Back to Overview
Executive Growth & AI Operating Systems

How businesses can grow without scaling their complexity

More orders can create disproportionately more work. How CEOs distinguish variety that customers value from additional internal work, assess the economics of variants and test a limited operational change.

Executive Summary

The essentials in one minute

The management problem

A business sells more, but bespoke commitments create additional approvals, data maintenance and rework. The capacity required grows faster than order volume. The problem does not necessarily lie with the team, but with the variety the operating model has to support.

Why it matters to leadership

An explicitly illustrative calculation shows the mechanism: if the number of orders rises from 1,000 to 1,500 and the proportion of exceptions from ten to twenty per cent, processing time grows from 300 to 600 hours under the assumptions set out in the article. That is 50 per cent more orders, but 100 per cent more work. It is neither an industry benchmark nor a revenue forecast. Meaningful differentiation for customers must be preserved.

What management should do

Select an area of the business that matters for growth. Examine the additional work and economic contribution of its variants, then decide what stays, what becomes a shared component, what is offered on different terms and what is withdrawn in an orderly way. Test the change within a limited scope; after 90 days, compare total work, quality and customer value. AI supports only the rules clarified beforehand. Available time is initially capacity, not automatically a saving.

Revenue is rising, but every additional order takes more coordination. Sales promises a new variant. Operations arranges a bespoke solution. Finance checks a different billing arrangement. The customer gets what was promised, yet nobody knows whether the order justifies the additional work.

Simply hiring more people or automating the queries at this point can mean expanding an expensive operating model. The question that matters is: Which differences create value for customers and which merely create additional work inside the business?

I would answer that question for one specific area of the business before the next decision on scaling. Not by imposing an across-the-board savings target. Some bespoke services are the reason customers buy. Others remain on offer even though nobody has examined their value or cost.

Growth needs variety, but not every bespoke solution

A business entering new customer segments or countries cannot expect to operate in exactly the same way everywhere. Legal requirements differ. A major customer may need a particular interface. A demanding service may justify a higher price. Removing these differences indiscriminately would not be sound management.

The problem arises when every new requirement gets its own process. A small commitment made by Sales then changes data fields, approvals, delivery rules and billing logic at the same time. A variant is not simply developed once. People need training in it, it needs maintaining and checking and it must be considered again whenever subsequent changes are made.

A team can appear highly efficient in this situation. It completes its work faster while the overall process needs more handovers. The costs are spread across Sales, Project Management and Administration. The costing may include only the direct work of delivering the service.

A study published in Management Science in 1996 helps make this distinction. In regression analyses of 57 car assembly plants, greater parts complexity was associated with more adjusted labour hours per vehicle. By contrast, there was no statistically significant relationship between model-mix complexity and labour productivity. The data were collected in 1989 and early 1990. This is a historical observational comparison, not evidence of savings for SMEs today. The parts index covers different forms of internal and customer-driven variety.1

For a CEO, the number of variants alone is therefore not the most useful measure. What matters more is what an additional variant triggers in the business. Does it need a different delivery route? Another data source? A new approval step? Or can it be delivered using components the business already knows how to manage?

Why the work can grow faster than the number of orders

An explicitly illustrative calculation makes the mechanism visible. It is not a client case, an industry statistic or a forecast for your business.

Suppose a business processes 1,000 orders a month. Every order requires twelve minutes of basic processing. Ten per cent of orders need an additional hour of exception handling, for example to clarify requirements, arrange bespoke delivery and carry out checks. These orders still require the twelve minutes of basic processing. The hour is counted on top.

Basic processing takes 200 hours. The 100 exception cases require a further 100 hours. Together, that makes 300 hours in the process being examined.

Now order volume grows to 1,500. At the same time, the proportion of exceptions rises to twenty per cent. Basic processing takes 300 hours and the 300 exceptions require another 300 hours. Total processing time is now 600 hours.

In this calculation, the number of orders grows by 50 per cent, but the processing workload grows by 100 per cent. The cause is not slower basic processing, but a higher proportion of additional work. The calculation assumes unchanged processing times and covers only this process. It says nothing about revenue growth or the staff capacity actually available.

