When businesses talk about "AI readiness," the conversation usually collapses into a binary: either you're using AI or you're not. Either you're ahead of the curve or behind it. Either your competitors are eating your lunch or you're eating theirs.
That framing is not just imprecise — it is actively unhelpful. It creates the impression that AI adoption is a switch you flip rather than a capability you build, and it gives business owners no guidance about what to actually do next.
Harvard Business Publishing Corporate Learning's 2025 research paper, Succeeding in the Digital Age: Why AI-First Leadership Is Essential, introduces a more useful model: an AI maturity pyramid with four distinct stages. Each stage builds on the previous one. Each has specific characteristics, specific outcomes, and specific requirements for moving to the next level. Understanding where your business sits on this pyramid is the most honest starting point for any AI strategy conversation.
Why a Maturity Model Instead of a Checklist
The reason a maturity model is more useful than a checklist is the same reason a growth framework is more useful than a revenue target: it tells you what the path looks like, not just what the destination is.
An organization at Stage 1 trying to skip to Stage 4 does not become AI-first faster. It becomes an organization with AI tools it does not know how to use, workflows it has not redesigned, and leaders who are being asked to drive transformation they were never equipped for. The gap between where most businesses are and where most AI vendors promise they should be is largely a maturity gap — and the Harvard model gives you a language and a map for it.
The model is structured as a pyramid because higher stages require everything below them to be in place. You cannot develop a genuine AI-first mindset without foundational knowledge. You cannot build AI-related skills without the right mindset. And you cannot reach AI-proficient leadership without the skills that make strategic AI embedding possible. The sequence is not arbitrary. It reflects how organizational capability actually develops.
Stage 1: Foundational Knowledge
The base of the pyramid is foundational knowledge: a working understanding of AI concepts, available tools, data analytics, and topics like cybersecurity and digital governance. In this stage, the primary organizational task is getting AI tools into the hands of employees and establishing clear parameters for acceptable use.
This sounds straightforward. It is not. Many organizations believe they have completed Stage 1 because they have subscribed to an AI tool and announced it to the team. But access is not understanding. Subscribing to ChatGPT Enterprise does not mean your team understands how large language models work, what they are likely to hallucinate about, or how to prompt them effectively for business tasks. Announcing an AI policy does not mean your leaders understand the data governance implications of feeding customer information into a third-party model.
Stage 1 is complete when:
- Leaders at every level can describe what AI does and does not do — not in technical terms, but in terms relevant to their operational area.
- Your team knows what tools are available, which ones are approved for which use cases, and what the boundaries are around data and customer information.
- There is a baseline fluency with at least one AI tool that people are actively using, not just aware of.
Harvard Business School professor Tsedal Neeley sets a useful benchmark for Stage 1 completion: "Everyone in your organization should be working toward at least 30% fluency in a handful of topics, such as systems architecture, AI, machine learning, algorithms, AI agents as teammates, cybersecurity, and data-driven experimentation." (references below) That is not a technical bar. It is a literacy bar — enough to engage meaningfully, ask useful questions, and recognize both opportunity and risk.
Self-assessment question: Can every leader in your organization explain — in plain language — what your current AI tools actually do, what they should not be used for, and where the outputs require human verification? If not, Stage 1 is not complete.
Stage 2: AI-First Mindset
Moving from Stage 1 to Stage 2 is not primarily a knowledge problem. It is a mindset problem. Stage 2 requires leaders and teams to shift from passive access — "we have the tool available" — to active experimentation: "we are actively trying it, learning from it, and building familiarity with what it can do in our specific context."
The Harvard paper describes this stage as developing an AI-first mindset that "embraces curiosity and experimentation." It also involves what the paper calls "letting go of limiting beliefs that may cause leaders to fear or avoid new technologies" and replacing those beliefs with "a balanced perspective of the technology's potential."
That is a more significant shift than it sounds. Many experienced professionals have deeply internalized the idea that good work requires deep human expertise — and that a machine producing a similar output is either a threat to that expertise or a shortcut that will produce inferior results. Neither belief is entirely wrong. But both beliefs, held uncritically, prevent Stage 2 from happening.
What Stage 2 looks like in practice:
- Employees are experimenting with AI tools within their own workflows without waiting for permission or a formal rollout.
- Leaders are sharing what they learn — what worked, what did not, where AI surprised them in either direction.
- The organization is getting better at using AI tools month over month, not just maintaining access to them.
- Individual productivity improvements are measurable — not transformational yet, but real.
The primary output of Stage 2 is familiarity at scale: enough distributed experience with AI tools that the organization can start having informed conversations about where to invest more seriously.
Self-assessment question: Are people in your organization actively experimenting with AI tools on their own initiative — not just using them when asked? Is there a culture where sharing an AI experiment (successful or not) is normal? If experimentation is still top-down and approval-dependent, Stage 2 is incomplete.
