Published 24 May 2026
Illustration of a team already working with AI tools while leadership plans at a whiteboard behind them
AI & the Workforce
The VALORICE Papers · Edition 02

Your People Already Beat You To AI.

The 70% of AI value that lives in your workforce, and why embedding AI in how your people work, decide, and lead is the only version of CX transformation that pays back.
By Valérie Vanderheyden — Founder & CX Lead, VALORICE SRL
70%
of AI value comes from people, roles and ways of working
BCG 10-20-70
84%
of GenAI users self-report already exposing company data
Qualtrics
~5%
of organisations have reaped substantial financial gains from AI
BCG
88%
of managers at leaders role-model AI use, versus 25% at laggards
BCG

VALORICE Papers · Edition 02 · 2026

Valérie Vanderheyden · MBA · CCXP · LSS Black Belt · PROSCI · Scrum PSM I

Founder, VALORICE SRL — CX & AI Transformation · valorice.be

Your people didn't wait for permission

Right now, somewhere in your organisation, an employee under deadline pressure is pasting customer data into an AI tool you never approved, never trained them on, and cannot see. They are not being reckless. They are being resourceful, solving a real problem with the only AI that is actually available to them: the one they brought themselves.

This is the real state of AI in 2026. Not the boardroom slide about agentic transformation. The shadow workflow at 11pm.

Qualtrics puts numbers on it. 52% of employees now use AI at work daily or weekly, and a growing share admit to sourcing their own AI tools to boost productivity, often because their organisation is not providing a more secure alternative (Qualtrics, 2026 Employee Experience Trends Report, February 2026). On the risk side, in Qualtrics' self-reported survey data 84% of employees who use generative AI at work say they have already exposed company data through unvetted tools. Meanwhile 95% of companies report having invested in AI, and 70% of leaders say their workforce is not ready to use it (Qualtrics, The AI Readiness Paradox, April 2026).

Read those two facts together. The workforce has already adopted AI. The organisation simply has not embedded it.

That gap, between bottom-up usage and top-down operating model, is where almost all your AI value is leaking out.

— The standpoint of this paper

This paper is about closing it. And the closing move is not another tool. It is embedding AI into how your people work, decide, and lead.

The 70% nobody is funding

Here is the single most important number in the entire AI-transformation conversation, and it has nothing to do with technology.

Boston Consulting Group's 10-20-70 framework splits where AI value actually comes from, and only one of the three numbers is about technology. It is a modelled estimate from BCG's transformation work, not a measured industry average, but it is directionally consistent with what the rest of the evidence shows (BCG, AI Transformation Is a Workforce Transformation, January 2026).

  • 10% from the algorithms — the models themselves.
  • 20% from the technology to implement them — platforms, integration, data plumbing.
  • 70% from rethinking the people — roles, skills, and ways of working.

Now look at where the money and attention actually go. Into the 30%: models, platforms, licences, proofs-of-concept. The 70%, the part that drives most of the value, is treated as an afterthought. A training calendar, handed to HR after the technology is already chosen.

The results are exactly what you would predict when you underinvest in 70% of the value.

The 5% who are compounding value share one characteristic: they did not add AI to their operating model, they rebuilt the operating model for AI, starting with their people. That group shows three-year total shareholder returns roughly four times higher than AI laggards (BCG, AI Transformation Is a Workforce Transformation, January 2026).

My position is simple: the 5% put their people, the 70%, at the front of the design, and rebuilt the model around them.

— The standpoint of this paper

What embedding AI actually means

Workforce AI integration gets reduced to a prompt-engineering webinar. That is not embedding. Embedding means AI becomes part of how the organisation observes, decides, and leads, visibly, measurably, from the top.

The evidence shows what separates the organisations that compound value from the ones that stall.

The lesson is direct: AI fluency is a leadership competency, not an HR deliverable. It must be visible, measured, and modelled from the top. When managers use AI in front of their teams and in their own decisions, adoption follows. When AI is delegated to a training module, it dies in the calendar.

And the workforce is not one homogeneous group. A Cencora case study points to mapping AI adoption personas, the Protectors, the Learners, the Balancers, the Shapers, so the integration plan meets each group where they are rather than broadcasting one message to a workforce that ranges from openly hostile to already running their own shadow stack (Cencora, AI adoption persona model, CXPA CX Leaders Advance Toronto, April 2026). Remember that nearly half of CEOs describe their employees as resistant or openly hostile to the tools being deployed (Qualtrics, The AI Readiness Paradox, April 2026). You cannot train your way past that. You design around it.

