How to Unstick Your AI Transformation

Introduction

Pilots are running. Employees are trying tools on their own. Leadership has announced an adoption goal. And yet the work itself looks much the same as it did a year ago.

The gap is real. McKinsey's 2026 survey found that 44% of respondents report enterprise AI scaling, but only 37% report a positive EBIT contribution. Activity is not impact.

The cause is rarely the tool. More often the blockers are unclear outcomes, uncertain roles, weak ownership, low trust, thin support, and managers carrying the change alone. This article shows how to diagnose those blockages and lays out a five-stage path to move past them.

Key Takeaways

  • AI stalls when it is run as a tech rollout instead of a change in how people work.
  • Diagnose clarity, ownership, and support before buying more tools or training.
  • Align, diagnose, redesign, experiment, then scale and measure.
  • Middle managers need decision rights, language, time, and backing.

Why AI Transformations Get Stuck in the First Place

Experimentation is not transformation

Experimentation produces isolated use cases and individual time savings. Transformation changes end-to-end workflows, roles, decisions, and outcomes. Many organizations stay in the first mode and assume they are in the second.

Common symptoms of a stall:

  • Scattered pilots with no shared goal
  • Wildly different usage from team to team
  • Repeated requests for more training
  • Managers giving conflicting guidance
  • Employees quietly sidestepping new workflows

The micro-productivity trap

The trap appears when you speed up individual tasks without redesigning the process around them: handoffs, incentives, responsibilities. McKinsey found nearly three-quarters of high performers fundamentally redesign workflows, versus about one-quarter of others.

BCG's 2026 research found 66% of employees get limited or no guidance on using the time AI saves, and more than half don't reinvest it in strategic work.

Hesitation is often legitimate

Hesitation is not always resistance. Pew found 52% of US workers are worried about AI's future impact in the workplace.

Slack's 2024 survey found 48% would be uncomfortable admitting they used AI for common tasks, often fearing they'd look less competent. People are asking fair questions about expertise, identity, workload, and who answers for the output.

Gallup's May 2026 indicator adds a clarity problem: only 25% of US employees report a clear plan for integrating AI into current practices.

Technology-first versus people-first

Technology-first response People-first response
More licenses and prompts Clear reason for the change
One-time training event Defined shifts in how work changes
IT owns the rollout Visible decision ownership
Usage dashboards Ongoing support and feedback

Technology-first versus people-first AI transformation response comparison

Diagnose the Blockage Before Adding More Tools

Start with a people-side reality check

Ask employees and managers four questions. Can they answer them?

  1. What is changing?
  2. Why is it changing?
  3. What remains human-owned?
  4. Where can people get help while they learn?

If answers vary by team, you have found your first problem. Tool-use data won't show it. Build the real picture through:

  • Focused conversations
  • Manager listening sessions
  • Workflow observation
  • Open employee feedback

Find the type of friction

Separate two kinds of friction.

  • Identity and trust friction: people fear becoming replaceable or feel their expertise is being devalued.
  • Capability friction: people lack the skills or the time to use AI responsibly.

Then look at leadership and sequencing:

  • Who sets priorities?
  • Who owns outcomes?
  • Who approves exceptions?
  • Who supports managers?
  • What happens when AI output is wrong?

Non-use can also signal relevance, not defiance. Gallup's 2025 survey found 44% of non-users don't believe AI can help with their work. Nobody has shown them where it fits.

Turn the diagnosis into a focused decision

Pick one high-value workflow where a clear business outcome meets a clear people outcome. Reducing avoidable administrative burden while preserving professional judgment is a good example.

Christina Muller's COUNTED at work™ Score is one diagnostic option. The free, five-minute assessment measures leaders' clarity about their role in AI adoption, their sense of ownership, and perceived organizational support. It gives HR and executives a baseline and flags where leaders feel stretched or unsupported.

AI transformation blockage diagnosis framework for leaders and teams

A Five-Stage, People-Centered Path to Unstick AI Transformation

This is a practical working model, not an industry standard. Scale each stage to your organization's maturity and risk level. A health system touching clinical workflows needs tighter guardrails than a firm piloting meeting summaries.

Stage 1: Align on the outcome and the reason for change

Name the business problem, the desired outcome, and the human impact before choosing a tool. State what success should improve for customers, employees, managers, or the mission.

Stage 2: Diagnose readiness, clarity, ownership, and support

Set a baseline:

  • What do employees understand?
  • Where do managers feel unprepared?
  • Which roles are most affected?
  • Where is accountability fuzzy?