Stage 3: AI-Related Skills
Stage 3 is where AI moves from individual productivity into organizational capability. The key transition is from experimentation within existing processes to piloting and scaling AI-enabled processes. This is where most SMBs stall — and where the investment requirement significantly increases.
In Stage 3, leaders need to be able to do things that Stage 1 and Stage 2 do not require: design an AI-enabled workflow from scratch, evaluate whether an AI system is performing correctly, identify failure modes before they create operational problems, and guide teams through the disruption that comes when a familiar process is replaced by an AI-driven one.
The Harvard paper describes Stage 3 as the point where "collaboration becomes increasingly important." Leaders who combine AI skills with traditional capabilities — customer-centricity, design thinking, the ability to build internal consensus and sell ideas across functions — become the change agents that carry transformation forward. Without that combination, AI pilots stay in the pilot stage indefinitely.
What Stage 3 looks like in practice:
- At least one core business process has been redesigned around AI — not just augmented with an AI tool, but fundamentally restructured so that AI is doing the heavy lifting and humans are handling exceptions and judgment calls.
- Leaders can articulate why an AI system is producing the outputs it produces — and can adjust the system when outputs are wrong.
- The organization is measuring AI impact on operational metrics, not just tracking adoption rates.
- AI-enabled processes are being scaled beyond the initial team or function that piloted them.
The outputs of Stage 3 are tangible: efficiency gains in core operations, better decision-making through AI-assisted analysis, and meaningful improvements in the operating model. This is the stage where AI starts to show up in the financial results — not as a cost line for tools, but as a positive delta in productivity and output per person.
Self-assessment question: Has your organization redesigned at least one core process around AI — not just layered an AI tool onto an existing workflow? Are you measuring the operational impact? If AI is still primarily a writing assistant or a search improvement, Stage 3 has not started.
Stage 4: AI-Proficient Leadership
Stage 4 is what the Harvard paper calls "AI-proficient leadership" — the full embedding of AI into the organization's strategic and operational framework. At this stage, leaders use AI insights to refine strategy, streamline processes, create new business models, and position the organization as a forward-thinking entity capable of rapid adaptation to external change.
The critical distinction between Stage 3 and Stage 4 is not the depth of AI integration — it is the organizational agility that comes from that integration. A Stage 3 organization has built AI-enabled processes. A Stage 4 organization can rapidly alter those processes in response to new information, new competition, or new opportunities — because AI is embedded at the infrastructure level, not bolted onto specific workflows.
For most small and medium businesses, Stage 4 looks like this:
- AI is part of how you make strategic decisions — market analysis, resource allocation, growth prioritization — not just how you execute operational tasks.
- New products or services are designed with AI capabilities as a core component, not added later.
- The business can respond to a change in customer behavior, a competitive move, or a market shift faster than it could without AI infrastructure — because the data is already being processed and the relevant signals are already visible.
- AI capability is a source of competitive differentiation that is difficult for competitors to replicate quickly.
Most businesses reading this are between Stage 1 and Stage 2. Stage 4 is not a distant abstraction — it is a direction. And understanding the stages in between is what makes it achievable rather than aspirational.
Where Most Small Businesses Actually Are
Based on the Harvard framework and our own experience working with SMBs, most businesses are in one of two positions: early Stage 1 (access without literacy) or transitioning from Stage 1 to Stage 2 (growing familiarity, inconsistent adoption).
Very few have completed Stage 2. Even fewer have reached Stage 3. This is not because AI is hard or because smaller businesses are behind — it is because the maturity model is genuinely sequential, and most organizations underestimate how much deliberate investment is required to move from one stage to the next.
The good news: moving from Stage 1 to Stage 2 does not require a large budget or a technology team. It requires leadership will and organizational permission — the decision to treat AI experimentation as a legitimate use of time rather than a distraction from "real work." That is a decision the owner or founder makes. No tool purchase required.
Moving from Stage 2 to Stage 3 does require more: specific skills, process design capability, and the willingness to redesign workflows that may currently feel functional. This is usually where external expertise adds the most value — not to hand off the transformation, but to accelerate the learning curve on what AI-enabled process design actually looks like in practice.
The Question That Actually Matters
Not "are we using AI?" but: "which stage are we at, what specifically is preventing us from moving to the next one, and who in the organization is responsible for closing that gap?"
The Harvard research is clear that the answer to the third question is not just the CEO or the founder. It is the leaders at every level who have the knowledge, the mindset, and the skills to carry the transformation into daily operations. Building those leaders — and giving them the mandate to act — is the work that moves organizations up the maturity pyramid.
As the paper concludes: "The time to act is now — delay will only widen the gap between organizations harnessing the transformative power of AI and those already falling behind." That is not a call to buy more software. It is a call to build more capability.