The shadow AI you're worried about is not a compliance problem. It is your readiness signal, your most motivated employees telling you exactly where AI creates value, if you'd only ask them instead of blocking them.

— The standpoint of this paper

This is also why McKinsey's headline holds: the value of AI comes from rewiring how companies run, and the redesign of workflows has the biggest effect on an organisation's ability to see EBIT impact. Yet only 21% of organisations have fundamentally redesigned any workflows in response to AI (McKinsey, The State of AI: How Organizations Are Rewiring to Capture Value, March 2025). Rewiring how people work is the workforce embedding. They are the same act.

Why bolting AI on fails

If the workforce is the engine, the method is the chassis. And most transformation methods were built before AI existed.

The standard enterprise transformation runs in a fixed sequence: map the current state (AS-IS), design the future state (TO-BE), analyse the gap, build the roadmap, implement. AI gets added at the end, a productivity accelerator bolted onto a method that was already finished producing its outputs.

That is the wrong architecture. It is the structural reason most AI programmes fail to pay back. You cannot embed AI in your workforce while running it through a method that never imagined AI in the first place.

The proof is in the experiment-to-scale gap: 62% of organisations are running agentic AI experiments, but fewer than 10% have scaled one enterprise-wide to deliver tangible value (McKinsey, Building the Foundations for Agentic AI at Scale, April 2026). The gap between experiment and scale is an architecture problem, not a technology problem. The tools work. The model around them does not.

And the clock is real. Gartner forecasts that 40% of enterprise applications will feature task-specific AI agents by the end of 2026, up from less than 5% in 2025 (Gartner, 40% of Enterprise Apps Will Feature Task-Specific AI Agents by 2026, August 2025). Forrester goes further: enterprise applications are moving from enabling employees to accommodating a digital workforce of AI agents, virtual team members to be managed rather than features to be deployed (Forrester, Predictions 2026: Enterprise Software, November 2025). You will be managing humans and agents in the same operating model within 18 months. A bolted-on method cannot hold that.

AI as a lens at every stage: the AS-IS and the TO-BE

AI moves from a workstream at the end to a lens at every stage. Each stage below shows the standard approach and the AI-embedded alternative, and the workforce thread runs through all five.

Stage 1, AS-IS: add one column to every process. The standard approach documents workflows, handoffs, volumes, and pain points. The AI-embedded approach adds a single column to every process thread, AI Adjacency: is this task automatable, augmentable (human plus AI assistant), or must it remain human? Generative AI can now convert process documentation into BPMN and swimlane diagrams automatically, detecting bottlenecks as it goes. The AI-adjacency score becomes the primary input for roadmap sequencing, and the foundation for simulation before any change is committed. This is also where you surface which tasks your people are already augmenting with their own shadow tools.

Stage 2, TO-BE: design AI-native, then design the human roles. The AI-embedded approach designs the TO-BE as an AI-native operating model and works backward to the human roles needed to govern, refine, and elevate it. AI becomes a decision layer across the entire technology stack, across the whole customer journey and not just the contact centre (Orange Business, CX Trends for 2026, March 2026). Static journey maps give way to dynamic, AI-orchestrated models, what BPM leaders now call operational digital twins: real-time simulation of what will happen, not just documentation of what exists.

The gap, the roadmap, and the end of change management

Stage 3, Gap Analysis: run an AI Readiness Gap in parallel. Alongside the usual gap, run a parallel AI Readiness Gap across BCG's three prerequisites: strategic alignment from the top, adoption architecture across the workforce (the AI persona mapping from Part Two), and an AI-enhanced operating model that redefines roles around human-AI collaboration (BCG, AI Transformation Is a Workforce Transformation, January 2026). Include a shadow AI audit. The employees already sourcing their own tools are not a risk-register entry, they are a map of where demand is highest (Qualtrics, 2026 Employee Experience Trends Report, February 2026).

