Opinionated SaaS Is the New Luxury
When you buy B2B software, do you actually want “flexibility”… or do you want fewer decisions to make?
The question I keep asking before I buy B2B software
Let me start with a slightly uncomfortable question: when you buy B2B software, do you actually want “flexibility”… or do you want fewer decisions to make?
Because most of the time, I'm not paying for a tool. I'm paying for an outcome. I don't want a “framework” that lets me design the perfect process if I'm honest with myself. I want something that gets my team from messy reality → repeatable results with the least drama possible.
And yet, a lot of SaaS still sells the same promise: “We won't force you into a process. Configure it your way.”
That used to be impressive. It still sounds polite. But in 2026 it's starting to feel like a red flag. Not because flexibility is bad. But because flexibility has become… cheap.
AI made configurable software less impressive
A big shift has happened quietly: the “build and customise” work that made configurable SaaS valuable is now much easier to do in-house. We're not guessing here. There's serious evidence that AI coding assistants speed teams up:
- A controlled experiment published by Microsoft Research found developers with GitHub Copilot completed a coding task 55.8% faster than the control group.
- A field-experiment write-up from MIT researchers found measurable productivity lifts after adoption—e.g., more pull requests per week in real organisations.
So when a buyer looks at a “configure-everything” SaaS, they're increasingly thinking: “Why am I paying per seat for a configuration engine… when my team can build 80% of what we need quickly, and own it?”
Even the classic poster-child platforms for configuration are basically saying the quiet part out loud. For example, Salesforce describes its platform as metadata-driven, enabling tenants to customise apps and logic quickly using metadata. And their own training materials highlight how much can be built with “no-code” / “point-and-click” customisation because of that architecture.
The uncomfortable implication:
If your core product value is “we give you knobs and a workflow builder”, AI is going to put downward pressure on what you can charge—because building knobs got easier. Meanwhile, the organisations that are getting value from AI at scale tend to be the ones with clear processes for validation, governance, and adoption. That's a hint about where the defensibility is moving.
Flexibility has a hidden tax
Even if AI didn't exist, extreme configurability has a long-standing problem.
It pushes complexity onto the customer
Configurable SaaS frequently behaves like unpaid labour. Someone becomes the admin, the workflow designer, and the translator of pipelines.
It creates messy data
Configurability changes what your data “means”. If every customer can redefine stages, your product ends up with “semantic drift”, ruining reporting and AI signals.
The shadow-system problem
When the official system is too rigid or too painful, people route around it. That's why spreadsheets never die, and “shadow AI” is growing rapidly.
Customise too much, and you make every future change harder. We've seen spectacular real-world versions of this pain, like Lidl cancelling an SAP rollout after reportedly sinking around €500 million into the effort.
And when systems fail to be usable without massive configuration, the failure modes can be horrific. In its internal report on the 2012 CIO trading losses, JPMorgan Chase & Co. describes discovering an operational error where a spreadsheet divided by a sum instead of an average—an error that muted volatility.
“Spreadsheets. In risk modelling. In a major bank. That doesn't mean 'never use spreadsheets'. It means: humans will use whatever lets them get the job done, governance be damned.”
What opinionated SaaS looks like in practice
The product has a point of view on what “good” looks like—and it nudges (or forces) you there.
Sometimes, that “forcing function” is explicit. For example, a Workday guide from a major university states plainly that Workday does not allow customisation. It's not because Workday dislikes customers—it's because the product is designed around operating in a world of frequent updates, controls, audit needs, and large-scale consistency.
- Opinionated SaaS reduces decision fatigue.
- It standardises data meaning.
- It makes onboarding faster because there are fewer branches.
- It supports automation/AI more reliably because workflows and states are predictable.
If you're selling into enterprise, customers will still ask about frameworks and certifications like SOC 2 and ISO/IEC 27001. But governance doesn't automatically come from buying a “flexible” platform. Governance comes from having an opinionated workflow that produces consistent records, clear audit trails, defaults that prevent risky behaviour, and limiting the ways people can “do their own thing”.
A practical blueprint for building opinionated SaaS
If I were building (or re-building) a SaaS product today, this is the approach I'd take—not as theory, but as an operational strategy.
Start with a narrow promise: Opinionated SaaS only works when you're honest about the customer: Who is it for? What job are they hiring it to do? What does “good” look like?
Replace options with defaults and constraints: Configuration says “tell us how you want it.” Opinionated design says: “here's how it works; tell us where reality breaks.” Have fewer custom fields and fewer workflow branches.
Keep one escape hatch, but make it boring: Fully locked-down products fail when reality is messy. So you need an escape hatch—like API-driven extensions or templated variations—without letting customers rewrite the meaning of your database.
My quick test for feature requests:
When a customer asks, “can you make this configurable?”, I ask: Is this a legal/regulatory requirement? Is this a true competitive differentiator? Will it fragment the data model? If it's not regulatory, not differentiating, and corrupts the data model, it's a “no”.
For example, on Feedsion, the core value is that the product chooses what matters (sources, structure, evidence) rather than forcing the user to build their own research workflow from scratch. The opinionated part is the point.
The unfair advantage: services first, then productise the playbook
You don't earn the right to be opinionated just by having convictions. You earn it by being right. Repeatedly.
Why “Product-led Services” work:
- Services teaches you the playbook by giving you real exposure to variability and edge cases.
- Software scales the playbook.
- Opinionated software is the packaging of earned judgement.
“We've seen this movie 200 times. Here's what works. Here's what doesn't. And our product won't let you step on the same rakes.” Are you selling knobs… or are you selling judgement?