The actual problem

Why AI implementations underdeliver

Most AI projects fail not because of the model — because of what was handed to it, and what was asked of the people using it. Three failure modes, occurring independently or together.

Failure mode 01

The data environment

Data that is technically clean but built for reporting — not reasoning. AI applied to this data produces outputs that are fluent, confident, and wrong in ways that erode trust before adoption can begin.

Failure mode 02

The organizational change layer

Change is announced, not managed. Awareness reaches leadership; it never reaches the person whose daily workflow has to change. Adoption plateaus at 30 days and quietly reverses by 90.

Failure mode 03

The individual competency gap

Everyone attends the same training. An executive, a manager, and a front-line operator need fundamentally different AI capabilities — and develop them at different rates. Uniform enablement produces uneven results.

The organization that addresses all three builds a compounding advantage. The one that addresses one — or none — repeats the same failed deployment on a longer timeline. Guided Change closes all three gaps.

The three-layer architecture

Each layer is distinct. Each is required.

None is sufficient alone. Guided Change integrates all three as a connected methodology — not three separate workstreams.

Layer 01 — Data foundation

Guided Discovery

The data environment AI needs to perform

Maps the gap between data built for reporting and data built for reasoning. Identifies the decisions that matter most, examines what data is ready for AI reasoning today, and produces a prioritized roadmap for the data preparation AI deployment requires. Conducted as a structured sprint — typically two to four weeks — before deployment begins.

Without this layer: AI produces fluent, confident outputs that don't reflect the organization's actual situation. Trust erodes before adoption begins. Clean doesn't correlate with correct.
Read the data layer methodology

Layer 02 — Organizational change

ADKAR

The change journey, structured and measurable

The Prosci ADKAR model — Awareness, Desire, Knowledge, Ability, Reinforcement — governs how an organization moves through change. Applied here to AI adoption: each stage has a defined failure mode, a set of interventions, and measurable signals that determine whether the organization has genuinely passed through it or only appeared to. ADKAR applies to any organizational change. Guided Change applies it specifically to the human and process adoption of AI tools, models, and agent workflows.

Without this layer: Training events are delivered. Behavior doesn't change. Adoption metrics plateau at 90 days. The initiative gets declared a success the week before it quietly reverses.

Layer 03 — Individual competency

4D AI Fluency

What each person needs to develop, by role

The Framework for AI Fluency — four interconnected competencies: Delegation, Description, Discernment, and Diligence — maps what individual practitioners need to develop to work effectively, efficiently, ethically, and safely with AI. Applied within ADKAR, the 4D framework differentiates the competency development required at each stage of the change journey by role. An executive's fluency gap is not a front-line operator's fluency gap. Uniform enablement misses both.

Without this layer: Everyone receives the same training. Power users emerge accidentally. High-risk users — those with strong task fluency but weak evaluative judgment — go undetected until an error surfaces.

How the three layers connect

Guided Discovery surfaces what the organization actually needs AI to answer — and where the data gaps will undermine that. ADKAR structures how the organization moves through the change required to close those gaps and adopt new AI-assisted workflows. The 4D AI Fluency framework maps what each individual needs to develop at each stage of that change journey, differentiated by role, function, and current capability. The data layer informs the change. The change framework governs the journey. The fluency framework develops the people.

Layer 02 in depth

ADKAR applied to AI adoption

Each stage of the ADKAR model has a defined question, a failure signal, and a 4D competency development priority. The stages are sequential — an organization cannot skip Desire and expect Knowledge to land.

A

Awareness

"Why do we need to change at all?"

4D priority

Description — the "why change" message must be articulated differently for each role. A managing partner needs competitive benchmarks and margin impact. A front-line operator needs a job-security narrative. The same message fails both.

Failure signal

"We've always done it this way." Rumors filling the vacuum where communication should be. Leadership has been briefed; no one else has.

D

Desire

"What's in it for me personally?"

