Amplification

AI: The Most Obvious Amplifier

Organisations spend enormous amounts of time looking for new capability. New technology, new people, new processes, new operating models and new programmes are introduced because something needs to become faster, stronger, cheaper or more scalable. Sometimes that is exactly what is required. At other times, however, the organisation already contains much of the capability it is trying to create. The problem is that the capability exists in the wrong place, is concentrated in too few people, remains difficult to access or is prevented from having a wider effect by the systems around it. Amplification is the deliberate act of increasing the impact of something that already exists. The principle sounds simple, but it has an uncomfortable characteristic: amplification is neutral. An amplifier does not decide whether the signal is good. It makes the signal stronger. Good judgement can be amplified, but so can poor judgement. Strong processes can be accelerated, but so can badly designed ones. Knowledge can be made available to hundreds of people, but incorrect knowledge can spread just as easily. A talented employee can become a force multiplier for a team, or the organisation can increase its dependence on that person until their capability becomes a structural weakness.

Artificial intelligence makes this principle unusually visible because AI can increase speed and scale dramatically. Yet AI did not invent amplification. Managers have always amplified people through mentoring, delegation, recognition and opportunity. Organisations amplify behaviours through incentives and measurement. Processes amplify decisions by making them repeatable. Knowledge systems amplify experience by allowing one person's learning to benefit others. Culture amplifies what people believe is safe, valued and expected. The leadership question is therefore not simply how to create more capability. It is how to recognise what already deserves a greater effect, how to create the conditions for that effect to spread and how to avoid multiplying the weaknesses surrounding it. Artificial intelligence is often discussed as though organisations are adding intelligence to an otherwise unchanged environment. A company has a process, introduces AI and expects the process to become faster, cheaper or more effective. The attraction is understandable. Generative AI can summarise information, produce drafts, analyse large bodies of text, automate repetitive work and make knowledge accessible in ways that would have seemed unrealistic only a few years ago. When a task takes hours and technology appears capable of reducing it to minutes, the business case can feel almost self-evident.

The more important question is what the organisation is asking the technology to amplify. If a process is well designed, ownership is clear, the underlying information is trustworthy and people understand what a good outcome looks like, AI can create enormous leverage. Work that once required repetitive manual effort can be accelerated without removing the judgement surrounding it. People can spend less time compiling information and more time understanding what it means. Knowledge can be retrieved more quickly. Routine preparation can happen automatically while the people responsible for the outcome retain control over the decisions that matter. The same technology behaves very differently when those conditions are absent. A badly designed process does not become well designed because AI is inserted into it. It becomes a badly designed process that can operate more quickly. Conflicting information does not become true because a model can summarise it elegantly. Unclear ownership does not disappear because a workflow has become automated. If nobody agrees which source is authoritative, who is responsible for maintaining it or what should happen when the automated result is wrong, the organisation has not removed complexity. It has hidden some of that complexity behind a faster interface.

This is why the quickest AI success can sometimes be misleading. Consider a team that spends several hours every week producing a management report. Information is collected from different systems, copied into a document, summarised and sent to leadership. The obvious AI opportunity is to automate the report. If the objective is simply to reduce the time required to produce the existing output, the implementation may be straightforward. The organisation can demonstrate that a task previously requiring hours now takes minutes and reasonably describe the difference as an efficiency gain. That gain may be completely real. Yet the implementation has answered only one question: can the existing process be made faster? It has not necessarily asked why the report exists, whether the information remains useful, whether the source data is consistent, whether recipients trust the output, which parts genuinely require human interpretation or whether the report is being used to make decisions at all. Automation can therefore preserve a process precisely because it has become cheap enough that nobody feels the need to challenge it.

The irony is that the technology may make it easier to avoid solving the more important problem. This pattern appears whenever organisations automate work before understanding it. A manual process contains friction, and friction is naturally interpreted as inefficiency. Sometimes it is. Someone copies the same information between systems every week because integration was never built. A manager spends hours reformatting data that already exists somewhere else. Employees repeatedly answer the same basic questions because organisational knowledge is fragmented. Removing those activities can create genuine value. At other times, friction contains information. A person manually reviewing a report may notice that two systems disagree. An experienced employee answering a colleague's question may realise that the documented process no longer matches what actually happens. A manager compiling customer information may recognise that the formally green account contains several signs of deterioration. If the organisation automates the visible task without understanding the judgement embedded inside it, the process can become faster while the organisation becomes less aware of the conditions underneath it.

The question is therefore not whether AI should automate work. In many cases it absolutely should. The question is whether the organisation understands the work well enough to know which parts are mechanical, which parts contain judgement and what assumptions are being transferred into the automated process. Those assumptions matter because AI increases reach. A poor judgement made by one person may affect one customer, one report or one decision. The same judgement embedded into a process can affect hundreds. An incorrect article in a knowledge base may sit unnoticed for months when employees rarely consult it. Connect that knowledge base to an AI assistant used across the organisation and the same error can suddenly appear in dozens of conversations with confidence and consistency. The technology has not created the mistake. It has amplified its effect.

Data quality therefore changes from an administrative concern into an operational one. Organisations have tolerated inconsistent documentation for years because humans are surprisingly good at compensating. Experienced employees know which document is outdated. They recognise terminology that changed two years ago. They know that one procedure is technically correct but not how the work is actually performed. They can distinguish between a formal policy and the practical exception everyone uses. This informal knowledge allows weak information systems to continue functioning because people provide the interpretation the system itself lacks. AI does not automatically possess that history. If an organisation wants technology to make knowledge easier to access, it also needs to decide what knowledge should be trusted. Duplicate documents, contradictory guidance, unclear ownership and material that nobody has reviewed in years become much more important once machines begin treating them as reusable organisational knowledge. The work of cleaning, governing and maintaining information may appear considerably less exciting than demonstrating a new assistant, but the assistant can only amplify what is available to it.

