Should You Buy SaaS, Automate It, or Build Custom Software?

A practical guide to deciding what your business actually needs

There is a particular moment in every growing business when technology stops feeling helpful and starts feeling like a small administrative hostage situation.

You have a CRM.

You have an accounting tool.

You have a project management tool.

You have a WhatsApp group called “URGENT.”

You have three spreadsheets.

One of those spreadsheets is maintained by a person who is on leave today, and nobody knows how it works.

Then someone says:

“We should probably get custom software.”

Maybe.

Or maybe you need automation.

Or maybe you just need to stop paying for six tools that do 80% of the same thing.

That is the real problem.

Most businesses don't struggle because they lack software.

They struggle because they don't know which parts of their business should be handled by software, automation, or custom systems in the first place.

And that distinction can save you a very large amount of money.

The three choices

When a business has a process that is slow, messy, expensive, or painful, there are usually three ways to fix it:

Buy it.

Automate it.

Build it.

They sound similar.

They are not.

Think of them like this:

SaaS solves a common problem.

Automation solves repetitive work.

Custom software solves a unique problem.

The mistake is using one to solve another.

Buying software for a process that is highly unique usually creates awkward workarounds.

Building custom software for a problem that thousands of businesses already solved is expensive reinventing of the wheel.

And adding AI to a process that should have been a simple rule like “when X happens, do Y” is just putting a tuxedo on a toaster.

So let's break it down.

1. Buy SaaS when the problem isn't special

Software as a Service, or SaaS, is usually the smartest option when you're solving a standard business problem.

Need accounting?

There is software for that.

Need email marketing?

Software.

Need employee payroll?

Software.

Need project management?

Software.

Need CRM?

Please, for the love of all remaining spreadsheets, do not build one from scratch.

If the process is common, mature SaaS products are usually better because they already include years of development, security work, integrations, maintenance, updates, and boring edge cases nobody wants to discover at 2:13 AM.

For example, suppose you run a small agency.

You need:

• project tracking

• invoicing

• customer communication

• internal task management

• file sharing

None of those things are your competitive advantage.

They are infrastructure.

You should probably buy them.

The important question is:

Does this software solve our problem well enough without forcing the business to change how it fundamentally works?

That last part matters.

Because SaaS has a hidden cost.

And it isn't the subscription.

It's the workaround.

2. Watch out for the “we'll just work around it” trap

This is where things get interesting.

A SaaS product might handle 80% of your needs.

That sounds fantastic.

Until your employees spend every day dealing with the other 20%.

Maybe your sales team copies customer information from the website into the CRM manually.

Then someone exports it to Excel.

Then finance cleans the Excel file.

Then operations uploads it somewhere else.

Then a manager asks:

“Why doesn't the dashboard match the report?”

And five people suddenly stare at each other like contestants on a game show.

The software itself isn't necessarily bad.

The problem is the gaps between systems.

Those gaps become manual work.

And manual work becomes expensive.

A useful way to think about this is:

Software cost = subscription + integration cost + human workarounds

Sometimes the subscription is the cheapest part.

3. Automate when humans are doing the same thing repeatedly

Automation is useful when the process already makes sense, but humans are repeatedly carrying it out.

This is a very important distinction.

You may not need custom software.

You may not need AI.

You may just need the computers you already own to stop watching humans click the same seven buttons every afternoon.

Suppose this happens:

New order arrives

→ Someone checks payment

→ Someone updates the order system

→ Someone sends confirmation

→ Someone updates inventory

→ Someone informs the team

→ Someone sends a WhatsApp message

→ Someone changes the order status

That is not necessarily a custom software problem.

It is an automation problem.

A useful automation often looks like:

Trigger → Data → Rules → Action

For example:

New order received
↓
Payment verified
↓
Inventory checked
↓
Order recorded
↓
Customer notified
↓
Operations notified

Once the logic is clearly defined, there is often no reason for a human to manually shepherd information from one system to another like an underpaid messenger pigeon.

4. Automation is not the same thing as AI

This distinction is becoming increasingly important.

