Why are we still renting software we can build?

I've been thinking a lot about what AI is going to do to enterprise software. How we use it will change, but I keep coming back to a bigger question: whether we'll need nearly as much of it.

For years, running a company has meant buying a collection of applications. Salesforce for sales, NetSuite for finance, ServiceNow for IT, Workday for HR, and something like Monday.com or Jira to manage projects. Then you hire people to administer those systems, train everyone to use them, and spend a small fortune getting them to talk to each other.

We've done this for so long that it feels like the natural way to run a business. I'm starting to wonder how much of it survives the next few years, and I don't mean because AI agents will suddenly be doing everyone's jobs. What if we simply don't need to buy so much software anymore?

Headless is interesting, but we're still paying the bill

There's been a lot of discussion lately about headless software, MCP, and AI agents that work across applications without anyone having to open them. I get the excitement.

Imagine closing a deal and having an agent handle everything that follows. It reads the contract, creates the customer record, sets up the project, checks who is available, prepares billing and tells the right people. Nobody bounces between Salesforce, NetSuite and a project management system to get things moving. That's a real improvement.

But something about it keeps bothering me. I'm still paying for Salesforce, NetSuite and the project management software, and now I'm also paying for the AI infrastructure sitting on top of all of it. I've made things more efficient, maybe a lot more efficient, without changing the economics of my software stack. In some cases I may have made it more expensive.

Which got me thinking about a different question.

How hard is it to build a CRM anymore?

Think about what most companies actually use their CRM for: accounts, contacts, opportunities, pipeline stages, activities, forecasting and reporting.

Salesforce obviously does a tremendous amount more than that, and nobody is recreating the whole Salesforce ecosystem over a weekend. But how much of Salesforce does the average company use? And how much of what it uses is genuinely hard to build? A CRM is largely a database with business rules, workflows and a user interface. None of that is new technology. What's changed is how quickly we can build it.

I've been experimenting with AI-assisted development: building dashboards, connecting systems, and creating things that would have needed a development team not long ago. Some of what I've built genuinely surprised me. I haven't suddenly become a software engineer; the gap between having an idea and getting something that works has just gotten a lot smaller.

There's a big difference between building something that works and building something you can trust to run a company. Security, reliability, maintenance and integration are real responsibilities, and I take them seriously. I also think we're underestimating how fast the economics of building software are changing.

If I can build a CRM that does exactly what my company needs, connects directly to delivery and finance, and doesn't charge me for dozens or hundreds of licenses, why wouldn't I at least consider it? Five years ago that wasn't a serious conversation. Today I think it is.

And then there's Adobe

Think about how much is packed into Photoshop, Illustrator, Premiere Pro and Acrobat. They're incredibly sophisticated, built over years of engineering and constant refinement.

Now imagine someone building alternatives that cover most of what people actually use, then giving them away or charging a fraction of the price. They wouldn't need every feature, because most customers don't use every feature anyway.

Click to enlarge

AI shifts the equation further, because we may stop working with creative software the way we do now. Instead of learning a complicated sequence of edits, I describe the result I want and an agent uses the underlying tools to get there. Designers and editors still matter; their judgment and craft are the point. But the software they use could look very different, and so could what they're willing to pay for it.

Business applications are the same. I don't really want a CRM. I want to manage customer relationships, understand my pipeline and grow revenue. I don't want project management software; I want to know the work is getting delivered, people's time is well used and customers are happy. And I certainly don't want to spend my time reconciling five applications just to see what's happening in my own business. The software has only ever been how we got to the outcome.

This is where it gets interesting for professional services firms

I've spent a lot of time on how professional services companies operate, and this is where the implications get big. We have a lot of people doing work that exists mainly because our systems and processes are disconnected.

Someone closes a deal. Someone else reviews the contract. Another person sets up the project and someone coordinates resources. Finance needs information from delivery, delivery needs it from sales, and leadership needs reports from everybody. Then we're surprised when things fall through the cracks. We've built whole processes, roles and sometimes departments around keeping these systems in sync.