This creates a recurring tension between visible innovation and invisible foundation. A successful demonstration can be created quickly. A strong organisational capability usually cannot. The demonstration shows what is technologically possible. The capability requires decisions about ownership, data, process, governance, risk, adoption, measurement and the responsibilities that remain human. It requires people to agree not only what the technology can do, but what it should do and what happens when the result is uncertain.

That final question is particularly important because automation can create a dangerous psychological effect. When a machine produces an answer consistently and quickly, people can begin treating consistency as evidence of correctness. Manual processes expose uncertainty because humans hesitate, ask questions and disagree. Automated processes can conceal uncertainty behind a clean result. Leadership therefore becomes more important as execution becomes easier. The more an organisation delegates to technology, the more clearly it needs to understand who owns the consequence. If an AI-generated recommendation is wrong, who is expected to notice? If a workflow takes an action automatically, who has authority to stop it? If the underlying information changes, who ensures the system changes with it? If the technology creates a result nobody anticipated, which function is responsible for deciding whether that result is acceptable?

These are not arguments against AI. They are the conditions that allow AI to become genuinely useful rather than superficially impressive. The strongest AI implementations I expect to endure are unlikely to be the ones that automate the greatest number of tasks as quickly as possible. They will be the ones that understand where technology provides leverage and where human judgement remains essential. They will use AI to remove repetitive work, make good information easier to access, surface patterns people would otherwise struggle to see and create more capacity for work that genuinely benefits from human experience. The organisation should not ask only, What can we automate? It should also ask, What are we about to make faster, more repeatable and more influential? Because AI is not a magic layer applied on top of an organisation.

It is an amplifier connected to the organisation that already exists.

You Cannot Amplify What You Cannot See

Technology makes amplification easier to recognise because the effect can be dramatic. A process that once required several hours can suddenly require minutes. A person who previously searched through dozens of documents can retrieve an answer almost immediately. The before-and-after comparison is visible enough that organisations can measure it. Human capability is often much harder to see. Every organisation contains people whose contribution exceeds what their title suggests. Some are obvious high performers. They lead meetings confidently, present well and accumulate visible achievements. Others create value in quieter ways. They are the people colleagues ask for when something unusual happens. They understand why a customer reacts differently from what the account plan suggests. They recognise technical patterns because they remember the last three times something similar occurred. They can explain complicated subjects without making other people feel foolish. They notice gaps between functions and step across them before those gaps become problems.

Much of that capability is difficult to capture through formal measures. A dashboard can show the number of customers someone manages, the tickets they close, the revenue associated with their portfolio or the projects they complete. It is much harder to quantify the number of incidents that did not escalate because they recognised the problem early. A metric rarely shows that a new employee became productive faster because an experienced colleague spent time helping them. It does not easily capture the customer who remained calm because somebody had built enough trust over several years that an incomplete update was still credible. The organisation therefore risks amplifying the people who are easiest to measure rather than the capabilities that create the greatest value. This is why Amplification depends so heavily on Visibility.

Before leaders can increase the effect of capability, they have to understand where capability really sits. That requires looking beyond formal structures and asking why certain people repeatedly appear around successful outcomes. Why does this person keep being asked to join difficult customer conversations? Why do new employees seek them out? Why does this team seem to recover faster when a particular individual is involved? Why does one manager consistently produce people who are ready for greater responsibility? Why does a process work smoothly in one region but struggle somewhere else? Why does one customer relationship remain stable despite experiencing similar operational problems to another? These questions are not attempts to identify heroes. They are attempts to identify mechanisms. A person may be successful because they possess unusual technical expertise. Another may be successful because they create trust. Someone else may have learned how to navigate the organisation and knows which people need to be involved before an issue becomes urgent. A manager may produce strong teams because they delegate differently, share information more openly or create psychological safety around difficult conversations.

The outcome is visible. The leadership work is understanding what produced it. This distinction matters because organisations often respond to good performance by increasing demand rather than increasing capability. A strong employee handles a difficult account successfully, so the next difficult account is also assigned to them. They resolve a complicated escalation, so they are asked to help with the next one. They mentor a colleague effectively, so every new person is directed towards them. They know the technology, so every technical question eventually reaches them. Each decision makes sense. The organisation is using the person best able to produce the required outcome. From the perspective of the immediate task, this is rational allocation of capability. From the perspective of Amplification, however, something is missing.

The organisation is consuming the capability without learning enough from it. A leader applying Amplification should eventually ask a different set of questions. What does this person know that others do not? Which behaviours make them effective? Which parts can be taught? Which conditions allow them to perform well? What responsibilities could be distributed? Which knowledge should be documented? Who else has the potential to develop similar capability? What should this person be doing next if the organisation did not require them to keep solving the same category of problem? This turns recognition into organisational learning. It also changes how leaders think about talent. Talent is often treated as an individual characteristic. Someone is described as high potential, technically strong, commercially minded or an excellent communicator. Those descriptions may be accurate, but they say relatively little about what the organisation should do next. The more useful question is how that individual's strengths can create value beyond the work they personally perform.