Because apparently every workflow now needs “AI” in its marketing description.

It doesn't.

Consider this:

When an invoice is uploaded, extract the invoice number and amount, then send it to the accounting system.

That's automation.

You might use AI for extracting messy information from an unstructured PDF.

But the overall workflow doesn't need to become an autonomous AI agent just because the invoice contains the word “Total.”

Traditional automation is often better when the rules are predictable.

AI becomes useful when the process involves things like:

• unstructured documents

• natural language

• classification

• summarization

• interpretation

• decision support

• reasoning across multiple pieces of information

Imagine a customer sends:

“Hey, the payment went through but my order still says pending and I need it before Friday.”

A normal rule-based system may struggle with that.

An AI-powered workflow can potentially understand the message, identify the problem, retrieve order information, determine what happened, and route or resolve the issue.

That is where AI starts becoming interesting.

The lesson is simple:

Don't use AI where a rule will do. Use AI where understanding is the problem.

That one decision can prevent a lot of unnecessary complexity.

5. Build custom software when your business is the exception

Now we get to the expensive one.

Custom software makes sense when your business doesn't fit neatly into existing software.

Not because custom software is cooler.

Not because your founder wants a dashboard with dark mode.

Not because someone on LinkedIn said “build your own internal platform.”

Build when the workflow itself matters.

A strong candidate for custom software usually has one or more of these characteristics:

Your workflow is genuinely unique

You operate differently from the average business in your category.

Existing tools force too many compromises

You keep adapting the business to match the software rather than the software supporting the business.

Multiple systems need to behave like one system

You have CRM + ERP + inventory + payments + communication + internal operations, and the glue holding everything together is five people.

The process is business-critical

If the workflow breaks, revenue, operations, customer experience, or compliance suffers.

The software itself creates competitive advantage

This is the biggest one.

If the software isn't just supporting your business but is part of what makes your business different, building custom becomes much easier to justify.

6. The real question isn't “Can we build it?”

You can build almost anything.

That's no longer the interesting question.

The interesting question is:

Should this specific problem become software?

A company can waste a fortune building something that doesn't need to exist.

Let's say your team uses a spreadsheet once a month to calculate an internal metric.

Could you build a beautiful dashboard for it?

Absolutely.

Should you?

Probably not.

Now suppose 30 employees use that spreadsheet every day, information is copied between systems, mistakes cause financial problems, and managers wait hours for reports.

That's different.

The process has become important enough that software can create meaningful value.

This is why the decision shouldn't start with technology.

It should start with economics and friction.

7. The hidden enemy: operational friction

Imagine a company where every customer requires 15 minutes of manual administrative work.

At 10 customers per day:

150 minutes.

At 50:

750 minutes.

At 100:

1,500 minutes.

You didn't necessarily need to hire another person.

You may have needed to remove the repetitive work.

This is why good software decisions aren't really about software.

They're about time.

Time spent copying.

Time spent checking.

Time spent reconciling.

Time spent searching.

Time spent asking someone for information that should already exist somewhere.

Time spent fixing mistakes caused by all of the above.

Businesses often calculate the cost of software.

They should also calculate the cost of not changing the process.

8. A simple example

Imagine an online business.

Their stack looks like this:

Shopify

Payment gateway

CRM

Accounting software

WhatsApp

Inventory system

Everything technically works.

And yet every day:

Someone exports orders.

Someone checks payments.

Someone updates inventory.

Someone sends customer messages.

Someone reconciles transactions.

Someone generates a report.

Nothing is “broken.”

But the workflow is broken.

This is where a hybrid system often wins.

You don't replace everything.

You connect what already works.

For example:

Shopify handles commerce.

Payment gateway handles transactions.

Accounting software handles finance.

CRM handles customer relationships.

Then a custom integration layer connects them.

An automation handles routine events.

AI handles the messy human-language or document-heavy parts.

Now the system is doing the work instead of your employees becoming the API.

9. The HikariWorks rule: Buy. Automate. Build.

At HikariWorks, we like a much simpler way of looking at the decision:

Buy the commodity.

