The Next Industrial Revolution — and Why Excel Professionals May Be Standing on the Wrong Side of It
Hiran de Silva | 8 October 2026
Is Reg Putting Excel Professionals Out of Work?
There is a question that will inevitably arise from my forthcoming Tim and Reg conversations.
Is Reg putting Excel professionals out of work?
And, by encouraging people to follow Reg’s example, am I contributing to the destruction of the bread-and-butter work on which much of the Excel community depends?
It’s a legitimate question.
After all, a great deal of Excel work consists of collecting spreadsheets, combining data, transforming it, reconciling it, and producing reports, dashboards and PivotTables.
People are employed to do this work. Consultants are paid to improve it. Trainers earn their living teaching people how to do it more efficiently.
Power Query, dynamic arrays, LAMBDA functions and countless other Excel techniques are promoted as ways of automating these activities.
And rightly so.
But Reg is demonstrating something fundamentally different.
He isn’t merely showing Tim how to perform the same work more efficiently.
He’s showing Tim how to make much of that work unnecessary in the first place.
And that creates a rather uncomfortable question.
If Reg’s approach becomes widely adopted, what happens to all those people whose livelihoods depend on doing the work that Reg has eliminated?
Before answering that, we need to understand why that work exists at all.
The 10% and the 90% — Two Very Different Worlds
I’ve argued for some time that the Excel landscape contains two fundamentally different environments.
For illustration, I call them the 10% and the 90%. These are conceptual proportions, not statistical measurements.
The 10% represents personal productivity.
One person. One computer. One task.
That person might receive spreadsheets, CSV files or reports from elsewhere. They might use Power Query to combine them, dynamic arrays to analyse them, and PivotTables to present the results.
But the essential characteristic is that the work belongs to that individual.
It is not part of a larger process involving other people who depend on its output.
In that environment, almost any method of working is legitimate.
If someone enjoys spending three hours building an elaborate formula to solve a problem that could have been solved in three minutes, that’s their choice.
If they want to automate their work with Power Query, wonderful.
If they prefer to do it manually, that’s their business.
But the 90% is different.
Here, spreadsheets participate in processes involving departments, business units, management hierarchies and entire organisations.
Information moves between people.
Data is collected, consolidated, reviewed, approved, corrected, redistributed and reported.
There are dependencies, deadlines, controls, audit requirements and consequences when things go wrong.
And crucially, the design of those processes is ultimately the responsibility of management.
Management’s job is to organise work.
It is not the responsibility of individual spreadsheet users to decide, independently, how an enterprise process ought to operate.
Yet that is effectively what happens in countless organisations.
People use the techniques they have been taught for personal productivity and apply them to processes that require an entirely different architecture.
The result?
Hundreds of spreadsheets. Endless consolidation. Repeated reconciliation. Fragile links. Manual intervention. And an army of people employed to keep everything functioning.
What I call Excel Hell.
Not because Excel is incapable of doing better.
But because the wrong approach has been applied to the wrong problem.
The Education Industry Has Created an Extraordinary Mismatch
Now consider how Excel education works.
Social media rewards content that attracts the largest audience.
The easiest content to produce and consume is visual, immediate and literal.
Click here. Select this. Enter that formula. Watch what happens.
It’s an excellent way of demonstrating a product feature.
But it is not necessarily an effective way of teaching someone how to design an enterprise business process.
And therein lies the problem.
The people at the broad base of the Excel skills pyramid are naturally interested in techniques that improve their own work.
The social media education industry targets that audience.
That is perfectly understandable.
But who is teaching the people responsible for the 90%?
Who is teaching them how to design an end-to-end process in which spreadsheets communicate with a central relational database?
Who is teaching them to separate the data from the spreadsheets?
Who is teaching them how to use Excel’s long-established capability to GET and PUT data, rather than repeatedly copying and consolidating files?
Who is teaching them to think beyond the boundaries of the workbook?
Very few.
And so an extraordinary mismatch has developed.
The techniques being taught for personal productivity are being deployed in enterprise processes, while the techniques required for enterprise processes are largely absent from mainstream Excel education.
This is not necessarily deliberate.
I believe most Excel educators genuinely want to help people become more productive.
But the cumulative effect of their teaching is that people become increasingly proficient at automating processes that should perhaps never have been designed that way in the first place.
And that’s where Reg comes in.
He doesn’t start by asking how to automate the consolidation of 400 spreadsheets.
He asks why there are 400 spreadsheets that need consolidating.
That is a completely different question.
And it leads to a completely different answer.