A strong technical specialist can solve more difficult problems. An amplified technical specialist can help other people become better problem-solvers. A trusted customer leader can maintain important relationships. An amplified customer leader can help the organisation understand what creates trust and build those behaviours into the wider customer model. A good manager can lead one successful team. An amplified manager can develop other managers who create successful teams of their own. The difference is reach. This does not mean every expert should become a trainer, every high performer should become a manager or every successful method should be standardised. Some capabilities depend heavily on personality, experience or judgement that cannot be reproduced through a checklist. People should also be allowed to remain exceptional individual contributors without being required to turn themselves into organisational development programmes.

The purpose is not to clone successful people. It is to understand whether some part of their success can improve the system around them. Sometimes the answer will be no. A person may possess specialist expertise that only a small number of situations require. The most efficient model may genuinely be to keep that expertise concentrated and ensure people know how to access it when needed. At other times, the organisation will discover that what looked like an individual strength is actually a missing organisational capability. If one employee repeatedly translates between technical and commercial teams, perhaps the organisation has a translation problem. If customers trust one individual far more than they trust the wider service model, perhaps the organisation has a relationship design problem.

If new employees can only learn the role by sitting beside a particular colleague, perhaps the organisation has a knowledge problem. If one manager consistently identifies strong internal candidates that the formal recruitment process overlooks, perhaps the organisation has learned something about what actually predicts success. Visibility makes these patterns available. Amplification decides what to do with them. There is also an important leadership humility involved. Managers are often under pressure to demonstrate that they are creating something new. New programmes, new frameworks and new initiatives are visible signs of action. Recognising that an existing employee has already discovered a better way of working can feel less dramatic. Yet some of the strongest improvements begin exactly there. A team has developed an informal process because the formal one does not work. Rather than immediately replacing it with a leadership-designed process, understand why the informal method succeeds. An employee has built a particularly strong customer relationship. Rather than simply celebrating the relationship, understand what behaviours created it. A manager has unusually strong retention or development within their team. Rather than attributing the result to personality, examine what that manager does differently.

The organisation may already contain prototypes of its own future. The challenge is seeing them before looking elsewhere. This is where visibility becomes more than observation. Seeing hidden capability creates a responsibility. Once leaders understand that value exists somewhere in the organisation, continuing to treat it as an accident becomes a choice. The question is no longer simply whether somebody is good. It is whether the organisation is becoming better because that person is there.

From Individual Capability to Organisational Capability

One of the best managers I worked with understood this distinction particularly well, although I do not remember us ever using the word Amplification. We worked in a large technology organisation and often recruited new people into our team from elsewhere inside the company. Support was a particularly useful source of talent because we already knew many of the people working there. We had seen their technical ability, but we had also seen something a CV or formal interview process could reveal less easily: how they behaved when things became difficult. We knew who remained calm under pressure. We knew who could explain complicated problems clearly. We knew who customers trusted. We knew who was curious enough to investigate beyond the obvious answer and who naturally helped colleagues without needing to be asked.

When we needed another person, my manager would sometimes ask me who I thought might be a good fit. The question itself was revealing. He was not only asking whether somebody met the formal requirements of the role. He had recognised from experience that the strongest people in our team tended to possess a particular combination of capabilities. Technical credibility mattered because customers and internal specialists needed confidence that the person understood the environment. Communication mattered because technical knowledge has limited value in a customer-facing role if it cannot be translated clearly. Curiosity mattered because complex problems rarely arrive with complete instructions. Judgement mattered because enterprise work constantly requires decisions in situations where the process alone does not contain the answer. My manager had noticed that combination.

I happened to know people elsewhere in the organisation who were already demonstrating it. So instead of simply continuing to rely on the few people in the team who possessed those strengths, he used what he had learned to identify additional people who might develop in the same direction. Sometimes I would encourage the people I recommended to apply. When they joined, I helped mentor them while they learned the role. At the time, I thought of this mostly as recruitment and coaching. Looking back, it was organisational amplification. My manager had recognised capability, understood some of the conditions producing good outcomes and used the authority available to him to increase the amount of that capability inside the team. He did not redesign the company's global recruitment strategy. He did not create a new competence framework covering thousands of employees. He did not launch a large training programme or attempt to turn everyone into the same type of person.

He simply paid attention.

He did not simply recognise that one person possessed a useful capability. He learned from it and deliberately created more of it.

That matters because leadership advice often makes organisational improvement sound dependent on authority that many managers do not possess. A manager may recognise that a role would benefit from different recruitment criteria but have no power to change the global job description. They may understand that the wider organisation needs better technical training but have no budget for a formal programme. They may see that a process should be redesigned but lack ownership over the function responsible for it. Those constraints are real. Amplification asks what can still be done within them. A manager may not be able to change the global hiring profile, but they may influence who joins their own team. They may not be able to create a corporate academy, but they can give experienced employees time to coach others. They may not be able to change every role structure, but they can make knowledge transfer part of how work is allocated. They may not control compensation, but they can make exceptional contribution visible and advocate for progression.

Small decisions can create significant leverage when they are repeated consistently. The mentoring element is particularly important because mentoring changes what a high performer produces. If an expert solves a difficult problem alone, the organisation receives the solution. If the expert solves the same problem while coaching someone else through the reasoning, the organisation may receive the solution and begin creating another person capable of solving similar problems in the future. The second approach is usually slower in the moment. The expert has to explain what they are thinking. The learner asks questions. Mistakes occur. Something the expert could have completed almost automatically becomes more deliberate because another person needs to understand the process. This can look inefficient. Many organisational systems actively discourage it. If people are measured mainly on individual output, mentoring reduces visible productivity. The expert closes fewer tasks because time is spent helping someone else. The learner also works more slowly because they are still developing. In the short term, the team appears to have consumed more capacity to produce the same result.