Now imagine starting a professional services company from scratch. Would you go straight out and buy Salesforce, NetSuite, ServiceNow, Monday.com and a handful of others? Maybe some of them. I'd still want a proven accounting platform; I have no interest in rebuilding anything where financial controls, compliance and auditability are on the line. But a traditional CRM and project management platform by default? I'm not so sure anymore.

I'd start by mapping how the business actually works. How do we win customers? How does an opportunity become a contract, and the contract a project? How do we decide who delivers the work, measure profitability and handle renewals? Then I'd work out the data, rules, workflows and automation that make all of that happen. Some of it I'd buy, some I might build, and more and more I'd look for places where agents could coordinate the work people now do by moving information between applications.

That's the reverse of buying software first and then fitting the company inside it.

Of course, there's a catch

This is where some of the AI conversation gets ahead of itself. Being able to build software more easily doesn't mean we should build everything. Someone still has to maintain it, secure it, test it, monitor it and fix it when it breaks. If your homegrown CRM goes down on the last day of the quarter, you can't exactly call Salesforce support.

Complexity matters too. A CRM that tracks opportunities is fairly easy. One that handles enterprise permissions, complex forecasting, integrations, data quality, audit trails and everything else a mature business eventually needs is much harder. Those are legitimate reasons to buy. I just don't think they settle the argument automatically anymore. The question is moving from whether you can build something to whether building it makes economic and operational sense, and that calculation is changing.

Professional services firms have one more problem. A lot of us haven't documented our processes nearly as well as we think. We rely on experienced people who know how things work, on exceptions nobody wrote down, on teams doing the same work in different ways, and on decisions that depend on knowledge living in someone's head.

You can't throw agents at that and expect a well-run company to come out the other side. If anything, AI is going to expose how loosely defined some of our operations really are. Before we can automate the business, we have to understand how it actually works, and that's probably a bigger job than building the software.

I think the economics of SaaS are going to get uncomfortable

For a long time, software companies had a simple advantage: what they'd already built was expensive to build again. If you needed CRM functionality, buying Salesforce was far more practical than making your own. The same went for project management, service management, creative software and countless other categories.

So what happens when the cost of building comparable functionality keeps falling? When small teams can build applications that needed much larger teams a few years ago? When companies realize they use only part of what they're paying for?

I don't think Salesforce, Adobe, ServiceNow or NetSuite are going to disappear. They have established ecosystems, customer relationships, deep functionality and years of investment behind them. But they're going to face a different kind of competition, and some of it will come from their own customers.

A company that had no realistic alternative to buying may suddenly have options. If enough customers start asking whether they need the full application, the full subscription or all those licenses, that puts pressure on the whole SaaS model. The software isn't becoming worthless, but some of the reasons we've paid so much for it are getting weaker.

What I'm really starting to question

The more I think about this, the less it feels like a conversation about software and the more it feels like one about how companies are structured. We've spent decades building organizations around applications. Sales runs the CRM, finance the ERP, IT the service management platform, and delivery the project management system. Each department has its own tools, processes, data and people keeping it all moving.

If agents can work across those boundaries, and the underlying capabilities can increasingly be built or assembled without buying an entire enterprise application, why would we keep designing companies the same way?

We won't wake up in three years to find everyone running a business with five employees and a hundred AI agents. There's a lot of distance between here and anything like that, and human judgment, relationships, expertise and accountability aren't going anywhere. But some of our assumptions about software, staffing and operations deserve a hard look.

“For years the advice was simple: why build something you can buy?”

That was usually the right answer. Now I find myself asking why we're still renting so much software when building exactly what we need keeps getting more practical.

Canceling every subscription and starting to code isn't the answer. Neither is buying the same applications, layering AI on top and calling it transformation. The more interesting opportunity is somewhere in between, and for professional services firms in particular, I suspect figuring that out will matter a lot more than which AI chatbot we're using this month.