Stage 4, Roadmap: sequence by task chains, not milestones. The breakthrough insight comes from MIT: the path to AI-driven productivity runs through deconstructing existing and emerging tasks to understand which activities can be substituted, augmented, or transformed (MIT Sloan Management Review, Want AI-Driven Productivity? Redesign Work, April 2025). The principle is task adjacency: if even one step in a chain is poorly suited to AI, it breaks the whole chain. So you build the roadmap around task-chain sequencing, clustering AI-compatible steps to eliminate friction at every human-AI handoff. Sequencing priority goes to the cluster that removes the most costly handoffs, not the most technically impressive use case.

Stage 5, Implementation: replace change management with continuous workforce integration. Stop running AI as a change-management project that ends. Run it as continuous AI workforce integration that never ends: upskill more than 50% of employees, have 88% of managers model AI use, and treat AI fluency as a measured leadership competency (BCG, AI Transformation Is a Workforce Transformation, January 2026). The 70% people dimension is the work, not the wrap-up.

Then a decision sandbox opens

There is one more reason embedding AI early changes everything: it makes a new capability possible, one big enough to deserve its own paper. When your entire customer database is fitted into a robust segmentation represented by living AI personas, you get a decision sandbox. Any change to pricing, product, packaging, messaging, service line, or channel can be pressure-tested against the full complexity of your customer base before a single real euro is committed.

The evidence that this works at enterprise scale is already strong. Synthetic customers trained on first-party data replicate roughly 90% of key outcomes from traditional research (Bain & Company, Synthetic Customers Earn Their Stripes, May 2026), and early enterprise adopters such as Unilever are already validating campaigns against AI-simulated consumer response before launch, moving from first insight to live execution in 6 months (Unilever, How AI Is Transforming Unilever's Personal Care Business, December 2025).

This is the subject of VALORICE Papers Edition 03, which unpacks those results in full. For Edition 02 the point is narrower, and it closes the loop on workforce: simulation is a capability your embedded, AI-fluent workforce operates. Without the people in Part Two, the sandbox is just another tool nobody is ready to use.

Workforce is the first design input

If 70% of the value is in the people, workforce embedding is not the last step of transformation. It is the first design input.

— The equation this paper proposes

AI-Embedded AS-IS + AI-Native TO-BE + AI Readiness Gap + Task-Chain Roadmap + Simulation-Led Decisions + Workforce AI Integration = Transformation That Compounds.

Every term is a place where AI is embedded into the method, and the workforce term is the one that makes the rest pay back.

Three things to do on Monday

  1. Treat workforce AI readiness as a design input, not a change output. Map your AI adoption personas before the roadmap. Surface and govern shadow AI: it is your demand map, not your risk register. Make AI literacy a visible, measured leadership behaviour modelled from the top (BCG, AI Transformation Is a Workforce Transformation, January 2026).
  2. Add one column to every AS-IS session: AI Adjacency. Automatable? Augmentable? Human-only? That single column reshapes your TO-BE design, gap priorities, roadmap sequencing, and workforce plan (MIT Sloan Management Review, Want AI-Driven Productivity? Redesign Work, April 2025).
  3. Make AI fluency a measured leadership behaviour this quarter. Require managers to use AI visibly in their most frequent decisions, then measure it. The 88% versus 25% gap on managers role-modelling AI is the single most actionable lever you have (BCG, AI Transformation Is a Workforce Transformation, January 2026).

Your people have already voted

Your people are already using AI, whether you have sanctioned it or not. The organisations pulling four-times-greater shareholder returns are not the ones with the best tools. They are the ones that embedded AI into how their people work, decide, and lead (BCG, AI Transformation Is a Workforce Transformation, January 2026). It does not require a bigger budget; it requires putting the 70% at the front of the design instead of the back of the rollout.

AI is not a tool you deploy after you finish designing transformation. It is the lens through which every step is redesigned, and above all a capability you embed in your workforce, because that is where 70% of the value has been waiting the whole time.

— The closing standpoint of this paper

Download this edition as a PDF →

Valérie Vanderheyden, Founder and CX Lead at Valorice
Valérie Vanderheyden
Founder & CX Lead, VALORICE SRL
Valérie Vanderheyden is Founder & CX Lead of VALORICE SRL, a Belgian boutique consultancy that helps organisations remove the friction that stalls customer journeys and turn it into measurable growth, retention and adoption. She works across diagnostics, board-ready blueprints and operating tools teams can run on Monday — in three languages, one delivery. CCXP, MBA, PROSCI, PSM I, Lean Six Sigma.

References