Yet the outcome is not the same. One approach completes a task. The other can build capability. This distinction becomes increasingly important as organisations grow. A company cannot scale indefinitely by finding one exceptional person and routing more work towards them. At some point, the person's time becomes the constraint. Every additional customer, escalation, new employee or complicated decision has to pass through the same limited capacity. Mentoring breaks that relationship. The organisation begins turning individual knowledge into shared knowledge, individual judgement into wider judgement and individual experience into organisational experience. Documentation can provide another form of amplification. Experienced employees accumulate a remarkable amount of knowledge that never appears in formal systems. They know which questions to ask before starting an investigation. They remember why a process contains an unusual exception. They understand which customer stakeholders need early communication. They have learned which apparent symptoms usually point towards a deeper problem.

When that knowledge exists only in someone's head, the organisation benefits only when that person is available. Capturing some of it changes the economics. A useful knowledge article does not eliminate the need for expertise. It allows expertise to travel further. A checklist does not replace judgement, but it can prevent new employees from repeating avoidable mistakes. A documented troubleshooting approach may allow ten people to begin an investigation with knowledge previously available only to one. This is one of the areas where AI can create particularly strong amplification when the foundation is good. Traditional knowledge systems often fail because finding information is difficult. People may know that useful documentation exists but not know where it is stored, what terminology the author used or which of several similar documents is current. A well-governed AI layer can reduce this retrieval problem dramatically. Instead of expecting employees to navigate the organisational filing system, the technology can help bring relevant knowledge to the point where it is needed.

But the valuable asset remains the knowledge. AI improves the distribution mechanism. The same is true of templates, playbooks, peer reviews and communities of practice. None of these is valuable simply because it exists. They become valuable when they allow good judgement developed somewhere in the organisation to influence work somewhere else. This is why I am cautious about describing amplification purely as standardisation. Some organisational knowledge should absolutely become standard. Security requirements, operational controls and repeatable processes benefit from consistency. Other capabilities are more contextual. Customer judgement, leadership conversations and complex troubleshooting often depend on adapting principles to circumstances. Trying to capture every successful behaviour in a mandatory process can destroy the very judgement that made the original person effective. Good amplification therefore preserves the difference between principle and prescription.

If a strong customer leader succeeds because they listen carefully, understand technical reality and communicate uncertainty honestly, the organisation should amplify those principles. It should not necessarily convert every conversation into an identical script. If an experienced engineer has developed a reliable investigative method, the organisation can capture the questions and reasoning that make the method useful without pretending every problem will follow the same sequence. If a manager develops strong people by giving them progressively more difficult work with support available when needed, the organisation can learn from that development philosophy without forcing every manager to follow an identical schedule. Amplification should increase reach without removing intelligence. The manager I described earlier did not attempt to produce clones. The people joining our team had different personalities, backgrounds and strengths. The value came from recognising a combination of capabilities that mattered and deliberately creating more opportunities for those capabilities to exist.

That made the team stronger in several ways. Knowledge became less concentrated. Customers had more than one person they could trust. New employees had access to experienced mentors. The team had a larger pool of people capable of handling difficult situations. And the people who had previously carried much of that responsibility gained more room to develop themselves rather than being permanently required to solve every problem personally. That final point matters. The objective of amplification should not be extracting more output from capable people. It should be increasing the organisational effect of their capability while allowing them to continue progressing. There is a significant difference between telling someone, You are valuable, so we need more from you, and telling them, You are valuable, so we want more of the organisation to benefit from what you know while creating room for you to grow beyond it.

Both statements begin with recognition. Only one creates capability.

When Amplification Creates Dependency

Amplification becomes dangerous when organisations mistake repeated use for multiplication. A talented person receives more difficult work because they are capable of handling it. Their output increases. Customers receive better service, colleagues receive help and managers experience fewer problems because somebody reliable is absorbing complexity. From a distance, this can look like successful amplification. The organisation has found a strong capability and increased its impact. The problem is that the capability has not actually spread. Demand has simply been concentrated around it.

This is how exceptional performance gradually becomes structural dependency. The pattern rarely begins with a bad management decision. It begins with a problem. A difficult customer situation appears and one employee knows how to resolve it. They help. Another customer experiences something similar and the same person becomes involved because the previous outcome was good. A technical question crosses organisational boundaries and they understand enough of both sides to translate it. A new employee joins and needs help, so the experienced person spends time mentoring them. Someone leaves unexpectedly and the capable employee covers part of the workload because the work still needs to be done. Every decision is individually reasonable. The person may volunteer. Capable employees often enjoy difficult problems. They like being useful. They take satisfaction from being trusted and may prefer solving the problem themselves to watching it remain unresolved. This is one reason dependency can be difficult to recognise while it is forming. The organisation is not forcing an obviously dysfunctional arrangement onto an unwilling employee. The arrangement can feel positive for everyone involved.

Customers receive help. Managers receive reliability. Colleagues receive support. The employee receives recognition, challenge and a sense of importance. The problem appears when the exceptions stop being exceptional. At some point, a manager should become curious about why the same person is required so often. Why does this customer relationship only seem stable when they are involved? Why are they carrying substantially more complexity than peers with the same role? Why does everyone ask them technical questions? Why do new employees struggle without their help? Why does a process repeatedly need them to bridge the same organisational gap? The answer may be that the employee is simply very good. But leadership cannot stop there. If the organisation requires one unusually capable person to keep compensating for the same weaknesses, then the person's strength is partly concealing the system's weakness.