Use existing software for problems that are already solved well.

Automate the repetitive.

Remove human involvement from predictable, repetitive workflows.

Build the differentiating.

Create custom software when your workflow, customer experience, or operational model genuinely requires it.

That's it.

Not every business needs an app.

Not every business needs AI.

Not every business needs custom software.

Sometimes the best technology decision is buying a boring piece of SaaS and going home.

That's a surprisingly difficult concept for technology companies to admit.

10. The seven-question test

Before spending money on development, ask these questions.

1. Is this a common business problem?

If thousands of businesses have the same problem, look for existing software first.

2. Are people repeatedly moving information between systems?

That's a strong automation signal.

3. Are humans making the same decisions using predictable rules?

Automate it.

4. Does the process involve messy language, documents, or interpretation?

AI may be useful.

5. Are we changing our business process just to fit the software?

That is a warning sign.

6. Does this process affect revenue, operations, customer experience, or competitive advantage?

The more important the process, the more valuable a custom solution can become.

7. Would owning the software make the business meaningfully better?

If yes, custom development becomes much more interesting.

If the answer to most of these is “no,” you probably don't need a custom platform.

Congratulations.

You just saved yourself a lot of money.

11. What businesses often get wrong

There are a few common mistakes we see again and again.

Building too early

A business discovers a problem and immediately wants software.

But sometimes the process itself is still changing.

Automating a bad process just helps you perform the wrong process faster.

First understand the workflow.

Then improve it.

Then automate it.

Then decide whether it deserves custom software.

Buying too much SaaS

The opposite problem is just as common.

A company keeps adding tools:

CRM.

Helpdesk.

Analytics.

Inventory.

Automation.

Communication.

Reporting.

Another CRM.

A specialized CRM for the CRM.

Eventually the business doesn't have a software stack.

It has software archaeology.

Each tool solves one piece.

Nobody remembers why half of them exist.

Adding AI because AI sounds impressive

This one is particularly dangerous.

A simple workflow doesn't become better because you insert a language model into it.

If the rule is:

“When payment succeeds, send confirmation.”

Use a rule.

Don't ask an AI model to contemplate the spiritual meaning of a successful payment.

Use AI where ambiguity exists.

12. Real businesses rarely fit into one box

This is the part that gets overlooked.

The answer isn't always:

SaaS OR automation OR custom software.

Often it's:

SaaS + automation + custom software + AI.

For example, a business could have:

CRM
for customer management

SaaS accounting
for finance

Custom backend
for proprietary business logic

Automation
for repetitive workflows

AI agent
for customer requests and operational tasks

That's not overengineering.

That's simply assigning the right tool to the right problem.

The trick is knowing where the boundaries should be.

13. So, what should you do?

Before buying another tool or commissioning a six-month software project, map your workflow.

Write down:

What happens?

Who does it?

What information is involved?

Which systems are used?

Where does someone copy something manually?

Where do mistakes happen?

Where do people wait?

Where do customers wait?

What decisions happen repeatedly?

You will usually discover one of three things.

A process that should simply use existing SaaS.

A process that should be automated.

Or a process that is important enough to justify custom software.

And sometimes you'll discover something even more valuable:

the process itself is the problem.

No software can save a workflow that nobody can explain.

The bottom line

The best technology isn't the most advanced technology.

It's the technology that removes the most unnecessary work while creating the most useful capability.

Buy software when the problem is already solved.

Automate when humans are doing repetitive work.

Use AI when understanding or reasoning is the hard part.

Build custom software when the way your business works is important enough to own.

And when all four belong together, use all four.

The goal isn't to have more technology.

The goal is to have less friction.

Because your employees should be spending their time making decisions, solving problems, talking to customers, creating things and growing the business.

Not copying a customer ID from Excel into a CRM.

Again.

For the 900th time.

That's not digital transformation.

That's just a person being used as an API.

At HikariWorks, we build technology around the business, not the other way around.

Buy the commodity. Automate the repetitive. Build the differentiating.