Bespoke vs SaaS is no longer the old trade-off
For a long time, the enterprise software decision looked simple. Want speed, predictability and a familiar operating model? Buy SaaS. Want something tailored to how your business actually works? Commission a bespoke build and accept the trade-offs: more time, cost, complexity and governance overhead. That framing is now changing.
I'm writing this now for three reasons. First, the market signals are becoming too consistent to ignore. Second, the underlying technology has shifted in a way that makes modular, governed bespoke delivery far more practical than it used to be. Third, our own experience at Gysho has shown this is no longer a theoretical debate. It's an operating reality. This doesn't mean every organisation should rush to build everything itself. It means the old assumption (that SaaS equals speed and bespoke equals pain) is no longer a reliable way to think about digital enablement. The real question now is more precise:
What should be standardised, and what should be differentiated?
01 · The market is changing the terms of the debate
The strongest reason to revisit the bespoke-versus-SaaS question isn't opinion. It's the direction of travel coming from the analyst market. Taken together, Gartner, Forrester and IDC point to the same shift: enterprise software is becoming more agentic, more modular and more process-centric. The relevant signals:
- Gartner (Aug 2025): 40% of enterprise applications will feature task-specific AI agents by 2026, up from less than 5% in 2025.
- Forrester: enterprise software is shifting from user-centric design to worker- and process-centric design.
- Forrester: 30% of enterprise app vendors will launch their own MCP servers.
- Forrester: half of enterprise ERP vendors will launch autonomous governance modules combining explainable AI, automated audit trails and real-time compliance monitoring.
- IDC: by 2029, 30% of global IT services will be delivered as modular, platform-enabled products.
- IDC: by 2029, 30% of contractual engagements with service providers will be outcome-based, driven by agentic AI.
Those statements matter for three reasons.
Software is moving closer to the workflow
The old software model was screen-based and user-based; the new model is increasingly process-based. That matters because many businesses don't win through generic screens or features. They win through the way they structure work, make decisions, apply judgement and govern outcomes. When software becomes process-centric, it's easier to design around a real operating model than to force the operating model to fit the application.
Modularity changes what "bespoke" means
If a meaningful share of IT services is moving towards modular, platform-enabled delivery, bespoke no longer has to mean building everything from zero. It can mean assembling proven building blocks into something tailored, governed and commercially useful: a very different proposition from the traditional image of a long custom build programme.
Governance is becoming part of the product
One historic objection to bespoke systems has been governance risk. That concern doesn't disappear, but the market shows governance becoming a core design requirement across enterprise software, not a separate compliance afterthought.
The market is not moving away from governance. It is moving towards governance by design.
That changes the quality of the conversation. The question is no longer just whether a system is custom or packaged. It's whether it can operate safely, transparently and effectively inside the real business environment.
02 · Where SaaS stopped working for us
This isn't just an external market story. Some of Gysho's technology exists because off-the-shelf AI and software platforms couldn't do what we needed in practice: a critical proof point of how this works, both functionally and as a feasibility study.
HubSpot is a great starting platform, but it hits a pricing cliff when stretched, and Breeze wasn't reading its own data. Microsoft Project and SharePoint required specialists to operate and wouldn't pick up a plain email. Off-the-shelf AI assistants indexed file names instead of contents. All of these are traditionally good SaaS solutions, but once we started pushing our own ambitions to perfect our processes, they no longer fit. The inflexibility meant we couldn't adapt their operation to our way of working, and agents had limited success.
Standard software works well until the workflow matters more than the tool.
That was the point at which configuration stopped being enough. So we built differently, a composable platform: a modular stack of 25 reusable backend services and application shells that can be assembled into unique client applications in days rather than months. That platform model matters because it changes what bespoke can look like: the application is tailored to the workflow, while the underlying infrastructure stays reusable, governed and scalable.
A practical example: why we replaced HubSpot
One of the clearest examples is our own CRM, built as a replacement for HubSpot in roughly three days. It includes automatic transcript ingestion, meeting summarisation, action-item extraction, follow-up email drafting, sentiment analysis, merged stakeholder briefs and a live decisions-and-risks log. It replaces HubSpot's core CRM behaviour, but is configured precisely to Gysho's preferred way of working. We value efficiency and want AI to flag important actions to us, instead of having to log every interaction and go into the CRM to work out what to do next.