4D priority

Discernment — the practitioner must read whether buy-in is genuine or performed. Surface compliance with no behavioral intent is indistinguishable from real Desire until the training ends. Misreading this is the most common ADKAR failure.

Failure signal

People attend every session and say the right things. Usage data at day 14 shows they've logged in once. Awareness was achieved; Desire was not.

K

Knowledge

"How do I actually do this?"

4D priority

Description and Discernment in tandem. A practitioner who develops strong task description without output evaluation becomes a high-confidence, low-accuracy user. In regulated or high-stakes environments, this is more dangerous than not adopting at all.

Failure signal

Enthusiastic adoption at week two. Workarounds appearing by week four. People can operate the tool — they cannot yet evaluate what it produces.

A

Ability

"Can I do this consistently?"

4D priority

Discernment and Diligence — Ability in an AI context is not fluency of interaction. It is accuracy of evaluation. Can the practitioner tell the difference between a correct output and a plausible-sounding wrong one? That is the measurable bar.

Failure signal

Training completion rates are strong. Usage data shows the tool is open. Output quality reviews show the same errors recurring — no one is catching them before they propagate.

R

Reinforcement

"What keeps this from slipping back?"

4D priority

Diligence at scale — the risk at this stage is normalization. Habitual use of AI outputs without the careful evaluation that characterized early adoption. Reinforcement must institutionalize the standard, not just monitor compliance with it.

Failure signal

Strong adoption metrics at month one. Plateau at month three. Regression by month six. The new behavior was never structurally reinforced — it was performed while attention was high.

The ADKAR model was developed by Prosci and is the foundational change management framework applied in Guided Change engagements. Sun Business Group applies ADKAR specifically to the context of AI tool, model, and agent adoption — not as a generic organizational change framework, but as a structured lens for the specific human and behavioral dynamics AI transformation requires.

Layer 03 in depth

4D AI Fluency — what each person needs to develop

Four interconnected competencies. Not a hierarchy or a sequence — a system. Strength in one without the others creates specific failure patterns that are predictable and preventable.

D

Delegation

Creative vision and selection of the right AI tools and modalities to realize that vision. Knowing when to automate, when to augment collaboratively, and when to configure AI for independent agency — and which tasks belong in each category.

Without it: AI is used as a faster search engine. The highest-value workflows — the ones that compound — remain untouched.

D

Description

Effectively communicating ideas, requirements, and constraints to AI systems. Crafting prompts that produce useful outputs — and iterating dialogically when they don't. The skill most directly developed through structured practice.

Without it: The practitioner gets generic outputs and concludes the tool doesn't work. The tool was never given what it needed to work.

D

Discernment

Critical evaluation of AI-generated outputs — quality, relevance, accuracy, and bias. The ability to tell the difference between a correct output and a plausible-sounding wrong one. The competency that determines whether AI adoption helps or creates liability.

Without it: High-volume, low-accuracy use. Errors propagate with confidence. In regulated contexts, this is the highest-risk competency gap.

D

Diligence

Responsible use of AI — ethical practice, transparency about AI involvement, and taking accountability for AI-assisted outputs. Fact-checking, accuracy validation, and vouching for what gets used. Not a policy — a practiced habit.

Without it: Habitual use without verification. The practitioner stops checking because outputs usually seem fine. This is the failure mode Reinforcement exists to prevent.

The Framework for AI Fluency is authored by Rick Dakan (Professor of Creative Writing, AI Coordinator, Ringling College of Art and Design) and Joseph Feller (Professor of Information Systems and Digital Transformation, Cork University Business School, University College Cork). The framework is published under CC BY-NC-ND 4.0. The current version and full documentation are available at ringling.libguides.com/ai/framework. Sun Business Group applies this framework within AI adoption engagements as a structured competency development layer within the ADKAR change journey.