Stage 3: Redesign roles, decisions, and workflows together

Map the current workflow end to end. Then specify:

  • Which tasks AI will assist
  • Which decisions stay human-owned
  • What judgment people must keep practicing
  • How responsibilities shift

Stage 4: Run supported experiments that create learning

Give teams protected time, clear guardrails, feedback channels, and permission to surface problems. Use regular demos, peer learning, and manager coaching rather than one launch-day session.

BCG's 2025 research links regular use with at least five hours of training and access to in-person coaching. That is an association, not a universal threshold.

An employer example: NewMark Merrill ran a four-hour foundation session followed by a six-session applied program for 13 cross-department participants. The company reports 3,783 annual hours of recovered capacity, which is employer-reported rather than independently verified.

Five-stage people-centered AI transformation implementation path

Stage 5: Scale what works and revisit what does not

Hold a recurring review covering:

  • Adoption quality
  • Workflow outcomes
  • Employee experience
  • Capability preservation
  • Decision ownership

Based on evidence, stop, adapt, or expand each initiative.

What Leaders and Managers Must Do Differently

The executive role

Executives own the conditions for adoption:

  • Set a focused ambition
  • Use AI responsibly in public
  • Explain what is changing and why
  • Protect learning time
  • Make ownership visible

Delegating the whole effort to IT leaves HR handling consequences it had no hand in shaping. Gartner calls work redesign a shared CHRO-CIO challenge.

The middle-manager role

Managers carry the day-to-day load:

  • Translate strategy into team expectations
  • Spot workflow friction
  • Answer employee concerns
  • Reinforce learning
  • Escalate risks before confusion turns into disengagement

They are often under-supported. Gallup found only 36% of employees in AI-integrating organizations strongly agree their manager supports team AI use. Those who do are 1.7x as likely to use AI several times a week (an association, not proof of cause). Separately, 50% of surveyed CHROs lack confidence in their managers' AI guidance.

Carrying the Rollout™ workshops, offered as half-day or full-day sessions, give managers practical support. They help leaders sort what they own, what to hand back, and who decides through each wave of AI change.

Psychological safety and participation

The people closest to the work know where AI helps and where it breaks. Invite that knowledge, and say plainly that sharing it will not make anyone expendable.

If people fear that explaining their process is the first step to being replaced, they hold back or hide their AI use. KPMG and the University of Melbourne found 57% of workers hide AI use and present AI-generated work as their own.

How to Measure Progress and Keep Transformation Moving

Activity is not progress

Logins, prompt volume, and pilot counts show activity. Microsoft's adoption guidance cautions that high usage doesn't necessarily mean better work or results. People can use AI without changing the steps that matter.

Build a balanced set of measures

Track across these areas, without borrowing universal targets:

  • Business outcomes and workflow performance
  • Adoption quality, not just frequency
  • Role clarity and manager confidence
  • Learning time and capability
  • Employee concerns and trust
  • Preservation of human judgment

BCG's upskilling guidance uses the Kirkpatrick levels: learning experience, demonstrated competency, behavior change, and business results. It also recommends controlled pilots or A/B tests, since attribution gets murky when several things change at once.

Review on a rhythm

Leadership should ask each cycle:

  1. What is working?
  2. What are employees learning?
  3. Where is ownership unclear?
  4. What risks are emerging?
  5. Should we scale, redesign, pause, or stop?

Organizations without internal change-management capacity can use outside help. COUNTED at work™ is a 90-day advisory engagement that turns a people-side diagnosis into a clear, defensible plan for the CEO and executive team.

Frequently Asked Questions

What are the five stages of AI transformation?

A practical working model is: align on outcomes, diagnose clarity and support, redesign roles and workflows, run supported experiments, then scale through continuous learning and measurement. It is a working framework, not an industry standard.

Why do AI transformation initiatives get stuck?

Leaders focus on tools and pilots while outcomes, ownership, role changes, trust, manager support, and learning expectations stay unclear.

How can leaders tell whether an AI adoption problem is really a people problem?

Look for uneven adoption, repeated confusion, quiet resistance, conflicting manager guidance, fear about expertise or job security, and no clear accountability for AI-supported decisions.

What role should HR play in AI transformation?

HR should help define the people roadmap, clarify role and capability changes, support managers, build feedback channels, and make sure adoption strengthens trust.

Why are middle managers so important to AI transformation?

They turn enterprise strategy into daily priorities, workflow decisions, coaching, and employee support. Without them, adoption rarely sticks.

How should organizations measure whether AI transformation is working?

Combine business outcomes, workflow performance, adoption quality, role clarity, learning, manager confidence, trust, and decision ownership. Tool usage alone isn't enough.