TL;DR:

• For 40 years, Excel has been the cockroach of enterprise software. It fills every gap that off-the-shelf tools can’t reach, and no platform has been able to kill it

• SaaS isn’t the problem. Salesforce, HubSpot, Monday.com fill a lot of needs. The problem is everything they weren’t built for. Companies are stuck choosing between generic SaaS that doesn’t fit, ugly vertical software that stagnates, or custom builds that cost too much. Most pick the first two and duct-tape the gaps with spreadsheets

• AI coming for software engineering in full has changed the math on option three. What used to take 18+ months and $750k+ can now be built in months for a fraction of the cost and more easily maintained

• But a working prototype isn’t a production system. The hard part isn’t building software. It’s understanding how work actually moves through your business and designing something people will use

• AI isn’t killing SaaS. It’s killing the spreadsheet’s role as duct tape between systems. The companies that move first stop fighting that gap and start building the connective software they’ve always needed but could never justify

“But where they have failed, you will succeed”

September 1985.

Back to the Future was still #1 at the box office, 13 weeks after release.

Weeks after the world met Marty McFly, it met something else.

Microsoft launched Excel.

At the time, Windows didn’t exist yet and MS-DOS was a text-based operating system. And Lotus 1-2-3 was crushing them in spreadsheet market share.

Rather than compete head-on, Microsoft flanked them, launching Excel on Apple’s Macintosh, a platform Lotus had completely ignored.

By the time Lotus realized graphical interfaces were the future, it was too late. Excel hit Windows in 1987 with a two-year head start.

Now, over 40 years later, Excel still dominates every person’s work week.

No matter the department, no matter what other software the company has, everyone falls back to Excel. And plenty have tried to kill it.

Software vendors have been playing Morpheus for decades, pitching their platform as The One.

In The Matrix, Morpheus tells Neo he’s the person who will finally defeat the machines. “I’ve seen an Agent punch through a concrete wall. Men have emptied entire clips at them and hit nothing but air. But where they have failed, you will succeed.”

Every few years, a new platform shows up with the same promise. The software that will finally take out the spreadsheet. CRMs, ERPs, project management tools, no-code builders.

Executives buy in, roll it out, and wait for Excel to die.

It never does. Because they’re fighting head-on. Trying to replace the spreadsheet instead of filling the gap it exists to cover.

Until now.

The same flanking maneuver Microsoft used in 1985 is available again. And it’s not another SaaS platform. It’s the ability to build custom software fast and cheap, something that wasn’t possible until AI showed up.

A Cockroach Infested House

LinkedIn is full of people claiming you can ‘build your own Salesforce in a weekend’ with AI. That’s mostly nonsense. But something real is happening.

I sit in the middle of this every day with clients: on one side, off-the-shelf SaaS that somewhat works but never quite fits. On the other, custom software that would solve everything but costs too much and takes too long.

In that gap? Excel. It’s always Excel.

And no one is ever happy.

The symptoms are as old as time:

A business development team that stopped using half the CRM fields because it’s faster to update their own spreadsheet, so now the weekly forecast meeting is just everyone arguing about which numbers are right.

An HR team planning headcount in a spreadsheet because no one can answer ‘are we on track with our hiring plan for Q2?’ without pulling data from three systems and five meetings.

A services firm managing jobs in Excel sheets passed around all day, each person adding their piece just to unlock the next step, with a daily meeting dedicated to figuring out which data points are actually current.

Project managers maintaining a ‘real’ resource schedule in Excel while the PM tool shows the ‘official’ one no one trusts, updating both every week just to keep the peace.

Customer success tracking renewals in a spreadsheet because their support platform doesn’t integrate with billing or provide any insights on customer churn.

I could keep going. You have your own versions of this.

More often than not, people forget to add in data points or change a formula that “only Frank in finance knows how to modify” and they spend hours every week chasing each other to fill in or update these spreadsheets.

Excel is the cockroach.

It fills the gaps SaaS can’t reach and custom software is too expensive to cover.

And like a cockroach, it’s almost impossible to kill. And it creates a host of issues that block teams from modernizing their business and its technology.