Heroic performance can create poor visibility. A process appears to work because someone keeps repairing it. A customer model appears effective because one individual supplies the relationship quality the model itself does not consistently produce. A team appears adequately staffed because a small number of people carry disproportionate complexity. A technical environment appears manageable because one expert knows how to navigate undocumented dependencies. The stronger the person is, the longer the organisation may be able to avoid confronting the weakness. This creates one of the strangest penalties of competence. A capable employee demonstrates that they can operate beyond the normal expectations of their role. They solve harder problems, customers trust them, colleagues learn from them and leadership increasingly depends on their judgement. Those are exactly the characteristics that should make progression more likely.

Then an opportunity becomes available and the organisation discovers how dependent it has become. Who will take the difficult customers? Who understands the technical history? Who will train the new people? Who will maintain the customer relationship? Who can replace the informal coordination that happens every week without appearing in anybody's job description? The employee has become simultaneously ready to progress and difficult to move. At that point, the organisation may genuinely value them while having accidentally created incentives to keep them exactly where they are. This is not successful amplification. It is capability consumption. The organisation has taken something valuable and increased the number of demands placed upon it without increasing the amount of capability available. The same principle applies to managers. A strong manager can become the person senior leaders call whenever a team is struggling. They inherit difficult employees, unstable functions or unresolved customer situations because leadership trusts them to restore control. This can be excellent development for a period. Eventually, however, the manager's reward for building stable teams can become a permanent supply of unstable ones.

Again, the organisation receives immediate value. But if nobody studies what the manager does differently, the organisation remains dependent on the individual rather than learning from their capability. Technology can create exactly the same form of dependency in less obvious ways. Imagine a manual process built around a series of undocumented exceptions. One experienced employee knows how to operate it because they understand why each exception exists. The organisation decides to automate the process and asks the same employee to explain the workflow. If the objective is simply to reproduce the current process electronically, the automation can lock years of workaround behaviour into code. The organisation has reduced its dependency on the person running the task manually, but may have increased its dependency on a process nobody fully understands.

Automation can therefore make dependency look like resilience. The person is no longer required every day, so the organisation believes the risk has disappeared. Yet the assumptions that person had been compensating for remain embedded in the automated system. When conditions change, nobody remembers why a particular decision rule exists or what exception was originally being handled. This is one reason good process improvement frequently begins by asking why. Why does this approval exist? Why is this information copied manually? Why does this customer require a different route? Why does this report contain this field? Why does this employee have to intervene? Some answers will reveal unnecessary friction that should be removed. Others will reveal important judgement that must be preserved. Amplification requires distinguishing between the two.

There is also a cultural dependency that appears when organisations celebrate heroics too enthusiastically. Every organisation needs people willing to step beyond normal boundaries when circumstances require it. Serious incidents, customer crises and unexpected operational problems sometimes demand unusual effort. Recognising people who respond well is entirely appropriate. The danger is allowing the heroic response to become evidence that the system works. If every major incident requires employees to work through the night, the organisation does not possess a strong incident model simply because the team repeatedly saves the situation. If customer relationships depend on senior people personally intervening whenever something becomes difficult, the existence of effective escalation does not necessarily mean the underlying service is healthy. If a manager consistently works extraordinary hours to keep the team functioning after headcount reductions, the absence of visible failure may say more about the manager than the sustainability of the operating model.

Heroics can hide cost because capable people absorb it personally. Time, stress, delayed development and reduced resilience rarely appear in the same financial report as the saving that created them. This creates another amplification problem. The organisation observes that exceptional effort prevents failure. It rewards the effort. The behaviour becomes culturally desirable. More people learn that recognition comes from rescuing difficult situations. Preventative work receives less attention because preventing the crisis creates no dramatic event. The organisation begins amplifying firefighting. The problem is not that the firefighters are doing anything wrong. They may be exactly the people the organisation needs in that moment. Leadership should appreciate them. Then leadership should ask why the fire keeps starting. Healthy amplification therefore contains an element of redundancy. That may sound inefficient because redundancy is often associated with unused capacity. Yet resilient organisations deliberately avoid concentrating every critical capability in one place.

A customer should ideally trust more than one person. A team should have more than one employee who understands a critical process. A technical environment should not depend entirely on one specialist's memory. A manager should be able to progress without the team becoming incapable of functioning. This does not require every employee to know everything. Specialisation remains valuable precisely because expertise takes time to develop. The objective is not eliminating dependency. It is understanding which dependencies are acceptable and which have become structural risks. One useful test is to imagine that the person is unavailable tomorrow. What becomes difficult? What becomes impossible? Which customers are affected? Which decisions stop? Which knowledge disappears? Which responsibilities suddenly have no obvious owner? The purpose of asking is not to treat employees as replaceable components. Quite the opposite. It makes their real contribution visible.

Once that contribution is visible, leadership has a choice. It can continue consuming the capability until something forces the issue. Or it can begin turning some of that individual strength into organisational strength while there is still time to do it deliberately.

Systems Amplify Behaviour

People are not the only amplifiers inside organisations. Systems amplify behaviour. Every target, incentive, approval process, reporting structure and leadership response teaches people something about what the organisation actually values. Formal policies matter, but employees usually learn faster from consequences than from statements. If leaders say they value transparency but punish people who surface bad news, caution is amplified. If leaders say they value ownership but require approval for every meaningful decision, dependency is amplified. If an organisation celebrates people for resolving crises while giving little recognition to those who prevent them, firefighting is amplified. If managers consistently reward people who develop others, knowledge sharing is amplified. Culture is partly the accumulated result of these signals. This makes amplification a useful way to think about organisational design because systems do more than control behaviour. They increase the likelihood that certain behaviours will repeat.

