Every week we speak to businesses asking the same question: should we just use ChatGPT, try Claude, or build something custom? The answer, unhelpfully, is "it depends." But it depends on specific, identifiable factors that we can work through together.
Having helped dozens of organisations navigate this decision, here is the framework we use.
When Off-the-Shelf ChatGPT or Claude Is Enough
Let's start with the simplest option: giving your team access to ChatGPT Team, Claude for Work, or a similar commercial AI product.
This works well when your needs are general-purpose — drafting emails, summarising documents, brainstorming, writing code, or answering questions that do not require proprietary knowledge. For many businesses, this is genuinely the right answer. Deploying an off-the-shelf tool takes hours, not weeks, and the cost is predictable (typically £15-25 per user per month).
ChatGPT excels at general knowledge tasks, code generation, and creative writing. It has the broadest plugin and integration ecosystem. Claude tends to perform better on long document analysis, nuanced reasoning, and tasks requiring careful instruction following. It also has a stronger data privacy posture, with Anthropic explicitly not training on business inputs by default.
The limitations of off-the-shelf tools become apparent when you need the AI to know about your specific data, follow your specific processes, or integrate into your existing workflows. A ChatGPT subscription cannot search your internal knowledge base, follow your compliance procedures, or connect to your CRM without significant additional work.
For a detailed feature comparison, see our Claude vs GPT comparison.
When to Build Custom
Custom AI solutions — purpose-built systems using APIs from OpenAI, Anthropic, or open-source models — make sense in several scenarios:
- Proprietary data integration: You need AI that answers questions based on your internal documents, customer data, or domain-specific knowledge. This typically means building a RAG pipeline or fine-tuning a model.
- Workflow automation: You want AI embedded in a specific business process, not as a standalone chat interface. For example, automatically classifying incoming support tickets, extracting data from invoices, or generating personalised client reports.
- Quality control: The stakes are high enough that you need guardrails, validation, human-in-the-loop review, and audit trails. Off-the-shelf tools do not offer this level of control.
- Data sensitivity: Your data cannot leave your infrastructure, or you need to demonstrate to regulators exactly how data is processed. Custom solutions let you control the entire data flow, including using local models that never send data externally.
Custom solutions cost more upfront (typically £10K-50K for an initial build, depending on complexity) but can deliver substantially higher ROI because they solve your specific problem rather than providing generic capabilities.
The Hybrid Approach
In practice, most of our clients end up with a hybrid approach, and this is usually the right answer.
The pattern looks like this: deploy ChatGPT Team or Claude for Work for general-purpose use across the organisation. Then build custom AI solutions for the two or three workflows where the impact is highest. The general-purpose tools handle ad-hoc needs while the custom solutions deliver measurable automation for specific processes.
A common hybrid architecture uses Claude or GPT APIs as the underlying model, wrapped in custom orchestration that handles data retrieval, prompt management, output validation, and integration with existing systems. You get the power of frontier models without building everything from scratch.
This approach also lets you manage costs effectively. General-purpose tools are a fixed per-seat cost. Custom solutions can be optimised — using smaller, cheaper models for simple tasks and reserving expensive frontier models for complex reasoning. We regularly see clients reduce their API costs by 60-70% through intelligent model routing.
Learn more about integrating commercial AI into your workflows on our cloud AI integration page.
Cost Comparison
The total cost of each approach varies significantly:
- Off-the-shelf (ChatGPT/Claude): £15-25 per user per month. For a 50-person team, that is £9K-15K per year. Predictable, low risk, but limited customisation.
- Custom solution: £10K-50K initial build, plus £500-3K per month for API costs and maintenance. Higher upfront cost, but the per-unit cost of processing drops significantly at scale, and the ROI from automating a specific workflow often pays for itself within months.
- Hybrid: Combines both cost structures. The general-purpose tools serve broad needs cheaply, while custom solutions target high-value workflows. This is typically the most cost-effective approach overall.
The key insight is that the cheapest option per seat is not always the cheapest option per unit of value. A custom solution that automates 30 hours of weekly manual work delivers far more value than a per-seat subscription, even if the absolute cost is higher.
Control vs Convenience: Making the Decision
The decision framework is ultimately about where you sit on the control-versus-convenience spectrum.
If your primary concern is getting AI into your team's hands quickly with minimal effort, off-the-shelf tools win. If your primary concern is solving a specific business problem with measurable ROI and you need control over the data flow, custom wins. If you want both — and most businesses do — the hybrid approach gives you the best of each.
Questions to ask yourself:
- Does the AI need access to our proprietary data?
- Is there a specific workflow we want to automate end-to-end?
- Do we have regulatory requirements around data handling?
- Can we quantify the value of automation for our top use case?
- Do we have (or can we hire) the technical capability to maintain a custom solution?
If you answered yes to three or more of these, a custom or hybrid approach will likely deliver better outcomes than off-the-shelf tools alone.
Not sure which approach is right for your business? We help organisations navigate this decision every week. Book a free strategy call and we'll give you an honest recommendation based on your specific situation.