We’ve Been Here Before: Fortress Wapping
To understand the consequences, I want to take you back to London in the 1980s.
I was working in the magazine publishing industry at the time.
Newspaper production was undergoing a technological revolution.
Historically, newspapers relied on skilled compositors and typesetters to prepare the material for printing.
The industry had developed around hot-metal typesetting, an elaborate process involving specialist equipment and highly skilled workers.
These were established trades, supported by established working practices and powerful trade unions.
Then computers arrived.
Journalists could prepare their articles electronically. Pages could be composed digitally. Much of the traditional typesetting process could be bypassed.
The implications for the workforce were enormous.
In 1986, Rupert Murdoch’s News International moved production of several newspapers to a new plant in Wapping, East London.
The dispute became known as the Wapping dispute, and the heavily secured production site acquired the nickname Fortress Wapping.
There were mass demonstrations, picketing and violent confrontations.
The underlying issue was not simply a dispute about wages.
It was a struggle over the consequences of technological change.
If technology could eliminate a substantial part of the traditional production process, what would happen to the people whose livelihoods depended on that process?
I witnessed a related phenomenon in the magazine industry.
A friend of mine, Tony Martin, operated a typesetting business called Topaz Typesetting, whose accounts I maintained.
I remember the arrangements surrounding union inspections and the preservation of what was known as the hot-metal rate.
The idea was that established rates of pay should be protected even as the technology and nature of the work changed.
The traditional skills and employment arrangements were being challenged by technology that could achieve the same output through a fundamentally different process.
The conflict was real.
The human consequences were real.
But the technological direction was unmistakable.
Today, digital production is normal.
We do not employ armies of people to assemble newspaper pages using hot metal merely because that activity once provided employment.
And this brings us straight back to Excel.
Are We Protecting Work That Should No Longer Exist?
Imagine a business employing several people to collect spreadsheets from different departments.
They spend their time combining files, correcting inconsistencies, reconciling differences and preparing management reports.
Someone introduces Power Query.
Wonderful!
The process becomes faster.
Someone else introduces dynamic arrays.
Even better!
Perhaps another person develops an impressive LAMBDA function.
The work becomes increasingly sophisticated.
But then Reg arrives.
He asks why the information isn’t being stored centrally, in a structured relational database, accessible to all the spreadsheets that need it.
He demonstrates a hub-and-spoke architecture.
The spreadsheets GET the information they need.
They PUT information back when authorised users make changes.
The central system maintains the data, permissions and audit trail.
The consolidation work largely disappears.
So does much of the reconciliation.
And the process becomes faster, more reliable and more scalable.
Now comes the uncomfortable part.
What happens to the people who previously performed all that work?
The same question was asked about typesetters.
And coal miners.
And workers in industries transformed by mechanisation, computerisation or changes in global production.
It is a serious social and economic question. Some people retrain successfully. Others struggle. Some lose their livelihoods.
Those consequences deserve attention.
But preserving an inefficient business process simply to preserve the work it generates is not a sustainable answer.
And there is another side to the story.
What if the people who currently perform that work could become the people who redesign it?
What if they could become Reg?
Tim’s Discovery — An Entirely New Career
This is the central theme of my Tim and Reg series.
Tim is a capable Excel professional.
He knows Power Query, dynamic arrays and many of the techniques promoted throughout the Excel community.
He is good at what he does.
But he notices something peculiar.
Reg seems to earn considerably more money.
And Reg doesn’t appear to be working nearly as hard.
Tim wants to know why.
Eventually, he has the courage to ask.
Reg explains.
He doesn’t possess some mysterious collection of advanced Excel formulas.
He simply approaches business problems differently.
He understands that Excel is not confined to being a personal document.
He understands how spreadsheets can participate in an enterprise process, sharing centrally managed information through a relational database.
And he knows how to demonstrate that capability to management.
Tim suddenly realises that he has spent years becoming increasingly proficient at solving problems within a boundary he never knew he could cross.
Reg hasn’t discovered a better way to do Tim’s job. He’s discovered an entirely different job.
And that job is far more valuable to the enterprise.
This is why I find the suggestion that Reg is threatening the Excel community so extraordinary.
The real opportunity is not to protect the existing work.
It is to equip Excel professionals to move into a much more valuable area of business problem-solving.
Tim’s future is not necessarily unemployment.
It may be the beginning of the most rewarding period of his career.
But there is another development coming that makes this entire discussion considerably more urgent.
The Next Frontier: Artificial Intelligence
Until recently, there was a practical obstacle to following Reg’s example.
