Q&A: Can AI preserve institutional knowledge before employees leave?

How can businesses use AI to preserve and scale the institutional knowledge that often lives with their most experienced employees? This issue of knowledge management is a pressing one for many firms as data becomes more complex.

In nearly every long-standing business, there are a set of employees who know how everything really works. They know which customer always pays on the twelfth day, which account needs a reminder on day five and which seemingly minor issue is actually a warning sign.

AI offers some answers. Partnering with the humans, AI can both preserve and scale institutional knowledge and surface insights that might be beyond the observation of even the best employees. By identifying patterns across thousands of transactions, communications and customer interactions, AI can augment the team, create scale and de-risk operations.

Lilac Schoenbeck, SVP, Product at Quadient recently spoke with Digital Journal on this subject of AI utilisation to capture knowledge. This includes understanding how AI can help organizations capture the patterns, judgment calls and hard-earned expertise employees build over years, while making those insights more accessible across the business.

Digital Journal: Why is so much critical business knowledge still concentrated among individual employees?

Lilac Schoenbeck: Many business processes have long relied on a combination of technology and human expertise. Over time, employees develop a deep understanding of how those processes work, recognizing patterns and making judgment calls that aren’t easily captured in a system.

That understanding develops over years of working with company data and seeing the same situations play out again and again. Employees learn to recognize exceptions, make decisions, and connect signals that may not be obvious to anyone else. Over time, they become the person everyone turns to because they know not only what the process says should happen, but what actually happens in practice.

The problem is that their knowledge has a scope, and it often lives almost entirely in their head. And these folks tend to be localized – not everywhere in the org. Eventually, that person retires, changes roles, or leaves the company, and much of that hard-earned wisdom goes with them. And the mysteries of how things ‘just happen’ become more and more opaque.

DJ: How can AI make that experience more broadly accessible?

Schoenbeck: AI doesn’t get overwhelmed by volume in the way people naturally do. A finance team may be looking at thousands of invoices, customer interactions and payment histories, making it difficult to spot the handful of details that actually deserve attention. AI can process all of that information at once and identify the patterns, exceptions and opportunities that would be nearly impossible to find manually.

That doesn’t replace human judgment. It gives people a much better starting point by surfacing where their attention is likely to have the greatest impact. This wont always be in the obvious places. For collections, for example, time is best spent on accounts that will pay with a timely reminder – not those who consistently pay 2 days late, or those who may never pay at all. Stack ranking effort by due date would be a colossal waste of resources. Creating an algorithm deterministically would likely yield the wrong outcome. Setting AI on pattern recognition, expanding the reach of the intuition of the team, could pay tremendous dividends.

DJ: Does that mean replacing the experienced employees who currently perform this work?

Schoenbeck: No – absolutely not.  Technology has long been used to accelerate or facilitate manual tasks, and most employees are grateful for the reduction in repetitive, tedious work. But, in this case, human collaboration is what creates the benefit.

That magical human with the insider knowledge – he or she probably can step back and define their operating rules. For example, they might say that “Accounting firms always pay on the final due date.” Or they might say “We wait for a certain cash flow trigger before we pay our invoices.” But it may turn out that this rule, applied manually, covers some small percent of the actual flow of transactions – because one person cannot reasonably apply a set of thoughtful but disparate rules to every single customer or vendor.

AI can. That individual can express their criteria and ask the technology to partner with them to expose the right next actions, based on those criteria. It can go a step further, sorting by size of the invoice or criticality of the account. It can crunch the entire data set – and propose further refinements, tests to be run, or observations that might evolve the rule-set to be more scalable, more accurate, or more repeatable by other employees.

This partnership between AI and the teams already in place can extend the impact that knowledgeable individuals can have within an organization – and, at the same time, capture some of that tribal knowledge in more durable ways.

DJ: What should organisations capture before their most experienced employees leave?

Schoenbeck: This may be the wrong question to ask – as it starts from a defensive stance. What if we asked ourselves: How can we expand the career of our most valuable employees by enabling them to have broader impact and engage with the newest technologies to do so?

The framing is intentionally more positive. The experience of engaging with AI to augment our own thinking is typically described as fun and interesting. Rather than thinking of this as a brain-dump, what if organizations approach employees with the desire to learn from them, strengthening their import.

Then, the collaboration with AI becomes career growth as much as risk mitigation. It’s refinement of ingrained processes – or even evolution of them – not rigidly documenting them. It’s an opportunity for everyone to excel and ultimately deliver better outcomes, revenue and customer experience.

DJ: What role does human judgment continue to play once AI is embedded in these workflows?

Schoenbeck: We have all spent enough time with AI output to treasure the role of human judgement. While AI can support decision-making, can create endless documents and spreadsheets, and can surface anomalies, regular users will routinely smack back its suggestions as incomplete or misguided.

Why? Because it lacks the full context. A person can then bring context that may not be visible in the data. For example, the customer recently spoke with a sales representative, but has an open service issue or is going through a contract renewal. Sending an aggressive collections message at that moment may be exactly the wrong choice.

So, we delegate only the most predictable choices to AI right now – under fairly understood rules. Most companies, however, are benefiting from the insights and recommendations that facilitate faster, more effective human decisions. But we’re a long way from “why didn’t you just do what AI said to do?”

DJ: What prevents companies from applying this kind of pattern recognition effectively?

Schoenbeck: One major barrier is that information is often fragmented across departments, people, and systems. Finance may not know that sales spoke with a customer yesterday. Marketing may not know that the customer received an incorrect invoice. Customer service may be dealing with an issue that none of the other teams can see.

The gap is that all these functions communicate with the same customer without necessarily knowing when the others are communicating or why. No amount of individual expertise can fully solve that problem when thousands of messages and transactions are involved.

AI can help bring those signals together, but organizations also need to improve the underlying connections between their systems and teams. They may not reach a perfect, unified view of the customer next year, but they can surface additional pieces of information that lead to better decisions. Moving the ball forward incrementally still creates real value.

DJ: How can companies modernize and implement AI responsibly and strategically?

Schoenbeck: Businesses should begin with the outcome they are trying to improve. Are they trying to reduce invoice errors, collect cash faster, identify accounts at risk or help employees prioritize their time? Adding AI without a clear business purpose is not innovation. It is adding technology because the market expects to see an AI label.

The right approach will also depend on the organization’s maturity. A large enterprise may have its own AI team, preferred models, and extensive governance and visibility requirements. A smaller business may not have any of those resources and will instead need a trusted technology provider to embed the capability responsibly and stand behind it.

Wise leadership knows the right answer for each organization. Most businesses do not need an AI laboratory. They need complete, accessible solutions that make their work easier, provide transparency around how the technology is being used, and protect their data.

DJ: Looking ahead, how could AI change the value of institutional knowledge within businesses?

Schoenbeck: Institutional knowledge will remain enormously valuable, but it will become less dependent on one person being the irreplaceable wizard behind the curtain. AI gives organizations a way to recognize patterns across years of activity and make those insights available to far more people.

The goal is to preserve what their most experienced employees have learned, combine it with a much larger view of the organization’s data and help every team member make decisions with more context. That creates an operational reality we have never really had before: insightful decisions made by capable employees, supported by the technology they need every step of the way.