The most common mistake: choosing the path before understanding the problem
Many mid-size companies arrive at the AI adoption decision with the wrong question. Instead of asking which process they want to improve and what outcome they expect, they ask: do we buy a SaaS, hire someone, or do it ourselves?
The order matters. The right option depends on the process, the available team, and how long the company can wait before seeing results. Without that prior clarity, any path can become spending without return.
This article describes the three paths with their real conditions for success, their risks, and a practical criterion for deciding.
Path 1: Internal build — when it makes sense and when it doesn't
Building internally means that someone on the team—or a technical profile hired for that purpose—develops the solution using AI tools available on the market.
When it makes sense:
- The process is highly specific to the business and no SaaS covers it.
- The company has, or can hire, a technical profile with dedicated time.
- The implementation horizon can extend to 3–6 months without operational impact.
The real risk:
In companies of 50 to 300 people, the technical team—when one exists—is occupied maintaining what already works. Assigning an AI project on top of their current responsibilities produces slow results, unmaintained code, and solutions that no one ends up using.
The cost is not just the technical profile's salary. It includes the time of the business owners who participate in the design, the opportunity cost of months without results, and the risk of building something that never gets adopted.
Cost hypothesis: A mid-level technical profile in Spain costs between 40,000 and 55,000 € per year. If that person dedicates 40% of their time to an AI project over 6 months, the direct cost is 8,000 to 11,000 €, not counting tools, infrastructure, or the time of department heads who must validate each iteration.
Path 2: Buying an AI SaaS — speed in exchange for flexibility
The market for SaaS with built-in AI has grown significantly. There are tools to automate reporting, manage documents, handle internal queries, or analyze sales data. Many are accessible and quick to implement.
When it makes sense:
- The process is standard and the SaaS covers 80% of the use case without customization.
- The team can adopt the tool without external technical support.
- The process volume justifies the monthly license cost.
The real risk:
The problem is rarely the tool itself. It's that the process you want to automate has exceptions, integrations with other systems, or business logic the SaaS doesn't account for. The result is a tool that works for the generic case but that the team ends up using halfway, keeping the manual process running in parallel.
Another risk: SaaS proliferation without governance. Each department buys its own tool, data doesn't connect, and the CFO loses visibility into what is being used, what it costs, and what it produces.
Cost hypothesis: An AI SaaS for reporting automation can cost between 200 and 800 € per month per company. If the process requires integration with the ERP and the SaaS doesn't offer it natively, the additional development cost can exceed the expected savings in the first 12 months.
Path 3: Hiring a consultant or specialized partner — when it's the most efficient option
An external partner brings the methodology, implementation experience, and the ability to integrate the solution with existing systems. They don't replace the internal team: they work alongside it.
When it makes sense:
- The process has a direct impact on costs, errors, or time to financial close.
- There is no internal capacity to build and operate the solution.
- Results are needed within 6 to 10 weeks, not 6 months.
The real risk:
The primary risk is hiring a consultant who delivers a prototype that no one operates afterward. This happens when the engagement ends with the technical delivery and does not include a genuine handoff to the team.
A partner who works well leaves the team capable of operating and extending the solution. One who works poorly leaves a permanent dependency or an abandoned system.
Concrete example: An industrial manufacturing company with 120 employees had a monthly close process that required consolidating production, purchasing, and logistics data from three different sources. The administration team dedicated between 25 and 35 hours per month to that process, with frequent errors in reconciliation.
With an AI agent that automated data extraction, transformation, and consolidation, the process was reduced to under 2 hours. The team validated the result instead of building it. By the third month, the same team had extended the agent to include automatic alerts on budget variances.
The project cost was recovered before the fourth month of operation.
How to decide: three concrete questions
Before choosing a path, answer these three questions:
1. Does someone have real, available time to build and maintain this?
If the answer is no, an internal build is not viable within the timeframe you need. This is not a permanent limitation—it's a current condition.
2. Is the process you want to automate standard, or does it have logic specific to your business?
If it's standard, a SaaS may be sufficient. If it has exceptions, integrations with your ERP, or proprietary logic, you need something built for your specific case.
3. How soon do you need to see results?
If the deadline is under 3 months, an internal build rarely delivers. A SaaS can be fast if it fits. A specialized partner can deliver in 6 to 10 weeks if the process is well defined.
What this is not: a technology decision
The choice between build, SaaS, or consultant is not a technical decision. It's a business decision that depends on three variables: team availability, process specificity, and expected return timeline.
The companies that achieve real results with AI are not the ones that choose the most sophisticated option. They are the ones that choose the right option for their current situation and operate it well.
If you're unsure which applies to your case, the first step is not to hire anything. It's to run a diagnostic of the specific process you want to improve, with concrete evaluation criteria.
Want to know which of the three paths applies to your situation?
Request a free diagnostic. No introductory call required, no commitment. Describe the process you want to improve and you will receive a concrete assessment with the option that makes the most sense for your company.
[Request a free diagnostic →]
