AI in SMBs fails when no one is responsible — the RACI matrix clarifies everything in 1 hour
In 64% of SMBs that have launched AI projects, no one has a formal mandate to steer the AI strategy. The result: orphan projects, duplicate tools, and an inability to scale. Yet structuring AI governance doesn't require creating an entire department. A simple RACI matrix (Responsible, Accountable, Consulted, Informed) and 3 to 4 key roles are enough to effectively steer the AI transformation of an SMB of 50 to 500 employees.
The problem
The question "who steers AI?" is the most frequent and least well-resolved question in French SMBs and mid-market companies. AI disrupts the traditional boundaries between IT and business, which creates organizational gray areas.
The dysfunctions we most frequently observe:
- Orphan AI: no one has an official mandate to coordinate AI initiatives. Each department experiments in its corner, creating duplicates and inconsistencies. The marketing team uses ChatGPT, accounting tests an OCR tool, HR experiments with a chatbot — without any coordination.
- The CIO / Business conflict: IT wants to control AI tools for security and architecture reasons, business teams want to move fast and find IT too slow. This conflict paralyzes projects and frustrates everyone.
- The absent CEO: executive management treats AI as a technical topic delegated to the CIO, whereas AI choices are strategic choices that impact the business model, the customer relationship and the organization of work.
- The lack of AI skills: even with the best intentions, SMBs lack profiles who can scope an AI project, evaluate a provider, or steer a deployment. And hiring a data scientist is out of reach for most.
Without clear governance, AI initiatives remain fragmented and don't create lasting value. The challenge is not to create a bureaucracy but to clarify who does what with minimal structure.
The AI solution
Our AI governance approach for SMBs rests on three lightweight, pragmatic pillars, designed to be deployed without hiring or restructuring.
The 4 key roles of AI governance
Strategic sponsor (CEO or executive committee member): vision, budget, trade-offs. AI lead (CIO, digital project manager or tech-savvy manager): project coordination, monitoring, provider management. Business champions (1 per department): use case identification, field adoption, surfacing needs. Compliance guarantor (DPO or legal counsel): GDPR, AI Act, usage policy. In SMBs, these roles are combined with existing functions.
RACI matrix of AI activities
A matrix that clarifies who is Responsible, Accountable, Consulted or Informed for each key AI activity: use case selection, choice of tools and providers, project scoping and budget, development and integration, user training, production monitoring, regulatory compliance. This matrix is built in a 1-hour workshop and eliminates 90% of scope conflicts.
Monthly AI committee
A lightweight governance body that meets 1 hour per month. Composition: strategic sponsor, AI lead, 2-3 business champions, CIO. Agenda: review of ongoing projects, validation of new initiatives, budget, monitoring. This committee is the single AI decision point in the company. Our training helps structure this committee.
Implementation
Here is the three-step plan to structure your AI governance in less than 2 weeks, without an external consultant or reorganization.
Appoint the AI lead and the champions (week 1)
The CEO designates an AI lead among existing profiles: the CIO (if the SMB has one), a digital manager, or a tech-savvy manager. Selection criteria: knowledge of the company, technological curiosity, ability to work cross-functionally. In parallel, identify 1 AI champion per key department (sales, production, finance, HR). These people will devote 10-20% of their time to this role. Formalize these appointments with an email from the CEO to provide the necessary legitimacy.
Build the RACI matrix (weeks 1-2)
Bring together the appointed people for a 2-hour workshop. List the 8-10 key AI activities of your company and assign the RACI roles for each. Important rules: only one "A" (Accountable) person per activity, the sponsor approves budgets > 10,000 euros, business teams are always "R" (Responsible) for use case identification, IT is "R" for security and integration. Share the matrix across the whole company.
Launch the first AI committee (week 2)
Schedule the first monthly AI committee with a concrete agenda: status review of current AI usage (15 min), presentation of 2-3 priority use cases identified by the champions (30 min), Go/No-Go decision and budget allocation (15 min). After the first committee, send minutes to the whole company to show that AI is a steered and structured topic. Regularity matters more than perfection.
Results
Frequently asked questions
Do you need to hire a CDO (Chief Data Officer) to steer AI in an SMB?
Not necessarily for an SMB of fewer than 250 employees. A part-time AI lead (20-40% of their time) is enough for the early stages. This role can be entrusted to the CIO, a tech-savvy business manager or a digital project manager. Hiring a dedicated CDO is justified from 5+ simultaneous AI projects or an annual AI budget above 200,000 euros.
How do you involve business teams in AI governance?
Three effective levers: appoint an AI champion in each department (an ambassador who surfaces use cases and facilitates adoption), include business managers in the bi-weekly project reviews, and entrust the sponsorship of AI projects to business directors rather than to IT. AI must be perceived as a business tool, not as an IT project.
What is the role of executive management in an AI project?
Executive management has three key responsibilities: define the strategic AI vision and embed it in the business plan, allocate the necessary budget and resources, and lead by example by using AI tools themselves. A CEO who uses an AI assistant to prepare their management committees sends a strong signal to the whole organization.
How do you organize an AI committee in an SMB?
An effective AI committee in an SMB meets once a month for 1 hour. It brings together the CEO (or a member of the executive committee), the AI lead, 2-3 business sponsors and the CIO. The typical agenda: review of ongoing projects (30 min), validation of new projects (15 min), technology and regulatory monitoring (15 min). No need for more formalism at the start.
For technical profiles
Comparison of AI organizational models by company size and maturity:
| Criterion | Decentralized model (SMB < 100) | Hub-and-spoke model (SMB 100-500) | AI center of excellence (mid-market > 500) |
|---|---|---|---|
| Dedicated AI team | No (shared roles) | 1-2 people (lead + data) | 3-8 people (dedicated team) |
| Annual AI budget | < 50,000 euros | 50,000 – 300,000 euros | > 300,000 euros |
| Number of simultaneous projects | 1-2 | 3-6 | 6+ |
| Governance | Lightweight monthly committee | Committee + formalized RACI | Complete governance + KPIs |
| Internal skills | Lead + providers | Data analyst + developer | Data scientists + ML Engineers |
| Agility | Maximal | High | Moderate (process) |
| Main risk | Dependence on a single lead | Hub / business tension | Disconnection from the field |
Our recommendation: start with the decentralized model with an AI lead and business champions. Evolve toward hub-and-spoke when you exceed 3 simultaneous AI projects. The center of excellence is only relevant from 500 employees and a substantial AI budget. The classic mistake is to want to create a center of excellence too early — it creates bureaucracy without added value.