Measurement provides a simple example. Imagine a support function primarily measured on ticket closures. The metric has legitimate value. Leaders need to understand workload, throughput and whether queues are growing. Problems begin when closure volume becomes a proxy for individual contribution. People adapt. Simple tickets become attractive because they improve visible productivity. Complex investigations become less attractive because they consume substantial time while producing only one closure. Helping a colleague may be valuable to the team but reduces the helper's personal output. Documenting a solution for future use creates no immediate ticket closure at all. Nobody needs to manipulate the system dishonestly. Rational employees simply learn what the organisation sees. The measurement amplifies the work most compatible with the measurement. The same thing happens with managers.

If managers are judged mainly on whether their immediate team hits quarterly targets, investing time in development can become a secondary priority. Coaching someone today may improve next year's capability but reduce this month's productivity. Giving a developing employee a difficult task introduces risk. Solving the problem personally may feel safer. The system encourages the manager to consume existing capability rather than build new capability. This is one reason organisations can sincerely describe people as their greatest asset while operating systems that consistently favour short-term extraction. The words and the incentives point in different directions. Amplification follows the incentives. Approval structures create another powerful signal. Controls are necessary. Financial commitments, security risks, customer obligations and significant operational changes should not depend entirely on unrestricted individual discretion. Good governance protects organisations from decisions whose consequences extend beyond the person making them.

But controls can also teach people not to decide. If every exception requires several layers of approval, employees learn that ownership means preparing information for someone else to choose. Managers become coordinators rather than decision-makers. Senior leaders become involved in increasingly operational choices because the organisation has taught everyone below them that authority sits somewhere higher. The immediate effect may be consistency. The longer-term effect can be reduced judgement. People gain experience making decisions by making decisions. If the organisation continually protects them from that responsibility, it should not be surprised when they remain dependent on senior leadership. Delegation is therefore an amplifier. Give capable people genuine ownership with clear boundaries and their judgement develops. Give them tasks without authority and their dependency develops. The same dynamic applies to knowledge.

An organisation where experts are rewarded primarily for possessing scarce knowledge may unintentionally encourage concentration. Scarcity creates importance. If sharing what you know makes you less indispensable, the rational behaviour may be to remain the person everyone needs. Most employees do not deliberately hoard knowledge in such an explicit way. The pattern is usually subtler. Documentation is never prioritised because immediate delivery receives more recognition. Mentoring happens in spare time. People are praised for knowing the answer rather than for making sure the team can answer without them. The system amplifies expertise while failing to amplify capability. Change the recognition and a different behaviour becomes possible. A manager who recognises the person who created a useful guide, coached a colleague into independence or removed a recurring dependency sends a different signal. The organisation is no longer saying only, We value what you can do.

It is saying, We value what becomes possible because you were here. That is a much more scalable definition of contribution. Recruitment systems also amplify organisational assumptions. When a role becomes difficult to fill, organisations often respond by making the specification increasingly precise. They search for someone who already possesses every technical, commercial, sector and leadership characteristic considered useful. This can feel efficient because the ideal candidate requires less development. Yet every additional requirement narrows the pool. The organisation can end up searching for a person who perfectly reflects today's operating model rather than someone capable of growing with tomorrow's. There is a reasonable alternative. Some capabilities are easier to teach than others. Technical knowledge can often be developed through training, exposure and practice. Product knowledge can be learned. Processes can be taught. Communication, curiosity, judgement and the ability to create trust can also develop, but they often require more time and are considerably harder to assess through certification alone.

A manager who understands which qualities create success may therefore hire for the harder-to-create characteristics and deliberately develop the rest. That is exactly what the manager in the earlier example was doing. He had seen that certain combinations of capability produced strong outcomes. Instead of waiting for perfect external candidates, he looked inside the organisation for people already showing some of those qualities and created an opportunity for them to grow. The recruitment process became an amplifier. The same principle applies to AI adoption. If an organisation introduces AI mainly as a cost-reduction programme, employees will interpret the technology through that lens. They may hide inefficiency because exposing it feels connected to headcount reduction. They may resist sharing knowledge if they believe that knowledge will be used primarily to automate their role. Leaders may then conclude that employees are resistant to innovation when the system has given them perfectly rational reasons to be cautious.

Introduce the same technology as a capability tool and behaviour can be different. Ask where repetitive work is consuming time that people would rather spend on customers, analysis or problem-solving. Let employees help identify tasks that should be automated. Make it clear which decisions remain human and why. Use the technology to make strong people more effective rather than positioning the people themselves as the inefficiency being removed. The technology has not changed. The surrounding system has changed what gets amplified. This is why leadership behaviour matters so much. Leaders create amplification through attention. What they repeatedly ask about becomes important. What they tolerate becomes normal. What they challenge becomes risky. What they recognise becomes desirable. A senior leader who consistently asks only whether a target was achieved can unintentionally amplify target compliance at the expense of understanding how the result was produced.

Ask instead what made the result possible, what was learned, where dependency exists and whether the approach is repeatable, and a different conversation begins. This does not mean leaders should interrogate every success until people regret succeeding. Celebration matters. The point is to recognise that successful outcomes contain information. A strong result is not only something to reward. It is something to understand. The same is true of failure. A repeated failure is evidence that the system is amplifying something undesirable. Perhaps the incentive is wrong. Perhaps ownership is unclear. Perhaps the process rewards speed over quality. Perhaps the organisation has created a workaround culture because resolving root causes is harder to fund than responding to incidents. Punishing the individual event may reduce the visible symptom without changing the amplifier.