Programming.
Even though Excel has had the capability to communicate with relational databases for decades, implementing that capability usually involves writing some code.
Perhaps VBA and ADO.
Perhaps Office Scripts.
Perhaps Google Apps Script and a Web API.
The programming is often surprisingly simple.
A GET operation might execute a SQL SELECT statement.
A PUT operation might call a stored procedure.
But someone still needs to write the code, understand the connections and handle the responses.
That technical requirement has discouraged many Excel users from venturing into this territory.
Now something remarkable is happening.
AI is beginning to remove much of that obstacle.
We are entering the era of what has become known as vibe coding.
You describe what you want.
AI generates the code.
You test it, refine it and direct the AI to make changes.
Of course, reliable business systems still require validation, security, testing and proper controls. AI-generated code is not automatically production-ready.
But the direction of travel is clear.
The ability to write every line of code yourself is becoming less important.
Which raises a fascinating question.
If AI can write the code, what determines whether you get a good solution or a bad one?
The answer is not necessarily your programming ability.
It is your understanding of the problem.
More precisely, your understanding of what the solution ought to achieve and how it ought to be designed.
And that brings us to what I believe will become the most important distinction in the future of Excel education.
Literal Thinking Versus Lateral Thinking
Much of Excel education is based on literal instruction.
Watch a demonstration.
Follow the steps.
Learn the technique.
Repeat the process.
There is nothing inherently wrong with this.
We all need literal instruction when learning how to operate unfamiliar software.
But knowing how to operate a tool is not the same as knowing when to use it.
And knowing how to automate an existing process is not the same as knowing whether that process should exist.
Lateral thinking requires us to question the underlying assumptions.
Why are we doing this?
Why is the information arranged this way?
Why do we need these intermediate steps?
What happens if the number of participants increases from six to six hundred?
What would the solution look like if we designed it from the beginning, rather than merely improving the existing method?
Most people are capable of both literal and lateral thinking.
But education can encourage one while neglecting the other.
And my concern is that the social media Excel education model overwhelmingly rewards literal demonstrations of product functionality.
It gives people practice in reproducing techniques.
It gives them far less practice in challenging the underlying architecture.
That distinction is about to become enormously important.
Because AI can generate code for almost anything you describe.
But if you describe the wrong architecture, AI may very efficiently build the wrong architecture for you.
Three Challenges That Could Expose the Difference
I want to demonstrate this rather than merely argue about it.
I have several practical examples in mind.
Each involves a business problem that can be approached through familiar Excel techniques, or through a fundamentally different way of thinking.
And each provides an opportunity to test what AI produces.
Challenge One: Account Reconciliation
Consider a familiar accounting problem.
You have two sets of transactions.
You want to reconcile them.
A conventional approach is to place the two sets side by side and find matching items.
Mark Proctor has demonstrated an account reconciliation technique using Power Query. He acknowledges that there are multiple ways to approach the problem.
But I want to challenge the underlying paradigm.
Why should we continue thinking in terms of ticking one list against another?
My alternative is based on three operations:
Reverse Sign. Append. Summarise.
Reverse the sign of the transactions in one dataset.
Append them to the other dataset.
Summarise by the appropriate reconciliation keys.
Where the net balance is zero, the transactions reconcile.
Where a residual remains, there is something to investigate.
Naturally, real-world reconciliation may require additional matching rules, tolerances and exception handling.
But the conceptual difference is enormous.
Instead of trying to reproduce the manual ticking process using a computer, we exploit the mathematics of reconciliation.
And that approach can scale.
I have already demonstrated an AI-assisted implementation.
Now I want to conduct a benchmark.
Give the same reconciliation requirement to several Excel professionals.
Allow them to use any AI system they choose.
Ask them to produce a scalable solution.
Then compare the results.
Will they independently discover the reverse-sign, append-and-summarise approach?
Or will they instruct AI to reproduce the familiar side-by-side matching process?
And if they do, why?
Challenge Two: The Footballers’ Dinner
My second example began as a simple shared-expenses problem.
Six friends go out for dinner.
Some pay for other people.
Money changes hands.
Who owes whom?
It sounds like a spreadsheet problem involving formulas, transformations and perhaps some elaborate Power Query operations.
But the essential problem is accounting.
It can be represented using a simple ledger and double-entry principles.
Once that conceptual leap is made, the problem becomes dramatically easier to scale.
Six friends can become sixty.
Sixty can become six hundred.
My more recent illustration involves Elen’s footballers and a much larger gathering.
