Guide · Build vs buy
Build your own tools before buying more licences
Ironhack for Business
- Buy a SaaS licence
- Build with Claude
Claude for business usually enters an organisation as one more licence: it gets bought, handed out, and everyone uses it their own way. Spain records Claude usage 2.69 times higher than its working-age population would suggest (Anthropic, Anthropic Economic Index, May 2026), and in September 2026 Anthropic opened its first office in the country. The value many companies still do not measure is somewhere else: using Claude to build the internal tools they currently pay a subscription for.
What is Claude for business, and what do teams use it for?
Claude for business is the use of Claude, Anthropic's AI assistant, with accounts managed by your organisation and applied to work tasks: writing, analysing, researching and programming. In Spain, 42.28% of Claude use is work-related, and the most frequent topics are content creation (22.06%), education (14.7%) and software development (11.34%) (Anthropic, Anthropic Economic Index, May 2026).
If you are choosing a plan or comparing tools, we cover that in our Claude vs ChatGPT guide: plans, data handling and the AI Act. This article is about a use few companies plan for: building.
Can you use Claude for business to build internal tools?
Yes: with Claude and Claude Code, people who are not developers can create prototypes and working tools for their team without waiting for capacity in engineering. It already happens in Spain: 4.37% of conversations with Claude produce an app or a website (Anthropic, Anthropic Economic Index, May 2026).
This is already changing purchasing decisions. In McKinsey's State of AI survey, 32% of respondents say their organisation has decided not to buy one or more software products or features because it could build them internally with agentic coding tools (McKinsey, The State of AI in 2026: 1,719 respondents across 97 countries, May to June 2026).
Anthropic documents the same pattern in its own non-technical teams. Its Growth Marketing team built a flow that reviews CSV files of all their ads, spots the worst performers and generates new variations: hundreds of ads in minutes instead of hours. Its legal team prototyped a tool that routes each query to the right lawyer (Anthropic, How Anthropic teams use Claude Code, Jul 2025).
When is it better to build with Claude than to buy a SaaS?
Building makes sense when the task is repetitive and specific to your company, you are paying for seats or features you do not use, and what you need fits into a few functions, with someone responsible for maintaining them.
The cost of not reviewing this is real. Across the organisations Zylo analyses, an average of 36% of SaaS licences go unused, and spending on AI-native applications grew 108% year on year (Zylo, 2026 SaaS Management Index).
Review these 4 criteria before renewing:
A repetitive, company-specific task
Your team does it every week in a way particular to your company, so a generic tool makes you adapt to it instead.
Underused licences
You pay per seat and only part of the team has access, or you use a fraction of the features.
A bounded scope
What you need fits into a few functions and one clear flow.
A clear owner
Someone on the team maintains the tool and decides what changes.
In our two cases all 4 held: repetitive tasks, per-seat licences and a scope that fitted into a week of work. To estimate how many hours your team recovers by automating a task, use our savings calculator. The result is an indicative estimate, not a guaranteed saving.
When is it better to keep buying software?
Keep buying when the tool is critical to the operation, carries regulatory requirements, certifications or SLAs, needs to integrate with many systems, or nobody in your company can maintain it.
Building with Claude lowers the cost of creating a tool, not of maintaining it. An internal tool needs someone to fix errors, update what stops working and respond when it fails, and it has to pass the same security and data controls as any software you buy. Without that owner, a tool built in a week becomes one more case of shadow AI.
Keep buying in these cases:
- Critical systems, such as billing, payroll or your main CRM, where a failure stops the operation.
- Software subject to regulatory requirements, certifications or SLAs your team cannot take on.
- Tools that depend on integrations with many systems.
- Any tool without an internal owner.
4 criteria to decide before renewing a licence
Build with Claude if…
- The task is repetitive and specific to your company
- You pay for seats or features you do not use
- The scope fits into a few functions and one clear flow
- Someone is responsible for maintaining it
Keep buying if…
- The tool is critical to the operation
- It carries regulatory requirements, certifications or SLAs
- It needs integrations with many systems
- Nobody in your company can maintain it
Decide go or no-go with numbers: the hours the task takes before and after, and the cost of building and maintaining against the subscription.
Ironhack for Business
What does it really cost to build a tool with Claude?
Building costs the hours of whoever builds it, the Claude licences your company already pays for and, above all, the maintenance hours every year. That total is what you compare against the annual cost of the subscription you are replacing.
Put the two options side by side:
| Buy (SaaS) | Build with Claude | |
|---|---|---|
| Direct cost | Annual licence per number of seats | Build hours at the loaded hourly cost |
| Recurring cost | Renewals and price increases | Maintenance hours per year |
| Hidden cost | Seats and features you do not use | Security and data review |
| Dependency | Your team depends on the vendor | Your team depends on whoever maintains the tool |
A practical rule: if building and maintaining costs less than the subscription over the first year and the tool covers the task, it is a go. If not, renew and look again at the next renewal. To measure the task's impact in hours, before and after, follow the method in our guide on training ROI.
Where to start: from licences to your first own tool
Start with one tool, not with a transformation programme. These 3 steps take you from the licence list to a decision backed by data:
- Take an inventory of licences and tasks. Cross what you pay each year against the tasks that eat the most hours in each team. To do it across the whole organisation, our AI consulting delivers a complete task-by-task automation map, 3 to 5 prioritised quick wins with estimated hours and euros saved, and a business case built from your own numbers, in a Lite (3 weeks) or full (6 to 8 weeks) format.
- Train the team on a real case. The first prototype comes out better when the team practises on its own process. In our Claude workshops, Claude is the main tool and each team takes away a replicable demo agent built on its own use case. It is a demo: deploying it into your systems is a separate step, and your company supplies the Claude licences.
- Decide go or no-go before scaling. Trial the tool with real users for a few weeks, measure the hours the task takes before and after, and decide whether it replaces the subscription, lives alongside it, or gets dropped.
Next step
Pick 1 subscription your company renews in the next 3 months and check whether it meets the 4 criteria for building. If you want to do it with a method and across the whole organisation, we will come back with a starting point within 1 working day.