.svg){kind=logo}
Introduction
Not more tools. A system that builds real AI capability. This is the playbook powering the companies who are actually getting value from AI and maturing in their AI journey.
Over the past few years, I have had the opportunity to watch how companies have both adopted AI successfully and wasted time and money in failed attempts. After working with dozens of organizations, from boardroom strategy sessions to real-world implementations, and conducting primary research on over a hundred organizations, I have seen a remarkably consistent pattern.
Buying AI tools doesn't make you an AI-mature company, any more than buying a gym membership makes you fit.
The most common story looks like this: a company has purchased Copilot licenses, maybe added a few other "AI-enabled" tools, formed an AI committee, and yet 50% of employees are still using free ChatGPT, and the other 50% are frustrated that they can’t get AI to work for them. Leadership feels they have checked the AI box. Almost no additional business value has been created.
This article is about why that happens, and what you can do instead. It is not a story about tools. It is a story about whether your organization is using the right playbook. Your AI Playbook will impact how you decide where to use AI, how you prioritize investments, and how to turn experiments into lasting impact.
What the Data Shows
The Widening Gap
When we ask executives to rate their AI maturity on a 0–10 scale, most land somewhere in the middle. Very few say zero; very few say ten. But there's a strong pattern in how they think about future investment: companies that already feel more mature when it comes to AI are planning to invest more aggressively, while companies that feel behind are more cautious, stuck in "wait and see" mode. The gap is not closing. It is widening.
From a distance, it can look like everyone is investing in AI. Up close, only a small set of companies are building real capability while a much larger set are simply adding AI to their budget. This investment typically lives as a one-off, disjointed AI initiative. Our research shows that without an internally aligned strategy, these initiatives typically die in the pilot stage. This sadly reveals that many organizations are running with the wrong playbook.
The Copilot Paradox
One finding surprised me more than any other: the mismatch between what companies buy and what employees actually use. Many organizations have made large top-down bets on Microsoft Copilot or similar enterprise tools.
At the same time, a significant portion of individual contributors still say their primary AI tool is ChatGPT, often the free version.
The patterns we see repeatedly:
- Copilot rolled out company-wide with no structured training on how it fits into daily work
- No clear expectations through performance reviews on AI usage in a given role
- Security fears around public tools, but no equivalent "safe, fast" alternative that people genuinely prefer
Executives assume they've checked the AI box because licenses have been purchased. But buying Copilot is not the same as adopting AI. The tool that wins the procurement cycle isn't always the tool that wins the workday.
The Strategy Gap
In private conversations, some version of this sentence comes up again and again: "We know we need to do something with AI. We're just not sure what, or how to sequence it." The biggest barriers sound like: we don't have a clear AI strategy; we're not sure where the real value is; we're nervous about data, security, and regulation. It's not that leaders don't care. It's that they don't yet have a playbook for AI. They have a shopping list.
The Four Traps
When executives first try to "get into AI," they usually fall into one of four patterns. None are malicious. They are simply what happens when you start with tools instead of with a system.
Side-of-the-desk AI. You encourage people to experiment when they have time. A few enthusiasts dive in. Most don't. Nothing mission-critical moves. The result is scattered small wins with no compounding learning.
Subscription sprawl. You start buying AI features embedded in every SaaS product your team touches. Each tool solves a tiny slice of the problem; nobody learns any of them deeply. The result is more logins, more confusion, and very little measurable impact.
Big-consulting theater. You hire a large firm. They run a maturity assessment, map your "AI gaps," and suggest you embrace "agents" as the future. The result is polished slides and some excitement. Twelve months later, those slides haven't changed how your core workflows run.
Throwing it over the wall to IT. You tell IT to own AI. They experiment with cool tools and find a big, hairy, data-intensive problem to solve. It involves 6 months of work rewiring the ERP or data lake before they can get started. Nothing your teams could actually use to be more productive or earn new business is even discussed. You have no idea how AI is being used, but IT promises that AI projects are under way. They worry about risk, and struggle to connect technical possibilities to frontline work and the P&L. The result is tools that don't move the needle.
The Three Layers
When we come into an organization, we always try to get the senior leadership team on the same page about AI. Almost always, they aren't. We start by asking: Who is supposed to be better at their job because of AI, specifically? What are you telling your organization about why AI matters? What departments and workflows are you willing to change?
Over time, we've found it useful to frame this as a simple AI playbook made of three layers:
- Personal Productivity Layer: how individuals use AI in their day-to-day work
- Workflow Transformation Layer: how teams redesign core processes with AI
- Owners/C-Suite, Data & Policy Layer: how leadership sets direction, guardrails, and backbone infrastructure
Most AI products are marketed as if they magically cover all three and they will also work effectively with whatever you throw at it. In reality, you have to design for each layer on purpose.
Layer One: Personal Productivity
This is where most AI journeys begin: people using assistants like Copilot or ChatGPT to draft, summarize, and analyze. The problem is that in many companies, this layer is completely informal, with no guidance on what "good" use looks like. It's unequal, with a few power users while everyone else feels behind. And it's disconnected from strategy, with no link between individual use and business outcomes.
When we run training programs or audits, we usually find three opportunities:
- Make existing tools twice as useful. Most teams are using a small fraction of what their core systems can do, AI features included. Deepening usage here is often a fast win.
- Give people patterns, not just access. Show them concrete approaches that work in their role: how to draft internal business cases, how to summarize meetings and turn them into action, how to do first-pass data analysis.
- Create a safe, sanctioned way to use assistants. Be explicit about what can and can't be shared with AI systems, and which tools are preferred for which types of work.
You don't need a custom model to double the impact of a motivated project manager or analyst. You need clear guidelines and training and set the expectation that people need to use it.
