AI can help mid-market organizations move faster and make better use of limited capacity. But when it replaces the customer conversations, investigation, and hands-on problem solving through which people develop judgment, it can also weaken how the organization learns. The goal is to automate and augment thoughtfully, keeping people close to the experiences that generate new knowledge and competitive insight.
For mid-market organizations, the opportunity presented by AI is practical. It can reduce repetitive work, speed up analysis, improve access to information and help stretched teams make better use of limited capacity. But as AI moves into research, planning, customer insight and decision-making, leaders need to look beyond efficiency and ask what people learn by doing the work itself.
AI can summarize, predict and generate at remarkable speed. What it cannot do is reassure a hesitant customer, notice an unexpected pattern on the shop floor or learn from a failed experiment. Before knowledge can be analyzed or reused, someone has to discover it.
A recent story about AI development illustrates this quite well; Amazon was reportedly purchasing physical books, removing their bindings and scanning their pages. Setting aside the copyright and training-data questions for a moment, it is quite revealing that technology companies are returning to books when the internet already contains more information than any person could consume in a lifetime.
AI can produce enormous quantities of new material. What it still needs is source material created by people who studied something, tested an idea, or developed a point of view.
What when organizations use AI outputs as inputs for the next round of work? A market scan becomes the starting point for a strategic plan. That plan shapes departmental priorities, which becomes the starting point for the next year’s planning cycle. If too little new evidence enters the process, the organization can end up refining the same assumptions instead of learning from the market.
Researchers have spent time examining what happens when AI models are trained increasingly on synthetic information rather than original source material. Terms like "model collapse" and "data degradation" can flatten a more nuanced discussion, but the underlying concern is practical: when information is repeatedly generated from prior outputs, some of the detail, context, and variation in the original material can begin to fall away. The photocopy becomes blurrier each time it is copied.
Something similar can happen inside an organization. Customer insight is reduced to summaries. Research starts with AI-generated answers rather than primary sources. Teams begin strategic discussions with a model’s recommendation before developing their own point of view. Over time, the organization becomes more efficient at processing information while weakening some of the activities through which it learns.
For a mid-market organization, that should grab your attention. Your advantage often comes from being close enough to customers and operations to notice what larger competitors miss, while having enough scale to act on those observations. AI should help you make better use of that knowledge without reducing the experiences that create it.
One way to evaluate an AI use case is to consider whether the work primarily processes existing knowledge or whether doing the work also creates new knowledge.
Plenty of processing work is well suited to AI. Turning meeting notes into action items, consolidating information across documents, or producing the first draft of a routine internal communication are all cases where much of the value lies in reaching the finished output efficiently.
Other activities are less straightforward. Consider a customer discovery interview. The visible output might be a transcript, a summary, or a set of themes, all of which AI can process quickly. Yet the person conducting the interview is doing more than collecting answers. They hear when a customer hesitates. They can pursue an unexpected comment and change the next question based on what they have just learned. Across several conversations, they may begin to notice a problem the organization was not originally researching at all.
Replacing that process with synthetic customer personas, or relying exclusively on AI summaries, would save time. It could also remove some of the learning that happens around the formal output.
Strategic planning presents a similar challenge. AI can help a leadership team research a market, explore scenarios, and pressure-test assumptions. If the team hands over the strategic question before it has developed a point of view, though, the starting point becomes a synthesis of information the model already has rather than the leadership team’s experience of its own business.
Employees are continually exposed to information that may not exist anywhere else. A salesperson hears a new objection from several prospects. An operations manager notices an unusual workaround developing on the floor. Someone testing a service realizes customers are using it differently than expected. An experienced employee recognizes that a familiar-looking problem has an important wrinkle this time.
AI can be particularly effective when it follows or accompanies firsthand learning. A team can bring customer interviews into a model to look for patterns they missed. Leadership can use AI to challenge a strategy developed from its own understanding of the business. The technology contributes considerable analytical capacity while the organization continues generating the context that makes the analysis useful.
“Can AI do this?” is rarely enough to evaluate a use case. Leaders also need to understand what their people learn by doing the work and whether that learning creates value elsewhere in the organization.
When evaluating an AI use case, organizations understandably look at time savings, implementation effort, cost, and risk. For some processes, those factors may make the decision relatively straightforward.
For work involving customers, judgment, or problem solving, there is another question worth asking: What do people currently learn by doing this work?
Sometimes the answer will be very little. If someone spends hours each week consolidating information from standard reports, automating much of that work may give them more time for analysis and decision-making.
In other cases, the question may uncover something that is not visible on a process map. A routine customer interaction may also be an informal source of market intelligence. Junior employees may be developing judgment by working through problems alongside experienced colleagues. None of that means the process should remain untouched. It changes how you might redesign it.
Perhaps AI handles repetitive preparation while a person retains the part of the process where judgment is developed. A team might automate the analysis of customer feedback while continuing to have people conduct enough conversations to stay close to what customers are experiencing. If an AI tool takes over work that previously exposed employees to useful information, the organization may need a new feedback mechanism so those observations still reach the people who can act on them.
This is where an AI impact assessment becomes more useful than a simple inventory of tasks that could be automated. It helps leaders look at what the organization gains from a use case and what could change around it.
Good prompting can improve the quality of an AI response, but technique only goes so far. The model also needs useful context.
There is a meaningful difference between asking an AI tool to develop a growth strategy and giving it customer research, operating data, strategic priorities, and the leadership team’s current assumptions, then asking it to identify contradictions or areas that warrant further investigation. The second approach gives the model something distinctive to work with: information generated through the organization’s own experience.
Effective AI capability building should help employees recognize when an AI-generated answer is sufficient, when internal evidence needs to be added, and when the output should lead to another customer conversation, experiment, or investigation.
Without that discipline, prompt engineering can become an exercise in producing increasingly polished answers from the same underlying information.
Efficiency is valuable, especially when people are stretched. There is little reason to preserve repetitive work simply because humans have historically done it. The more difficult decisions involve work where effort and learning are intertwined.
Before changing those processes, look at where new information enters the organization, how employees develop judgment, and which interactions provide insight into customers or operations. Then decide how AI can support the work without closing those learning loops.
Used well, AI can give people more capacity to investigate an unexpected result, spend time with customers, or work through problems that require experience. It can help teams connect information that would otherwise take hours to review and challenge ideas before the organization commits resources to them.
As you assess the next internal process for AI, add one question to the usual discussion about efficiency and ROI: Where does doing this work help our organization learn?
The answer can help determine what to automate, what to augment, and where people still need to stay directly involved.
AI adoption can move quickly from individual experimentation to larger questions about workflows, decisions, employee capability and governance. Those choices are easier to make when each use case is considered in the context of the business rather than evaluated solely on what the technology can do.
Stratford works with leadership teams to assess AI opportunities, understand their organizational impact and build the capabilities needed to use AI effectively. This can include AI impact assessments, practical prompt engineering and capability building, governance, and connecting AI investments to broader business priorities.
If your leadership team is considering where AI can create capacity without weakening the knowledge, judgment and experience your organization relies on, let's talk.
| Natalie Giroux is the founder and former President of Stratford Intellectual Property. Retiring from that role in 2025, Natalie continues to support Stratford as an executive advisor. With deep expertise in strategic IP management and a business-first approach, Natalie has supported over 100 companies in aligning their IP portfolios with growth objectives. She has been internationally recognized multiple times as a leading IP strategist, including being named to the IAM Strategy 300 list. She is passionate about maximizing the value of innovation. |