If the proportion of exceptions at 1,500 orders were to return to ten per cent, processing would take 450 hours. The difference of 150 hours would represent capacity released in the calculation. A test must establish whether that is achievable. Whether it produces savings or more revenue is a separate question.

This distinction matters. A free afternoon does not, in itself, reduce staff costs. It may help clear a backlog or process more orders. That use of the capacity then needs to be demonstrated. Development, training and maintenance costs also belong in the assessment.

The cause often lies in individually reasonable decisions

A bespoke offer may make sense to Sales. A separate template saves the project team time in the short term. A local data list helps a branch complete an urgent task. Each decision is understandable in isolation.

But nobody may be deciding on their combined effect. A one-off exception becomes a permanent service commitment. A makeshift list becomes a second data source. The original justification disappears; the maintenance work remains.

Incentives add to this. Someone measured on revenue sees bespoke requests differently from the team that will have to fulfil them. This is not automatically misconduct. But the basis for the decision is incomplete if the sales decision does not account for the work it creates later.

The response can also be wrong. A central function required to approve every deviation creates a new queue. The business needs a rule for routine decisions and a limited route for genuine exceptions. It does not need a meeting for every order.

First, understand the economics of one area of the business

I would start with an area where growth is wanted and queries regularly absorb capacity. This might be the process from order to delivery, customer onboarding or billing for a recurring service. These are possible areas to examine, not claims about specific clients.

Review completed transactions from the first promise to any rework. Record what was ordered, what deviated from it, who had to do additional work and why. Distinguish active processing time from waiting time. A two-day delay may involve ten minutes of work, although both measures can still matter economically.

A few clearly defined measures are enough for the initial baseline: the proportion of exceptions, additional processing time per exception, lead time and the economic contribution of each order. Use observed times for additional work or clearly labelled estimates. Finance must be able to explain which costs have already been accounted for and which are additional, so that nothing is deducted twice.

Do not simply count cases. Assess their importance. A rare customer-specific integration may be worth more than a hundred straightforward small orders. Equally, an apparently profitable service may tie up scarce specialist staff who could make a greater contribution elsewhere.

Missing data is a task to address, not permission to produce a supposedly precise margin figure. The first useful question is: which commitments do we still understand too poorly, in terms of the work they require, to plan further growth?

Four possible decisions about a variant

A variant that customers demonstrably need and the business can reliably deliver should stay. This includes differences required by regulation. Set the conditions under which it will be delivered rather than improvising each time.

Recurring requirements may be organised into shared components. An additional service remains visible to customers but uses the same data fields, checks and handovers internally. For a service business, this could mean standard onboarding with an approved add-on module. Modularisation is worthwhile only if the components are genuinely reused and maintaining them costs no more than the work they replace.

A valuable but demanding bespoke service may need its own price, a longer delivery time or a limit on available capacity. That is a commercial decision. Sales must test whether customers will accept those terms. Evidence of internal costs does not, by itself, establish willingness to pay.

If a variant has no identifiable customer value or economic justification, an orderly withdrawal may be appropriate. Review existing contracts and service commitments. A properly managed transition matters more than an attractive figure in the next report.

None of these decisions should be based solely on the variant's revenue. What matters is its contribution after a reasoned assessment of additional costs, the bottleneck capacity it consumes and its strategic importance. This turns a discussion about reducing variety into a decision about the business you want to run.

A manageable standard and an accountable route for exceptions

The selected area needs a clear standard. What inputs are required? Which services are offered on what terms? Which decisions can the team make itself? At what point is an exception needed?

The route for exceptions should begin before a binding commitment is made to a customer, not only at delivery. For a new bespoke service, record the expected customer value, additional work, pricing assumption and person responsible. A review date prevents a temporary solution from continuing indefinitely.

When exceptions recur, someone must decide whether they should become a new standard component. The overall process is what counts. Faster quotation is not a success if delivery later generates more queries or rework.