$63 Billion in Ugly Software

So if Excel keeps filling the gaps, why hasn’t better software come along to close them?

There’s a company called Constellation Software most people have never heard of.

They have a market cap of about $63 billion. They got there by acquiring hundreds of small software companies that make tools that look like this:

Ew. Software for home service companies. Software for agencies. Software for logistics companies. Software for campgrounds, storage facilities, and golf courses. Purpose-built tools for specific industries that are large, but seemingly not large enough to hire an actual product designer or respond to a customer’s feature request submitted this decade.

Their whole model is buying niche software companies and holding them forever. They’ve acquired over 500 of them. And they keep printing money while customers are neglected and the product stagnates.

Why do companies buy this stuff at all?

Because the alternative of “general software” isn’t much better.

Salesforce wasn’t built for insurance brokerages. Monday.com wasn’t built for property managers. HubSpot wasn’t built for third-party logistics companies.

These general tools work great for the narrow use cases they were designed for, but the minute your workflow gets industry-specific, they break down. You end up forcing your process into someone else’s software, turning off half the features (that they still make you pay for), and filling the gaps with spreadsheets.

So companies turn to vertical software. It’s ugly and stagnant, but at least it works a bit better for their industry. And once they’ve paid to implement it and adapted their processes to it? Switching feels harder than suffering. So they stop questioning it.

Vertical software exists because generic SaaS doesn’t fit. Constellation got rich because they recognized the gap and bought up every solution filling it then put minimal effort into improving it.

So companies have three options:

Force-fit generic SaaS and live with the gaps (hello, spreadsheets)

Buy vertical software that’s expensive, ugly, slow to update, and barely integrates with anything else

Build custom software, which historically cost exorbitant amounts

Most non-enterprise companies pick option 1, supplement with option 2, and rarely, if ever, consider option 3.

But first, let’s talk about what this all leaves you with.

DataLake.xls

A few months ago I was on a call with an exec exploring our services. Mid-sized company, just over 150 people. He mentioned they had great data infrastructure. “We have a data lake,” he said. At that team size, I was impressed. I asked if I could see it.

He shared his screen and pulled up an Excel file called DataLake.xls.

Twenty tabs that their entire team spent hours every day chasing each other to update. Connected to nothing. Dozens of people across the company spent hours every week exporting data from other tools, pasting it in, cleaning it up, and hoping nothing broke or was missed. That was the “data lake”.

This is what option 1 + option 2 gets you. A patchwork of SaaS tools that don’t talk to each other, vertical software that doesn’t integrate with anything, and Excel duct-taping it all together.

This same exec wanted robust reporting dashboards, AI-powered insights, and automated workflows that didn’t require human babysitting. But you can’t build any of that on a foundation of spreadsheets held together by copy-paste and good intentions.

You can’t automate a workflow that lives in six spreadsheets emailed between departments. You can’t train AI on data that “lives in Frank’s head.” You can’t build a dashboard when half the numbers are in a neglected CRM and half are in Sarah’s Excel file on her desktop.

The spreadsheets aren’t just annoying. They’re the bottleneck blocking everything else.

There’s a scene in The Office where Michael Scott stands up and yells “I declare bankruptcy!” Oscar has to explain that you can’t just say the word bankruptcy and expect anything to happen. Most companies treat data strategy the same way. They say “we’re data-driven” while fifteen versions of “the real numbers” float around in inboxes and folders.

This is what’s actually missing in most businesses: an orchestration software that connects workflows across teams and systems, and a data layer that gives you one source of truth instead of fifteen spreadsheets.

Without those two things, every modernization effort stalls. Execs fall prey to shiny tool syndrome, buy a new SaaS that doesn’t integrate, and six months later there’s another spreadsheet filling the gap.

Until now.

The Flamethrower

For most of business history, custom software was a luxury. You needed a small army of developers, millions of dollars, and a prayer that what got built actually matched what you asked for.

It was like funding an expedition. You’d assemble a team, pour in resources, and hope they came back with what you asked for instead of scurvy and excuses. So most companies just...don’t. They buy off-the-shelf tools, live with the gaps, and hire people to be the glue between systems. The spreadsheet became the universal adapter.

