Custom Automation & AI for SMBs in Barcelona

The floor is €2,500 for small, well-scoped projects (1-2 automations, one integration). Typical SMB projects sit between €3,500 and €8,000, depending on how many flows we cover and which systems we integrate. Custom-model AI implementations or complex ERP integrations can reach €10,000-15,000.

Price is closed in the initial proposal, not by estimated hours that later grow. What drives project cost up: number of automated flows, integration complexity (documented APIs vs. legacy systems), data volume, special compliance requirements (high-risk AI Act, GDPR with sensitive data), and whether the AI model is self-hosted or consumed via API. The first diagnostic call is free and ends with a clear range before you sign anything.

The processes with the highest return are usually the ones you already do many times a month and that consume expensive people's time. The four typical candidates: invoice preparation and sending, responses to recurring queries (via email or chat), data extraction from documents (quotes, delivery notes, contracts), and internal reporting currently built by hand in Excel.

For payback to arrive quickly, it's better to start with one or two processes, not everything at once. If the team spends 10 hours a week classifying emails or writing standard replies, automating that usually recovers the investment in 2 to 4 months. Larger projects (generative-AI assistants, semantic search over private documents, decision-making agents) can be very profitable, but they require more design and aren't usually the recommended first project.

2-4 weeks for a small automation with well-documented integrations (one flow, one or two APIs). 4-8 weeks for typical SMB projects (multiple flows, integration with your ERP or accounting software). 8-16 weeks when there's custom-model AI, legacy-system integrations, or AI Act requirements demanding specific documentation.

Timelines assume your API access credentials (Holded, your ERP, Google Workspace, whatever applies) are available in week one. The usual bottleneck isn't development but access: getting API keys, admin permissions, and real test data tends to take longer than expected in companies with multiple departments. If we work in parallel from the start, timelines hold.

Yes. The Kit Digital AI category offers grants of up to €6,000 from April 2026 for AI-solution implementation. The Business Intelligence and Process Management categories offer up to €19,000 each in Segment V (100-249 employees), with a total available grant for that segment of €29,000.

The typical process: we review which categories fit your project (it can be one or several combined), I help you draft the technical documentation required for the application, and we coordinate with an accredited digital agent to submit the grant. The program is in force under Orden TDF/39/2026 until funds run out, so there is no fixed deadline. If your SMB has already used its full grant, the project is still viable; only who pays for it changes.

Because several legal obligations converge around that date and most of them require touching invoicing and process software. Verifactu is mandatory on January 1, 2027 for companies and July 1, 2027 for self-employed. B2B electronic invoicing (Ley Crea y Crece) is mandatory from July 2027 for companies with revenue over €8M, and from July 2028 for the rest. The AI Act reaches full application in August 2026.

If you're going to touch your invoicing system to comply with Verifactu or B2B, it's the natural moment to add the automations you've been putting off. Implementing two separate changes costs twice as much as implementing one well-planned change. For companies already thinking about AI, the AI Act marks the difference between implementing with compliance from the start (cheaper) and having to redo audits retroactively (much more expensive).

It depends on the risk classification of the specific use case. Most SMB implementations fall under limited risk (chatbots, internal assistants, summaries) or minimal risk (automatic classification, internal search), with manageable obligations: user transparency, human oversight on relevant decisions, basic activity logging. Only uses in sensitive sectors (healthcare, hiring, credit) fall under high risk, with impact-assessment requirements.

What changes in August 2026: obligations for AI users (not just providers) become fully applicable. That means if your company uses an AI assistant with customers, you have to be able to show there's human oversight, transparency, and basic logging. Building the solution with these requirements from the start is relatively cheap. Retrofitting them later because an audit lands is much more expensive. I design every project with the risk class identified and the corresponding controls documented.

The typical automation agency in Spain resells subscriptions to Make, n8n, or Zapier, charges a one-time implementation plus the platform's monthly fee. As a freelance developer, I write your code in your repository, with known operating costs (servers and APIs you pay directly) and no functionality ceiling when you reach complex flows.

There are cases where the agency is the right answer: teams without an internal technical profile, simple flows that no-code platforms cover well, and low volume where monthly subscriptions don't scale. There are others where custom code wins: integrations with Spanish systems (Verifactu, electronic invoicing, ERPs not friendly to no-code), flows with logic the platforms don't express well, sensitive data you'd rather not move through third-party servers, and high volume where fees stack up. The first call ends with an honest recommendation, even if it's 'agency'.

No. I always work by integrating with what you already use: Holded, Factura Directa, Anfix, FacturaScripts, Google Workspace, Notion, Slack, your in-house CRM or ERP if it has an API. Changing the base software is a separate, usually larger project and isn't justified by automation alone.

What I do evaluate in the initial diagnosis: whether your current software has a documented API that allows what you want to automate. Most modern SaaS does (Holded, Factura Directa, Google Workspace). Some older ERPs don't, and in those cases it's often more cost-effective to build a bridge or, in extreme cases, consider replacement. But that decision is yours, not the automation project's.