We didn't replace HubSpot because bespoke is inherently better. We replaced it because the off-the-shelf route wasn't supporting the way we actually worked, and an AI-first, agentic-ready system was more useful in practice. We now run our commercial operations, internal project management and customer intelligence on the same platform we sell to clients. In other words, we eat our own cooking. SocialOS is another example: our social-selling and commercial-signal intelligence application, built because HubSpot and its peers couldn't offer that capability at any price.
This is our lived operational example, not a claim that every company should replicate it exactly. But it demonstrates something useful: when the workflow is strategically important, and the market tools cannot support it properly, a bespoke route no longer has to be slow, expensive or impractical.
03 · Why bespoke is becoming easier, and more valuable
This is where the discussion gets more commercially interesting. The case for bespoke isn't that everything should become custom. It's that in a commoditised market, firms need to be much more deliberate about what they standardise and what they protect as distinctive. Three reasons bespoke is becoming both easier to achieve and more strategically useful.
The plumbing can now be reused
Modern bespoke doesn't need to start from a blank page. Reusable components now handle a large share of the foundational work (orchestration, connectors, workflow controls, security patterns, monitoring and deployment support) so the bespoke layer can focus on what actually creates value: proprietary frameworks, decision logic, review workflows, governance rules, firm-specific data structures and client-facing operating flows. Custom at the value layer; reusable at the infrastructure layer. That's why bespoke is easier to achieve now than it used to be.
Differentiation is moving away from generic tooling
In many sectors, the base technology is becoming more available and more similar. That's not a problem for end users. In fact it helps, since technologies are more likely to integrate and there's more choice, making them a commodity that should be treated as such. The strategic question is whether a firm's distinctive expertise is being flattened into generic software in the process, and that matters especially for consulting, advisory and research firms, whose value is often built on proprietary frameworks, structured methods, specialist taxonomies, repeatable decision processes, experience-based judgement and governed review models. If those assets stay trapped in documents, spreadsheets, project teams and institutional memory, they're harder to scale, protect and apply consistently. A bespoke product can operationalise that IP without forcing it into a generic system that weakens it.
That's the logic behind our own offer: we help consulting, advisory and research firms embed differentiated IP into secure, governed agentic products: starting with a rapid, outcome-driven 5–7 day proof of concept and scaling through subscription-based continuous improvement.
A better balance of trust, governance and agility
One of the biggest misconceptions is that bespoke and governance pull in opposite directions. They don't have to. A well-designed bespoke model can create stronger control over where human review is required, how outputs are explained, which sources are authoritative, how audit trails are maintained, how confidentiality boundaries are enforced, and how the workflow connects to the wider business. Governance is part of the operational design of a system, which matters even more in an agentic environment.
Do not customise commodity. Differentiate where your business actually wins.
This is also why "build or buy" is increasingly the wrong question. The better questions are: what is truly commodity in our operating model? What is truly differentiating? And how do we govern both properly? Those lead to better decisions than a default assumption that buying packaged software is always safer, faster or more commercially rational. Sometimes it is; sometimes it isn't.
Conclusion: standardise what's common, build where it matters
The old bespoke-versus-SaaS debate was built for a market that looked different. Today, applications are becoming more agentic, services more modular, governance embedded, and the real competitive advantage is moving away from generic tooling towards the way firms structure work, apply expertise and turn knowledge into repeatable outcomes. That's why the decision needs reframing: not a binary choice between custom and packaged software, but a more disciplined question of standardise versus differentiate.
From our own experience, the lesson is straightforward: off-the-shelf tools are useful until they stop fitting the workflow, the governance model or the intelligence the business actually needs. When that happens, a composable bespoke approach can now be a credible option in a way it simply wasn't for many firms a few years ago. That doesn't mean customising everything. It means being clear about three things: what is common, what is core, and what must be governed.
Standardise what is common. Protect what is core. Govern both properly.