This is one reason organisational problems survive multiple restructures. People change. Titles change. Reporting lines change. Yet the incentives, measures, approval patterns and decision habits remain similar. The system continues producing familiar behaviour because the amplifier has not changed. Good leadership therefore requires a systems view. When something repeatedly happens, leaders should look beyond the person performing the behaviour and ask what conditions make that behaviour rational. Why are employees reluctant to take ownership? Why does every risk become an escalation? Why is knowledge concentrated? Why do teams optimise locally rather than collaborate? Why do strong performers become overloaded? Why do people wait until problems become urgent before raising them? Sometimes the answer will involve individual performance. Often it will involve the environment. Amplification makes the distinction particularly important because systems operate continuously. A single poor leadership decision has limited reach. A poorly designed incentive can influence hundreds of decisions every week.

The most powerful amplifiers in an organisation are therefore not always technological. Sometimes they are the rules everyone has learned to follow.

The Economics of Amplification

Amplification ultimately becomes a financial question because leaders allocate finite resources. Training costs money. Coaching consumes time. Documentation consumes time. Cleaning data consumes time. Creating proper governance consumes time. Building a durable process is often slower and more expensive than creating a workaround. Giving a developing employee a difficult task may take longer than giving it to the expert who can already solve it. A careful AI implementation with well-governed data, testing, ownership and controls will almost certainly cost more initially than a quick demonstration built on whatever information happens to be available. The short-term calculation can therefore make weak amplification look efficient. Consider a simple example. A difficult problem appears. An experienced employee can solve it in two hours. A developing colleague could probably learn to solve similar problems, but coaching them through the investigation might require four hours. The expert has to explain the reasoning, allow the learner to attempt parts of the work, correct mistakes and remain involved long enough to ensure the outcome is safe.

If the organisation measures only the cost of today's problem, the answer is obvious. Give it to the expert. Two hours is cheaper than four. Tomorrow another problem appears. Again, give it to the expert. The decision remains efficient. Repeat that calculation fifty times and the organisation will reach the same conclusion fifty times. It will also still have one person capable of solving the problem.

This is where financial analysis can become locally accurate and organisationally misleading. The first approach optimises the cost of individual tasks. The second may be investing in future capacity. After enough coached problems, the developing colleague begins solving some independently. The expert is no longer required for every case. Add several people who have developed similar capability and the organisation begins changing the relationship between demand and specialist capacity. Work no longer queues entirely behind one person. The expert has more time for genuinely exceptional situations. Customers gain access to more capable people. The team becomes less vulnerable to absence or attrition. Future recruitment can draw from a stronger internal talent pool. The initial coaching cost was higher. The long-term organisational cost may be considerably lower.

This does not mean coaching always pays back financially or that every activity should be turned into development. Sometimes the expert should simply solve the problem. Customers should not become training environments for people who are not ready. Urgent situations require speed. Certain expertise may be so specialised that distributing it widely makes little sense. The point is that the calculation should include more than today's task. The same principle applies to knowledge. Writing a useful knowledge article takes longer than answering one colleague's question. If the question is asked only once, documentation may genuinely be inefficient. If the same question is asked fifty times, the economics change. The organisation has a choice between repeatedly purchasing the expert's time or investing once in making some of that knowledge reusable.

Again, the strongest model is rarely documentation alone. Written material becomes outdated, context matters and people still need access to experienced judgement. But the organisation has altered the baseline. Future conversations start from a higher level of shared understanding. This is what amplification does economically. It accepts some cost today in order to change the cost structure of tomorrow. Technology follows the same pattern. A quick AI implementation can look extremely attractive because the visible saving appears almost immediately. A task that took five hours now takes thirty minutes. Multiply the difference by the number of people performing the task and the potential annual saving becomes substantial. That calculation is useful. It should not be the end of the calculation. What does the output require before someone is comfortable using it?

Does a human have to verify everything? How often is incorrect output generated? What happens when the underlying knowledge changes? Who maintains the source data? How many separate teams build slightly different solutions to the same problem? What happens when the person who created the workflow leaves? Is sensitive information being handled appropriately? Does the automation remove work or merely move effort into checking, correcting and maintaining it? A fast implementation can still be the correct decision. Not every use case needs enterprise architecture. The problem appears when the organisation compares the visible implementation cost of a strong foundation with the visible implementation cost of a shortcut while ignoring the repeated costs that the shortcut may create later. This is how quick successes become expensive operating models.

The initial project is cheap. Maintenance becomes somebody else's problem. Human review becomes part of somebody else's workload. Incorrect output becomes somebody else's rework. Security exceptions become somebody else's investigation. Fragmented tools become somebody else's consolidation programme two years later. The cost has not disappeared. It has travelled. This is a pattern organisations experience far beyond AI. A lower-cost implementation can transfer effort into Support. A reduced headcount model can transfer workload into overtime and management attention. Removing a specialist can transfer work into several more expensive generalist roles. Leaving a process undocumented can save time today while increasing every future onboarding cost. Keeping the strongest person permanently assigned to the hardest customers can avoid investment in broader capability while making future progression, succession and customer transition much more expensive.

The recurring error is comparing immediate costs that appear in the same budget while ignoring consequences that appear later, elsewhere or in forms that are difficult to measure. This does not make the short-term decision wrong. Leaders sometimes need short-term outcomes. A customer issue has to be resolved now. A commercial commitment has to be met. A new capability needs to reach the market quickly. The organisation may knowingly choose a workaround because the cost of building the perfect foundation before acting is greater than the risk of moving quickly. That can be sound leadership. The important distinction is whether the shortcut is understood as a shortcut. Temporary decisions become expensive when the organisation forgets that they were temporary. A manual workaround becomes a permanent process.