Applied across sectors

The framework in practice

Guided Change applies across any organization undergoing AI-driven transformation. The three layers are constant. The role profiles, ADKAR interventions, and 4D development priorities shift by sector and function.

Context

Enterprise cybersecurity — AI adoption across 2,000+ employees

A large enterprise technology organization deploying AI across security operations, sales, and customer success functions simultaneously. Multiple internal stakeholders with competing priorities. L&D infrastructure exists but lacks AI-specific fluency frameworks. Executive sponsor has approval; adoption has stalled at the management layer.

Guided Change application

Layer interaction

  • Data layer: Security data environments have strict access controls that create data preparation constraints unique to the sector. Guided Discovery surfaces these before deployment begins.
  • ADKAR: At enterprise scale, the management layer is the critical failure point. Middle managers who are themselves not fluent become blockers of team adoption — their Desire gap is the leverage point.
  • 4D: The engagement typically begins with Delegation and Diligence for leadership, then builds Description and Discernment for the operational teams who use AI daily.

Sector example — representative of enterprise technology deployment contexts.

Context

GTM / Revenue teams — Sales, Marketing, Customer Success

Growth-stage B2B organizations ($10M–$50M ARR) deploying AI across commercial functions. Sellers and CSMs have high AI enthusiasm and low AI discipline. Marketing teams are adopting AI tools in silos. Revenue operations teams are attempting AI-assisted forecasting on data that was never built for reasoning.

Guided Change application

Layer interaction

  • Data layer: CRM data is almost always built for reporting — deal stages, close dates, activity logs. Guided Discovery reveals the gap between what the CRM contains and what AI needs to produce reliable pipeline analysis or renewal risk signals.
  • ADKAR: GTM teams move fast and lose patience with structured change programs. The ADKAR application here emphasizes speed to Ability — quick wins that build genuine fluency, not training that feels like compliance.
  • 4D: Sellers need Delegation (which tasks to give AI) and Discernment (evaluating AI-generated outreach or proposals). The Diligence gap — sending AI-generated content without review — is the most common and most reputationally costly failure mode.

The most developed proof point in the Sun Business Group portfolio — from individual seller enablement to enterprise GTM transformation.

The practitioner commitment

Committed, not just invested

A consulting firm's reputation survives a failed engagement. A named practitioner's does not. That asymmetry is not a liability — it is the reason the work lands differently.

"Most practitioners can teach AI fluency or deploy it operationally. Few can do both — and fewer still have run the commercial functions they are now helping transform."

John Williams is the principal of Sun Business Group and the operating entity behind Guided Change engagements. The methodology is not a framework handed to a delivery team — it is applied directly, by a practitioner who has carried quota, managed churn, owned forecasting, built comp plans, and integrated acquisitions. That operator depth is what distinguishes framework knowledge from deployment capability.

For enterprise organizations, Guided Change can be delivered as a scoped project engagement, a fractional change leadership role, or an interim AI adoption function — depending on the scale of change required and the internal capacity already in place.

AI
Anthropic Partner Program — fxops.ai Member of the Anthropic Partner Program. AI Fluency for Educators certified. Hands-on Claude API and MCP deployment experience.
CH
ADKAR — Prosci framework Applied across post-acquisition integrations, enterprise sales transformations, and multi-function AI adoption programs at organizations from $20M to $350M+ ARR.
AI
Pavilion AI School — instructor Faculty instructor, "AI Tools in Action" — Pavilion executive community of 1,420+ GTM leaders. AI fluency developed at practitioner scale, not classroom scale.
OP
Commercial operator background E2open (40 sellers, 4 BUs, 90%+ renewal, 107% net growth). N3/Accenture (global scope, 5 clients). LivePerson, REW, Checkmarx (140 global sellers). The functions being transformed are functions that have been run.

The first conversation is a fit assessment, not a sales pitch

Guided Change works when the organization is ready to address all three layers. The discovery call determines whether that's true — and if so, where to start.

Book a Discovery Call View Services