The MER Gap: A Better Way to Find AI Business Opportunities

Stop asking which experts AI can replace. Ask where expertise actually starts to matter.

The expensive person may be entering the process too early

Many AI business ideas begin with the wrong question:

Can AI replace the expert?

Usually, that is too crude.

A lawyer, engineer, consultant, analyst or auditor may perform a process that takes ten hours, while only two of those hours genuinely require the professional judgment for which the customer is paying.

The other eight hours may consist of finding documents, checking versions, moving information between systems, cleaning data, reconciling formats, constructing tables, preparing comparisons and turning unstructured material into something that can finally be analysed.

The economically interesting question is therefore not:

Can we automate the expert?

It is:

How much of the process can we complete before expert judgment becomes marginally valuable?

The Marginal Expertise Boundary

I call this point the Marginal Expertise Boundary, or MER.

MER is the point in an information-processing chain where the participation of a particular expert begins to materially change the relevance, meaning, reliability assessment, risk evaluation or decision recommendation.

Imagine a simplified consulting process:

Raw information → collection → validation → structuring → comparison → MER → interpretation → recommendation

Everything before MER is not necessarily trivial. It may be necessary, difficult and costly. But it does not necessarily require the expensive expert who currently performs it.

That distinction matters.

The MER gap

Now compare two points in the process:

1. Where the expert currently enters the workflow.
2. Where the expert actually begins to create additional value through specialised expertise.

The distance between them is the MER gap.

A large MER gap means that scarce and expensive expertise is being used for a substantial amount of work before that expertise becomes economically necessary.

This does not automatically mean the work is waste.

The work may still need to be done.

The opportunity comes from changing who or what does it.

Before MER, the lowest-cost competent performer might be:

• a junior employee,
• an assistant,
• a specialised software tool,
• a deterministic automation,
• an AI agent,
• an LLM with human verification,
• or a redesigned process that eliminates the step entirely.

This changes how we should search for AI opportunities

The usual search strategy is:

Find tasks AI can perform.

That produces endless lists of summarisation, document generation, classification and chatbot opportunities.

A more commercially useful search strategy is:

Find expensive outputs where high-cost experts perform a large amount of work before the marginal expertise boundary.

That immediately changes the target market.

You are no longer looking for impressive AI capabilities. You are looking for misallocated expensive labour.

A simple opportunity equation

A rough way to think about the opportunity is:

Business opportunity ≈ Pre-MER cost × volume × repetition × expert scarcity

adjusted downward by:

automation cost + integration cost + verification cost + exception cost + error risk

This produces a much better filter than simply asking whether AI can technically perform a task.

A task can be technically automatable and commercially irrelevant.

Another task can require only modest automation, yet create enormous value because thousands of expensive professional hours are currently consumed before MER.

Example: the €5,000 expert report

Suppose a customer pays €5,000 for a specialist report.

The expert spends:

• 3 hours finding information,
• 2 hours checking sources and versions,
• 2 hours cleaning and combining data,
• 1 hour building comparison tables,
• 2 hours interpreting the evidence and forming a recommendation.

The customer sees a ten-hour expert service.

But perhaps the real MER appears only after hour eight.

If the first eight hours can be transformed into a reliable machine-assisted preparation pipeline, the innovation does not need to replace the specialist.

It only needs to deliver the expert a better starting point.

That can be much easier technically, much safer operationally and much more credible to customers.

MER as a competitor-analysis tool

MER can also be used to analyse competitors.

Traditional competitor analysis asks:

• What does the competitor sell?
• What features does it have?
• What does it charge?
• How is it positioned?

MER analysis asks a more structural question:

Where does the competitor still require expensive human expertise?

Two companies may sell nearly identical outputs at the same price while having radically different internal economics.

Company A may use a senior consultant for twelve hours.

Company B may use software and AI for eight hours of preparation and the senior consultant for four hours of judgment.

To the customer, the service may look similar.

Economically, it is not the same business.

Look for expert-heavy markets with narrow expertise cores

The most interesting markets may therefore be those in which the visible product appears highly expert-intensive, while the genuinely expertise-sensitive portion of the workflow is relatively narrow.

Potential examples include:

• engineering documentation,
• technical design preparation,
• legal research,
• compliance work,
• financial analysis,
• insurance processing,
• due diligence,
• procurement analysis,
• consulting research,
• audit preparation,
• planning and permitting processes,
• and other information-intensive professional services.

The strongest opportunity is not necessarily where AI is most intelligent.

It may be where the current production system uses the most intelligence unnecessarily.

MER also exposes complexity premiums

There is another useful consequence.

Companies often accumulate layers of technology, workflow, coordination, specialised software, review procedures and professional roles around an output.

Some of this complexity is necessary.

Some of it may be historical residue.

MER gives us a way to ask:

If the economically essential expertise is concentrated in only a few decision points, how much of the surrounding architecture can be removed without materially reducing customer value?

That is not merely an automation question.

It is a business-model simplification question.

Do not automate the workflow before understanding the boundary

MER analysis also prevents a common AI mistake: automating an existing process simply because the process exists.

Before automating a step, ask:

1. Does this step affect the final decision?
2. Does it genuinely require specialised expertise?
3. Can its input and output be specified?
4. Can exceptions be detected?
5. Can the result be tested?
6. What happens if it is wrong?
7. Can a cheaper competent performer handle it?
8. Should the step exist at all?

Sometimes the correct solution is AI.

Sometimes it is conventional software.

Sometimes it is delegation.

Sometimes it is a better form, better metadata, a shared data model or a standardised input.

And sometimes the correct solution is simply to delete the step.

A new way to search for businesses

MER suggests a systematic business-discovery process:

Market → expensive output → workflow → individual transformations → MER → pre-MER cost → cheapest competent performer → new production model

This is a more disciplined approach than brainstorming AI products.

Instead of starting with technology and searching for a problem, start with a costly production structure and search for the point where expensive expertise becomes genuinely necessary.

The strategic question

The next generation of AI-native professional services may not win because they remove experts.

They may win because they use experts only where experts create marginal value.

That creates a simple strategic question for almost any knowledge-intensive industry:

How far can the work travel before it needs the expensive human?

Wherever the answer is “much further than it travels today,” there may be a business opportunity.

Flutterby Ninja Consulting
Simplify the system. Find the real bottleneck. Put intelligence where it actually changes the outcome.