Caricamento...
Three approaches to bringing AI into your professional firm: custom consulting, DIY and vertical SaaS. An honest comparison of costs, risks and when to choose which.
Every professional firm that wants to adopt AI faces the same question. Not "whether to use AI" — that question is already settled — but "how to structure adoption". And the options on the table are essentially three: do it yourself, buy a vertical SaaS, or engage a specialist consultant.
There is no universally correct answer. There is a correct answer for the size of your firm, the level of regulatory risk you manage, and the resources you can dedicate. This article helps you identify it.
What it means in practice: the principal or a colleague starts using freely available AI tools — ChatGPT, Gemini, Copilot — for daily activities. No structured training, no written policy, no formal governance. Learning by doing.
Who is suited to DIY: firms of 1-3 people, with low regulatory-risk work, where the principal is directly involved in every output and can exercise direct oversight over every use. Also firms that want to experiment before committing to a more structured approach.
The real risks of DIY:
DIY is not wrong as a starting point. It is problematic as an end state.
What they offer: vertical software for the legal and professional sector that integrates AI into existing features. Tools such as Harvey, Kira, or AI modules in accounting management software fall into this category.
The advantages:
The limitations:
What it means in practice: a specialist consultant — external or fractional — designs and implements AI adoption tailored to your firm. The work includes: initial assessment, tool selection, governance (policy, training), phased implementation, and measurement of results.
The advantages:
The limitations:
Who it is suited to: firms with 5+ people, or firms of any size that handle particularly sensitive data or high-risk work (corporate restructuring, tax disputes, M&A files).
| Criterion | DIY | Vertical SaaS | Custom consulting |
|---|---|---|---|
| Initial cost | Low | Medium | High |
| Implementation time | Immediate | 1-4 weeks | 1-3 months |
| Customisation | High (but ungoverned) | Low | High |
| AI Act compliance | Partial | Partial | Complete |
| Governance risk | High | Medium | Low |
| ROI measurement | None | Limited | Structured |
| Suited to | 1-3 people, low risk | Firms wanting low friction | Structured firms or high-risk work |
Choose DIY if: you are a sole practitioner, you use AI only for low-risk activities (regulatory research, generic internal drafts), and you have the time to experiment without immediate regulatory pressures. But treat DIY as a phase, not a permanent arrangement.
Choose a vertical SaaS if: your firm already uses an established management system that is integrating AI, the required compliance level is low to medium, and you want the fastest adoption with the least organisational effort.
Choose custom consulting if: you have more than 5 colleagues, you handle particularly sensitive data, you have already been subject to inspections or disputes, or you want to build a real competitive advantage — not just "use something AI".
The most reliable signal: if you already have a colleague using AI and you do not know exactly with which data and for which activities, you are already in territory that requires structured governance.
There is a common misconception about AI consulting: that it creates dependency on the consultant. The model we practise is the opposite.
Our objective is for your firm to have the competencies to govern AI independently at the end of the project: written policy, trained team, selected and configured tools, active monitoring metrics. The consultant must become superfluous — that is the signal that they have done their job well.
The Fractional AI Officer figure — a part-time external AI responsible — serves precisely for the transition phase: in the first 6-12 months, when governance is being built and questions are frequent. Then the knowledge stays internal.
Not sure which approach is right for your firm? A free 30-minute assessment clarifies the current situation and indicates the most efficient path — with no commitment.
If you want to explore the structured methodology we use once the approach is decided, the AIRA Method explains how we move from assessment to measurable results in 90 days. To get a sense of the full range of services — from compliance to ongoing strategic support — browse what we offer. Or if you prefer to talk through your specific situation first, contact us.
It depends on the SaaS. Some management software for accountants or law firms has integrated AI with adequate GDPR clauses, but rarely covers AI literacy obligations (documented staff training), an internal AI policy, and client disclosure under Art. 50. In other words: a SaaS solves the AI use problem, not the compliance problem.
It depends on how it is done. ChatGPT Plus used for low-risk activities (regulatory research, internal drafts, brainstorming) with already anonymised data is acceptable. But using it to process client data (accounts, declarations, documents) without verifying the OpenAI contractual terms, without an internal policy and without team training — is a GDPR and AI Act risk that many firms underestimate.
Three questions to ask any consultant: (1) Have you worked with firms like mine (same sector, same size)? (2) How do you measure the results of your work? (3) What do you leave me at the end of the project so I can continue independently? A good proposal always includes an initial assessment, concrete deliverables (policy, training, roadmap) and a plan for the client's autonomy.
Ogni settimana: guide pratiche, novità normative e casi d'uso per studi professionali. Niente spam.
Most AI implementations in SMEs do not reach their stated objectives. Here are the 3 mistakes we see repeated most often in professional firms, and how to avoid them.