Embracing AI at Work: HR Transformation and Productivity

Training employees in AI at work boosts productivity and reduces resistance to change. HR guide to successful adoption.

The DATALIA team · Published August 2025 · Updated August 2025

Adopting AI at work involves three key components: AI training (AI literacy), deploying custom AI assistants, and structured change management. Without these three elements, even the most powerful AI remains underutilized.

Contents

• Basics: Why training comes before the tool
• Training in AI: The foundation of AI literacy
• Deploying an AI assistant at work
• Change management and adoption
• Best practices and pitfalls to avoid

Basics: Why training comes before the tool

In 68% of companies complaining about low AI adoption, the finding is the same: employees have access to an AI assistant but don't use it. The main reason? Lack of understanding of the tool's capabilities and fear of automation.

Our field diagnosis, drawn from audits conducted at accounting firms and real estate agencies, shows that the absence of training during the first months reduces effective adoption to less than 20%. Yet, organizations that train their teams before deployment see regular AI usage reach 75% within three months.

Training is therefore not a luxury: it is the main driver of successful adoption.

Read more: The three pillars of adoption

Adopting AI at work rests on three pillars:

• AI literacy : understanding the capabilities and limitations of AI;
• AI assistants at work : tools tailored to repetitive tasks;
• Change management : guiding the transition without threatening jobs.
• These three dimensions form an integrated whole: ignoring one blocks the other two. Training without a tool leads nowhere. A tool without training remains unused. And change management without training generates resistance.
• The concept of AI literacy refers to employees' ability to understand, use, and evaluate artificial intelligence systems. It's not a technical goal: it's an organizational skill.
• The European Commission estimates that at least 60% of European employees will need ongoing retraining related to AI by 2030. Meanwhile, a McKinsey survey published in 2024 shows that only 21% of employees reported having followed AI training in 2023, despite growing demand for these skills.
• AI training should cover four areas:
  1. Understanding AI : difference between generative and analytical AI, concepts of prompt and hallucination;
  2. Using AI in daily work : writing, summarizing, translation, data analysis;
  3. Working with AI : validating results, iterating queries, documenting learning;
  4. Managing risks : confidentiality, bias, source traceability.
• Each area should be illustrated with a concrete case related to the employee's role. For example, a customer service representative learns to rephrase a complaint email with AI assistance, then verifies the legal compliance of the result.
• In-person training remains preferable for initial acquisitions. Then, peer mentoring — "learning by doing" — is more effective than a lecture. 45-minute workshops where each employee tests AI on their own job yield an assimilation rate twice as high as a lecture session.
• A model training program:
• In 71% of corporate AI assistant deployments, the tool is abandoned within six months. The cause: it's not integrated into existing workflows. An AI assistant at work must be naturally embedded in the employee's typical day to be adopted.
• At DATALIA, we've observed that adoption of a sovereign AI assistant — hosted locally, connected to internal data — increases by 300% when deployed after targeted training. The key: the assistant complements, it doesn't replace.
• In the restaurant industry, a voice assistant connected to the reservation system lets waiters ask about allergens in a dish in real-time without interrupting service. Result: a 40% reduction in back-and-forth trips between the dining room and kitchen, observed at a DATALIA client.
• In the real estate sector, an assistant incorporating solvency analysis automates buyer pre-qualification. A Franco-Belgian agency reduced its processing time from 5 days to 2 hours with this system, while keeping humans in client relationships.
• For an AI assistant at work to be adopted, it must meet three principles:
  • Contextual : it understands business vocabulary and internal processes;
  • Transparent : it indicates its sources and signals its limitations;
  • Secure : it never transmits sensitive data outside.
• These requirements are hard to meet with a general-purpose assistant. At DATALIA, our AI assistant is designed to meet these imperatives: private hosting, connection to internal ERPs like Odoo, complete traceability of exchanges.
• Change management is often the neglected pillar of AI adoption. According to a Deloitte study published in 2024, 62% of employees surveyed fear that AI may threaten their jobs. This perceived risk blocks adoption, even when the tool is perfectly integrated.
• Successful change is based on three steps:
  1. Communicating intentions : explaining that AI automates tasks, not jobs;
  2. Involving teams : co-constructing use cases with employees themselves;
  3. Measuring impact : tracking adoption and satisfaction metrics.
• In a DATALIA client accounting firm, automation of journal entry recording freed up 15 hours per month for each staff member. These hours were redirected to accounting variance analysis — a higher-value activity. The employee thus gained autonomy without their position being threatened.
• This model also applies to other sectors. In a CPTS (Center for Social Action and Senior Support), the AI assistant enabled centralizing administrative and medical data, reducing data entry errors by 60% and improving coordination between care and administration.
• p>For measuring AI adoption at work, track:
• During our missions with SME leaders and CFOs, we identified five frequent errors when deploying AI at work:
• Error: installing the assistant, then training. Result: the employee doesn't know how to give appropriate instructions.
• Fix: train in AI literacy, then deploy the tool with use case scenarios.
• Error: imposing ChatGPT or another public tool across the entire organization.
• Fix: prefer a sovereign AI assistant, hosted locally, incorporating internal data.
• Error: automating a task without rethinking the process. Automation applies routine; transformation rationalizes it.
• Fix: map the full workflow before automating, as recommended by DATALIA's VASPIS method.
• Error: allowing employees to paste sensitive documents into a public assistant.
• Fix: implement a private AI assistant, compliant with GDPR and the AI Act.
• Error: delivering the tool, then disappearing. Adoption quickly declines without follow-up.
• Fix: appoint an "AI champion" per team, implement a peer-learning program, and set up regular reporting.
• AI adoption at work is not a race to the most powerful tool, but a gradual transformation of teams. By starting with training, deploying a private AI assistant, and supporting each employee through change, you build an organization ready for the future.
• Discover how DATALIA supports organizations in their AI transformation, from consulting to training through integration of custom assistants.
• Automate your business with AI through DATALIA: DATALIA →
• No. The principle of a workplace AI assistant is to be accessible to everyone, without technical skills. It's the concrete use cases, illustrated during training, that allow each employee to adapt quickly.
• By preferring a sovereign AI assistant, hosted locally, and connected only to necessary internal data. At DATALIA, our assistant complies with GDPR and the AI Act, with complete traceability of every exchange.
• Automate your business with AI through DATALIA: DATALIA →

How to ensure data confidentiality with an AI assistant?

Do employees need technical skills to use an AI assistant?

Frequently asked questions

Next step

Key takeaways

5. No post-deployment follow-up

4. Neglecting data confidentiality

3. Confusing automation with transformation

2. Imposing a single public assistant

1. Deploying the tool before training

Best practices and pitfalls to avoid

Key performance indicators to track

The real impact of AI on employees

Change management and adoption

Assistant design: 3 requirements

Concrete integration cases

Deploying an AI assistant at work

Effective training methods

Training content: What to teach?

Training in AI: The foundation of AI literacy