Layer Two: Workflow Transformation
The second layer is where AI moves from "cool tool" to actual leverage: the level of workflows, not tasks. This layer is owned by managers and directors, the people responsible for business development, project scheduling, customer service, AP/AR and billing, compliance, and reporting.
From our experience and extended research, we have determined that the projects that consistently generate real ROI share a few traits:
- They focus on a single, clearly defined workflow (that can be extended with further ways AI can help)
- That workflow is critical, and happens hundreds or thousands of times a year (this ensures there’s ROI)
- There's already a sense from directors that "if we could fix this, it would really move the needle" (existing pain points)
Examples include a system that drafts and standardizes project documentation so senior staff only review and tweak, an assistant that triages customer inquiries to the right team with context attached, or a tool that reads invoices, catches anomalies, and passes to another system.
The key at this layer: Don't start by asking "What can AI do?" Start by asking "Where are the friction, delay, or error-prone workflows today?" and then "What parts of this could AI reliably support?"
Layer Three: Owners/C-Suite, Data & Policy
Owners and C-Suite often experience AI in two modes: as a hype topic they're expected to talk about, and as a risk they're expected to contain. Both are incomplete, and both ignore the third piece that determines whether AI works over the medium- to long-term: your data architecture.
At this layer, we want to bring together a long-term horizon of success in an organization which includes direction (where AI is meant to change how the business works), data design (how you shape and govern data so AI can keep getting smarter), and policy or governance (how you keep customers, employees, and the company safe while you move fast).
When we work with executive teams, we push them to think of data not as a back-office IT concern, but as the fuel and memory of their AI strategy. Practical questions surface:
- If the pilots we're running today actually work, will our current data architecture support scaling them?
- Do we know which systems are "systems of record" vs "systems of engagement," and how AI should interact with each?
- Are we designing new workflows so they create cleaner data we can use in the future?
You don't need a perfect data warehouse to start using AI. But you do need to be intentional about choosing a small number of "AI-critical" data domains with clear owners, aligning access and security with how AI will actually be used, and designing new AI solutions so they improve the data over time.
Without this layer, the other two stay fragile and will see diminishing returns over time. Individual productivity wins remain personal hacks. Workflow improvements stay one-off projects. With it, every experiment has somewhere to land to build a sustainable advantage with AI in the future.
Putting It Into Practice: A 90-Day Reset
When an executive team talks with us, we often lean them toward a version of a reset because they are missing a critical piece in their AI efforts. Then, we recommend a 90-day sequence that touches all three layers.
Week 0: Find the Truth
Before you convene leadership, get as close as possible to what's actually happening with AI in your organization. Don't rely on what people tell you in meetings. Walk the floor, sit with employees, and ask them what they did with AI last week. You'll be surprised how different the reality looks from the executive summary.
Better yet, issue a structured AI Maturity Survey across the organization. A good survey reveals the truth about how people are using AI tools, what they're using them for, and where the gaps are between what leadership thinks is happening and what's really going on. It also surfaces risk: employees using unsanctioned tools, sharing sensitive data with public models, or working around policies that don't match the reality of their jobs. The findings from Week 0 will directly inform your enablement program and help you update policies before they become liabilities.
Week 1: AI Leadership Working Session
Not a keynote. A working day with three outcomes:
- Shared understanding of what today's AI can and can't do
- Alignment on the top use cases where AI investment would make the biggest difference
- A lightweight intake process for AI ideas from the rest of the organization
By the end of the day, you want something you can say out loud and everyone nods their head: “For the next twelve months, here's where AI matters most to us, here’s why we are doing it, and here's how we'll make decisions on investment.”
Weeks 2–12: One Focused Workflow Sprint
Pick a workflow in one to two functions and run an eight-to-twelve-week sprint to redesign it with AI. Scope the project ruthlessly: one department, one workflow, one measurable outcome. Build something that goes into real use, even if it's not perfect. After implementing dozens of these systems, they are never perfect, but they are always representative of a tangible movement in the way you work and aligned beliefs that more is possible with AI. The goal is less “the perfect solution” and more “proof that we can change how work happens with AI.”
In Parallel: Practical Enablement Program
Run a practical enablement program focused on individual productivity, informed by what you learned in Week 0. Short live sessions tailored to specific roles (project managers, finance, sales) with exercises people can apply the same day. Internal champions collect and share the best patterns. The point is to create a sense that AI is not an extra thing on top of work; it's part of how the work gets done.
Resource Allocation by Layer
- Layer 1 (Personal Productivity): Invest in an AI assistant (ChatGPT, Copilot, Claude) and role-specific enablement on the tools you already own
- Layer 2 (Workflow Transformation): Resource and enable 1–3 focused workflow builds that will materially change cost, speed, or experience
- Layer 3 (Owners/C-Suite, Data & Policy): Ensure the CEO, CFO, CIO, and business leaders are aligned on priorities and guardrails, with a future that will support AI
The Culture Shift
Most organizations aren't short on tools. We have lots of tools that make big promises. Organizations are short on effective processes to initiate behavior change.
If you don't explicitly plan for:
- Training and coaching
- Time to learn and experiment
- Recognition for teams that lean in
…then AI becomes “one more thing” instead of “the way we do things.” The fastest way to stall an AI program is to treat it as a software rollout instead of a culture shift. You don’t need the smartest person in AI to roll out a tool. You need people who understand how to get change done.
You also don’t need a million-dollar ML project to become an AI-mature company. You do need to execute an AI Playbook: a way to turn money, tools, and ideas into better, faster, more resilient operations. That's the shift—from tools to transformation.
Josiah Shelley is the CEO of ForwardPath, an AI consulting firm serving the construction, architecture, and engineering industries.