Here's a thing that's true now and wasn't ten years ago: almost any single software feature can be rebuilt by a competent team in a few weeks. A kanban board, an email composer, an invoice generator, a booking calendar — these were defensible products once. Today they're commodities, table stakes, weekend projects with AI assistance. If your moat is a feature, you don't have a moat. You have a head start, and head starts expire.
Where the value went.
Value doesn't disappear when something commoditizes; it moves up a layer. The features commoditized, so value moved to the thing features can't capture on their own: workflow — the way jobs connect into a coherent flow across a business. Anyone can build an invoice. Almost nobody can make the invoice know about the deal that created it, the project that justified it, and the support history that follows it. That connective tissue is the product now.
A feature is a noun. A workflow is a sentence. Competitors can copy your nouns. They can't copy how your sentences connect.
This is why a pile of best-of-breed tools, each excellent in isolation, can lose to a platform whose individual features are merely good. The platform wins on the sentences — the way a closed deal becomes a project becomes an invoice becomes a renewal — not on any single word. The best-of-breed stack has better words and no grammar.
Why workflow can't be commoditized.
Features commoditize because they're self-contained — you can specify one completely and rebuild it. Workflow resists because it's relational. It depends on a shared data model, consistent identity across jobs, and a thousand small agreements about what a “customer” or a “project” means everywhere it appears. You can't clone that in a weekend because it isn't a thing — it's the relationships between things, and relationships are exactly what doesn't copy-paste.
What this means for buyers.
Stop shopping for features. You will always find a tool with a deeper version of any single capability, and it will always be tempting, and it will almost always be the wrong call below 50 people. The question that actually predicts whether software will help you is: does this connect the jobs I do, or does it do one job in isolation and leave the connecting to me? The connecting is the work. The connecting is the cost. The connecting is the moat.
We build features, and we try to build them well. But we don't pretend the features are the moat. The moat is that they share one data model, so the workflow is native instead of assembled. In a world where software is a commodity, that's the only thing left worth charging for.