The leadership team should therefore appoint a process owner who can assess these effects with the functions involved. This does not mean that the process owner personally approves every exception. It means that changes to the process are not made without considering their downstream costs.

Where AI can help and where it creates new complexity

AI can search completed transactions for recurring reasons for exceptions, make approved service rules easier to find or flag missing information before a handover. This support is useful when sources, access rights and professional review are clear.

But it can also draft new bespoke commitments faster or turn local working practices into additional automations. The work involved in one step then falls while the number of rules and interfaces to maintain grows. An agent for each department is therefore not, in itself, a scalable operating architecture.

Before a test, I would specify which decision is being supported, which source is authoritative and what the system must do when conditions are unknown or contradictory. NIST's AI Risk Management Framework calls for a documented context of use, clear responsibilities and ongoing risk assessment, among other things. It is a voluntary US framework, not evidence of savings or a guarantee of legal compliance or certification.2

The Fassbender AI Operating System connects this work across six layers. Strategy & Value determines which variety supports the business. Processes & Decisions defines the standard and authority to make exceptions. Data & Knowledge keeps service rules and their rationale available. Technology & Integration avoids unnecessary bespoke connections. Governance & Evaluation assesses effects and errors. People & Operating Model ensures that the team can work effectively in the new way.

This is a method for organising how the business operates, not a single software product. AI is used within it only where it better supports the chosen way of working. Removing an unnecessary activity can be more valuable than automating it.

A realistic start over 30, 60 and 90 days

In the first 30 days, management and the process owner select one area of the business. Sales, Operations and Finance establish the baseline together. They document recurring exceptions and check which commitments are covered by existing contracts. At the end, they agree a limited change, such as a shared onboarding module with clearly defined additional options. If there is no adequate cost baseline, establishing it is the first decision.

By day 60, the team tests the new rule on a limited set of transactions. It must remain possible to track comparable cases, customer requirements and volumes. New forms or AI support are used only if required for the test. The previous process remains available as a controlled fallback. Check whether work is simply being shifted to another team and whether customers still receive the service.

By day 90, management decides whether to extend, adjust or reverse the change. A lower proportion of exceptions is not enough on its own. Processing time, errors, customer value and economic contribution must be considered together. Released capacity needs a specified use. If the change is extended, the plan must include training, maintenance of the rules and a further review.

These ninety days are a possible starting point for a clearly bounded process, not a promise to simplify the entire business. Contractual changes, complex integration or missing data may require more time.

When an independent diagnostic is useful

When several functions report additional work but nobody can explain its origin or economic contribution, the Executive AI Diagnostic is a suitable engagement. It provides an independent management diagnostic with an Executive Summary, a baseline across the layers of the AI Operating System, an Opportunity & Risk Map and clear management visualisations. A robust competitive comparison adds context where public information allows. The result is direction for the next ninety days, not a blanket promise of savings.

The starting question remains: which next stage of growth can our operating model support economically? Before buying more capacity, you should know which work actually needs to grow with the business.

FAQ

Does a business have to reduce its offering to become simpler?

No. It can preserve differentiation for customers while using shared components internally. Where a bespoke service makes a demonstrable contribution or is strategically necessary, its complexity is not automatically waste.

Is a higher proportion of exceptions always a warning sign?

It is a reason to investigate, not a verdict. New markets or a different customer mix may explain it. What matters is why the exceptions arise, the capacity they consume and their economic value.

Can we book the processing time saved as a cost reduction?

Not automatically. It initially represents potentially released capacity. A cost reduction requires expenditure that can actually be avoided; additional business requires demand and a positive contribution. The costs of making the change must also be included.

Should simplification come before an AI project?

First, be clear about which work the business wants to do and which rules apply. AI can then support that work. A limited AI analysis may already help with the diagnostic, but it does not replace a decision about what the business wants to offer in future.

Sources

Related Insights

Strategic Advisory

From insight to a decision you can stand behind.

We translate AI potential into clear architecture, measurable priorities and an executable roadmap for your organization.