The question I want to ask is this:
If someone presents that problem to AI without understanding the underlying accounting principle, what solution will AI produce?
Will it construct increasingly elaborate spreadsheet transformations?
Or will it recognise that the problem is fundamentally a ledger problem?
And if it doesn’t recognise that, what happens when the scale increases?
Challenge Three: Excel as a Database
My third example concerns a familiar subject.
Using Excel to maintain a collection of records.
Mynda Treacy has demonstrated Excel’s built-in Data Form as a way of managing a list of information.
Imagine a chairman maintaining a collection of CDs.
At first, a worksheet might be perfectly adequate.
A list of artists, albums, genres and other details.
The Excel Data Form can help someone navigate and edit that list.
But what happens when the application grows?
What if multiple people need to access the information?
What if the collection becomes part of a larger business process?
The issue is no longer simply how to improve the worksheet.
It is where the information should live and how multiple users should interact with it.
My answer is the Digital Librarian.
A central relational database that stores the information and makes it available to Excel and other applications.
The spreadsheets become clients.
They GET and PUT information through a controlled interface.
Now let’s ask AI to develop the original CD collection application.
Without architectural guidance, will it build a more elaborate standalone spreadsheet?
Will it suggest tables, forms, formulas and perhaps Power Query?
Or will it recognise when a central database becomes the more appropriate solution?
I don’t know.
And that is precisely why I want to test it.
The AI Benchmark That Excel Education Needs
These challenges raise a question that goes far beyond Excel.
Suppose two people have access to exactly the same AI system.
They have the same software.
The same computing resources.
The same business problem.
One understands only the familiar techniques used to solve similar problems in the past.
The other understands the underlying business principles and can recognise alternative architectures.
Will AI give them equally good solutions?
Or will their different understanding produce dramatically different outcomes?
I suspect the latter.
But that is a hypothesis worth testing, not merely asserting.
We can establish a common problem specification, define objective criteria and allow participants to use their preferred AI tools.
We can measure correctness, scalability, maintainability, auditability, security and the amount of manual intervention required.
We can examine the prompts and the reasoning that led to each result.
We can also compare what happens when participants receive additional architectural guidance.
That would be a meaningful test of the relationship between human understanding and AI-assisted software development.
And I believe it could expose a serious weakness in the way Excel professionals are currently being educated.
AI Will Not Automatically Give You Reg’s Understanding
There is an important distinction here.
AI may know how to write VBA.
It may know how to create stored procedures.
It may know how to build a Web API.
It may know how to connect Google Sheets to SQL Server.
But the person directing AI still has to recognise the relevance of those capabilities to the business problem.
Otherwise, the conversation may never go in that direction.
Someone accustomed to thinking of Excel as a standalone document may ask AI to improve that document.
Someone who understands enterprise architecture may ask AI to redesign the entire process.
The first person may receive an excellent spreadsheet.
The second may receive a completely different system.
Both may be technically correct responses to the prompts.
But their value to the business could be vastly different.
And this is the crucial point.
AI does not eliminate the value of knowledge. It changes which knowledge is most valuable.
The ability to remember the syntax of a programming language becomes less important.
The ability to recognise the correct architecture becomes more important.
The ability to follow a tutorial becomes less of a competitive advantage.
The ability to question assumptions, define requirements and evaluate alternative solutions becomes more valuable.
This is not the end of professional expertise.
It is a transformation of professional expertise.
What Does This Mean for the Excel Influencers?
I have repeatedly challenged the direction of Excel education promoted by people such as Paul Barnhurst, Christopher T. Finnell, Mark Proctor and Wyn Hopkins.
My criticism is not that the techniques they teach are necessarily wrong.
Many are entirely valid and useful within the contexts for which they are designed.
My criticism concerns what is missing.
Where is the equivalent enthusiasm for teaching Excel professionals how to participate in enterprise business processes?
Where is the education on separating data from spreadsheets?
Where are the demonstrations of GET and PUT?
Where is the recognition that Excel has long been capable of leveraging relational databases to extend its capabilities far beyond the boundaries of an individual workbook?
Where is the education that connects spreadsheet skills to the responsibilities of senior management?
And now, with AI becoming capable of generating the technical implementation, where is the education in architectural thinking?
These are not peripheral questions.
They go directly to the future career prospects of the people who follow Excel education.
If we continue to train people primarily to reproduce familiar spreadsheet techniques, we risk preparing them for work that will increasingly be automated or redesigned out of existence.
Not because Excel is dying.
But because the nature of valuable Excel work is changing.