But something has shifted.

Think about what happened to publishing. For centuries, producing a book meant hand-copying manuscripts. Monks spent years on a single text. Then the printing press showed up, and suddenly books could be produced in days instead of decades. The skill wasn’t gone, it was just no longer the bottleneck.

AI coding tools are the printing press for software.

A year ago, building a custom tool to sync your CRM with your invoicing system meant hiring developers, writing a spec, and waiting months. Now, someone can describe what they need in plain English and iterate on working software that can scale to a whole company in weeks. The code still needs to be reviewed, tested, and maintained properly. But the gap between “I need this” and “I have this” has collapsed.

You know that scene in Raiders of the Lost Ark? A swordsman does an elaborate, intimidating display with his blade. Indy just pulls out a gun and POW! For decades, software-as-a-service has been the swordsman. Impressive demos, complex implementations, long, expensive contracts for inflexible software you couldn’t truly customize. Companies watched the performance because they thought it was the only option. Now there’s a gun.

What used to require a team of developers and a year and a half can now be built in months. What used to cost $750k+ can be done for a fraction of that.

And unlike the vertical software Constellation has been acquiring for decades, this stuff can actually look good, integrate with modern tools, and be updated without submitting a ticket and waiting three months for a canned reply of, “we’ll think about putting that on the roadmap.”

The gap that Excel has been filling for 40 years? It’s finally closeable.

Not because AI can “build Salesforce in a weekend.” That’s still nonsense. But because the custom tools and unique orchestration layer your business actually needs, the one that connects your CRM to your invoicing to your capacity planning, the one no vendor will ever build because your workflow is too specific? That’s now buildable. And easily maintainable.

Here’s what makes this moment different from the usual hype cycle: the number of processes in work and life that haven’t been meaningfully improved by software is still far greater than the number that have. We’re nowhere near saturation. The demand for custom software has always been massive. The bottleneck was never the need. It was the cost and timelines.

This won’t happen overnight. These transitions never do. But the economics have shifted, and the companies that figure this out first are building exactly what they need, fast, while everyone else argues about which SaaS platform to buy.

The cockroach finally has a flamethrower predator.

“Roads? Where we’re going, we don’t need roads”

AI didn’t hand everyone a “build me perfect software” button.

You’ve probably seen the demos. Someone describes an app in plain English, and ten minutes later they have a working prototype. That’s real, and it’s exciting. I’d genuinely encourage anyone curious to try it.

But there’s a meaningful gap between a prototype running on your laptop and a system your whole company depends on.

A prototype doesn’t have real data. It doesn’t handle the edge case where a client has two accounts with slightly different names. It doesn’t know that your invoicing workflow has a manual approval step that only applies to projects over $50k. It doesn’t sync with your CRM, your billing platform, and your capacity planning sheet in real time. And it definitely doesn’t hold up when forty people are editing data, running automations, and relying on it to be accurate at the same time.

It’s a bit like WebMD. You can Google your symptoms and get surprisingly far. Sometimes that’s all you need. But when the stakes go up, when it’s your company’s core operations and not a side project, you want someone who’s seen the patterns across dozens of companies and knows which symptoms point to which root causes.

The hardest part was never the code. It was figuring out how work actually moves through a business. Where data breaks, where people have quietly built workarounds no one talks about, and designing something they’ll actually use instead of route around. AI made the building radically faster. It didn’t change the diagnosis.

That’s what we do at Switchboard. I’ve spent over a decade building software for Fortune 50s and startups alike, and the last few years specifically helping mid-sized companies unpack their operations and build the connected systems they couldn’t justify before. Real data, real integrations, real business logic. Took 5 months. Three years ago, that same project would’ve taken 18+ months. If it got approved at all.

The economics of custom software have fundamentally changed. The companies that move first will stop fighting spreadsheets and start building competitive moats. They’ll have systems that actually talk to each other. They’ll make decisions with data they trust and own.

Forty years is a long time for a cockroach to run the place. We think it’s time for an exterminator.

Ps, we’re doing exactly this at Switchboard if you wanna learn more. Or ping me on Twitter or LinkedIn any time.