An exceptional employee's additional responsibilities become the expected role. A prototype becomes production infrastructure. A one-off customer exception becomes the precedent for future contracts. Amplification increases the importance of these decisions because once the workaround begins spreading, the cost of correcting it increases. There is therefore a financial value in good foundations that is easy to underestimate. A foundation rarely creates the most exciting immediate result. Good documentation does not produce revenue simply by existing. A clean knowledge base does not impress customers on its own. A second employee learning a difficult skill may initially reduce productivity. Clear process ownership may require meetings that did not previously exist. Governance can look slower than simply allowing everybody to move independently. Yet these investments change how easily the organisation can scale good outcomes.

This is where the difference between cost and capability becomes important. Cost is what the organisation spends. Capability is what the organisation becomes able to do repeatedly. The two are related, but they are not the same. An organisation can spend very little solving one problem because one exceptional person happens to know the answer. That does not mean it possesses strong capability. It possesses access to a capable person. An organisation can produce an impressive AI demonstration in a few weeks. That does not necessarily mean it possesses an AI capability. It possesses a demonstration. An organisation can deliver a project by relying on extraordinary employee effort. That does not mean it possesses a sustainable operating model. It possesses people willing and able to compensate.

Strong financial leadership should understand those distinctions because capability affects future optionality. If several people can manage complex customers, the organisation can grow more easily. If knowledge is distributed, people can move roles without creating operational crises. If data is governed, new technology can be adopted more quickly because every initiative does not begin by reconstructing the information landscape. If managers develop other decision-makers, senior leaders do not need to scale their own involvement at the same rate as the organisation. Those outcomes have economic value even when the value cannot be expressed with perfect precision. This creates a temptation in the opposite direction as well. Once leaders accept that capability has value, there is a risk of attaching exaggerated financial benefits to everything. Every training programme becomes a multimillion-euro productivity opportunity. Every process improvement is described as transformational. Every avoided incident is assigned the value of the worst possible failure.

That weakens credibility. Amplification does not require pretending the future is certain. A more disciplined approach is to make the cost structure visible. How much expert time is repeatedly consumed because capability remains concentrated? How many tasks require senior review because judgement has not been delegated? How often do new employees ask questions that could have been answered through stronger knowledge? How much rework is generated because an automated process begins from poor information? How much customer risk sits around relationships dependent on one person? How much time do managers spend compensating for hiring decisions that looked cheaper on paper? These questions do not always produce a perfect ROI calculation. They do produce a more complete picture of what the organisation is paying for. Hiring provides a good example.

Imagine two candidates for a customer-facing technical role. One has every technical certification and immediately satisfies the formal requirements but struggles to communicate clearly with senior customers. The other has strong communication, judgement and customer presence but requires additional technical development. The first candidate may look cheaper because the training requirement is lower. But if senior people repeatedly attend customer meetings to compensate, managers rewrite communication, difficult conversations are routed elsewhere and customers escalate because confidence is weak, the organisation has not avoided development cost. It has moved the cost into other people's time. The salary and training budget were visible. The surrounding compensation was not. The opposite can happen as well. Hiring somebody with strong communication but insufficient willingness or ability to develop the required technical understanding can create a different form of dependency. The goal is not privileging one capability universally.

It is understanding which combinations matter for the outcome and how realistically the organisation can develop what is missing. Good amplification therefore requires financial patience, but not financial naivety. Not every investment in people creates a return. Not every foundation deserves to be built to the highest possible standard. Not every process should be automated. Not every expert should spend half their week coaching others. Resources remain finite and leadership still requires trade-offs. The useful question is whether the organisation is repeatedly choosing the cheapest way to complete work or deliberately deciding when work should also create future capability. Those are different objectives.

Sometimes the answer will be no, and efficiency should win. Sometimes the answer will be yes, and paying slightly more today may be the cheaper decision over the life of the capability. The strongest organisations learn to distinguish between the two. They understand that there are moments to consume expertise and moments to multiply it. They know when a quick workaround is sufficient and when the underlying process deserves redesign. They recognise when AI is removing genuine repetitive work and when it is simply accelerating an unstable foundation. They see that giving a difficult task to the best person may produce the safest immediate outcome while giving that same person time to develop others can change what the whole team is capable of next year.

They also understand that none of this happens automatically. Capability does not multiply merely because capable people exist. Knowledge does not spread merely because documentation exists. AI does not create value merely because it is available. Leaders create amplification by connecting useful capability to conditions that increase its reach. That can mean technology. It can mean autonomy. It can mean mentoring. It can mean recruitment. It can mean documentation. It can mean better incentives, clearer ownership or simply noticing that someone has discovered a better way of working and asking what the organisation can learn from it. The financial discipline is understanding what should be built, what should be consumed and what the organisation is really paying when it repeatedly chooses not to build capability.

The cheapest way to use something valuable is often to consume it. Give the problem to the expert. Use the workaround again. Copy the existing process into the new technology. Buy the quick solution. Move on. The more valuable approach can initially look slower. Understand why the expert succeeds. Give them room to develop others. Fix the information before asking technology to rely on it. Build ownership around the process. Accept that the first few repetitions will cost more because the organisation is learning. Eventually, however, something changes. The expert is no longer the only person who knows. The process no longer depends on the workaround. The technology no longer requires constant human compensation. The organisation has not simply completed more work. It has become more capable.

That, ultimately, is what Amplification means to me: recognising what creates value, increasing its reach deliberately and making sure that what becomes stronger is the capability you actually want more of.