The educators who can help their audiences understand that change will be doing them an enormous service.
Those who continue to teach only how to perform existing tasks more efficiently may find that their audiences are becoming increasingly proficient in activities of declining economic value.
That is the uncomfortable possibility I want the Excel community to confront.
The Human Problem Cannot Be Ignored
I want to return briefly to the people who may find this transition difficult.
Throughout history, technological progress has created winners and losers.
Some people adapt quickly.
Others struggle.
I remember a colleague, Len Ford, who worked as a credit controller.
He was comfortable with the printed records and accounting-machine systems he had used throughout his working life.
The arrival of computers was deeply unsettling for him.
The difficulty wasn’t simply learning to operate a keyboard.
It was translating an entire familiar way of working into a completely different conceptual environment.
I remember another historical example: decimalisation in Britain.
On 15 February 1971, the United Kingdom moved from pounds, shillings and old pence to decimal currency.
For people who had spent their lives calculating in pounds, shillings and pence, the change could be disorienting.
I had grown up with decimal currency, so I didn’t face the same mental transition when I later came to Britain.
The arithmetic of the new system was simpler.
But simplicity does not necessarily make a transition easy for someone who has spent decades thinking in another system.
And that is what we must recognise about the coming AI revolution.
People who have spent years learning to solve spreadsheet problems in a particular way may find it difficult to abandon those methods.
Some will embrace the change.
Others will resist it.
And some may be left behind.
That is not a reason to abandon technological progress.
It is a reason to rethink education.
We should be teaching people how to make the transition, not merely teaching them to become faster at the work that is being transformed.
The Future Belongs to Those Who Understand the Problem
The Tim and Reg story began with a simple question.
Why is Reg earning so much more than Tim?
The answer was not that Reg knew more Excel functions.
It was that Reg understood a different way of solving business problems.
He saw beyond the spreadsheet.
He understood the architecture of the process.
He recognised the value that management would place on eliminating unnecessary work, reducing risk and improving control.
And he knew how to demonstrate that value.
Now AI is arriving.
It is removing some of the technical obstacles that previously separated Tim from Reg.
Tim no longer needs to spend months mastering every detail of VBA, SQL or API development before attempting a useful solution.
He can increasingly use AI to help implement what he understands.
But that creates a new dividing line.
Not between people who can program and people who cannot.
Between people who understand what needs to be built and people who know only how to reproduce what already exists.
That, I believe, will become one of the defining professional distinctions of the AI era.
And it has enormous implications for Excel education.
We need to move beyond teaching people how to operate the tools.
We need to teach them how to understand business processes, recognise unnecessary work, challenge existing assumptions and design better systems.
We need to teach them the principles of data management, accounting, process architecture and enterprise information flow.
We need to teach them how Excel can leverage the capabilities of relational databases and other technologies.
And we need to teach them how to direct AI intelligently, rather than merely asking it to automate the familiar.
The Next Industrial Revolution Has Already Begun
The typesetters of the 1980s were not threatened because they had become less skilled at typesetting.
They were threatened because technology had changed the nature of newspaper production.
The same distinction applies today.
Excel professionals are not necessarily threatened because they lack knowledge of the latest Excel functions.
They may be threatened because the business processes in which those functions are deployed are becoming candidates for fundamental redesign.
And AI is accelerating the possibilities.
The question is not whether people should learn Power Query, dynamic arrays or LAMBDA functions.
Of course they should, when those techniques are appropriate.
The question is whether learning those techniques is sufficient preparation for the future.
I don’t believe it is.
The greater opportunity lies in learning to think like Reg.
To see the enterprise process rather than merely the spreadsheet.
To understand the problem before selecting the technology.
To recognise when the correct solution is not a more sophisticated workbook, but an entirely different architecture.
And then to use AI to help build it.
That is where the professional value will increasingly lie.
That is where I believe the next generation of enterprise Excel specialists will emerge.
And that is the opportunity I want to open up to the Excel community.
The future of Excel automation is not simply about automating the work we do today.
It is about discovering how much of that work never needed to exist — and using our understanding, now amplified by AI, to create something better.
Tim has discovered that opportunity.
Reg has been demonstrating it for years.
The question now is whether the wider Excel community is ready to follow.
And perhaps the most revealing way to find out is to give everyone the same business problem, the same access to AI, and see what they build.
That is the experiment I intend to conduct.
Hiran de Silva
Enterprise Excel Consultant | Author of Triple Your Pay with Excel
Exploring the future of Excel, enterprise process design and AI-assisted automation.
Add comment