engineering, not tool deployment. The FDE treats activation as a governance and change problem before a technical one. Ownership, the delivery sequence, and the work-chart redesign are owned jointly with business and human-resources leadership, which is the partnership this paper has argued the rebuild requires.
Human-agent teaming, grounded in the platform. Copilot and the agentic layer are provisioned, governed, and supervised on a remediated estate, so the human-agent teams of the work chart are assembled on data the organisation can trust, with oversight matched to autonomy.
Adoption discipline. Behaviour change is engineered through a structured change model, persona-based enablement, and a champions network, and it is judged on an adoption funnel rather than on licences deployed.
Embedded delivery. The FDE builds inside the client's environment and stays to the working operating model, which is what separates a delivered design from an adopted one.
Capability to outcome
The map below traces each capability the FDE delivers to the operating-model outcome it serves.
Figure A2: Volge Modern Work capabilities mapped to operating-model outcomes. Outcome figures, where cited elsewhere in Volge materials, are illustrative and should be reconfirmed against a specific estate.
A first step. A Modern Work readiness assessment, run by a forward-deployed engineer, is the recommended starting point. Over a short, bounded engagement embedded in the client's estate, the FDE plots the tenant against the adoption-maturity model, surfaces the oversharing that would otherwise block Copilot, and returns a sequenced roadmap that names the operating-model owners and the first workstream suited to hybrid teaming. The assessment is designed to be self-contained, and it commits the organisation to nothing beyond the findings. Volge would welcome the opportunity to scope it against a specific operating context.
Prepared by Volge · June 2026. This document is a point of view intended to inform executive decision-making. It is not legal, financial, or human-resources advice. Statistics are attributed to their sources and were accurate to the best of available knowledge as of June 2026; readers are advised that the agentic-AI and future-of-work fields are moving rapidly and figures should be reconfirmed before they are relied upon for organisational or investment decisions. "Frontier Firm", "Work Trend Index", and related terms are used as defined by Microsoft and are attributed accordingly.
Voice of Volge: the firm's point of view on the technology it delivers. Sourced, year-stamped, and built to be handed to a board.
Volge is an independent deep-tech enterprise service provider and premier Microsoft delivery partner. That interest is disclosed; the analysis in this book is independently sourced.