Tech trends: which skills does your project need?

How do tech trends translate into your priorities?

High demand for a technology does not automatically make it your next priority. Start with a problem you can observe: requests waiting to be handled, conflicting information, unclear costs or actions that carry too much risk. Then identify what prevents your team from resolving it. That gap determines the work to assign and the experience to look for.

The World Economic Forum’s Future of Jobs Report 2025 highlights expected growth in the importance of AI and data skills, networks and cybersecurity, and technological literacy. It presents surveyed employers’ expectations for 2025–2030. These international perspectives measure neither French demand for freelancers nor activity from WheelOfWork assignments.

This guide offers a decision framework for your business, not a technology ranking. The situations below are planning examples, not client results. You may need to examine just one area: not every issue calls for a new hire or a new solution.

This guide offers a decision framework, not a technology ranking.

AI agents: which actions should they be allowed to take?

An AI agent uses a model to decide on actions and call tools within a defined scope. Once it can act on an application, the questions change: you must define permissions, approvals and stopping conditions. A convincing demonstration is not enough to decide how much autonomy the system should have in day-to-day use.

Consider a tool that drafts replies to customer requests. Reading a knowledge base, sending a message and changing an order are three different permissions. Separate them first, then test errors and recovery before expanding the scope.

ANSSI, France’s national cybersecurity agency, recommends limiting interconnections to operational needs and retaining human control over critical interactions. Planning should involve AI integration specialists, application owners and security teams.

Look for experience of integration, evaluation and deployment that matches the work. Our Data & AI expertise helps clarify these contributions. The title “AI expert” alone does not tell you which activities someone has performed.

Process automation: who handles the exceptions?

Automation connects steps and applications to execute a process. Before selecting a tool, describe both the standard workflow and what should happen when data is missing, an application stops responding or an approval is refused. The skills you look for should cover these exceptions and recovery, as well as getting the initial demonstration to work.

To transfer a request from a CRM to a tracking tool, you need to know which system is authoritative, how to prevent duplicates and who handles failures. A business owner must explain the rules; the specialist translates them into integrations, checks and operating instructions.

A visual tool does not remove these decisions. For example, n8n can require human approval for selected AI agent tool calls. This is a capability to configure for the use case, not a guarantee that the entire process is safe.

Depending on the dependencies, you may need a developer experienced with APIs, an automation specialist or someone to coordinate business and technical work. Our Product & IT project expertise distinguishes delivery from coordination.

Data: build the use case or fix its foundations?

A data or AI use case depends on sources the team can locate, understand and keep up to date. If two departments define the same measure differently, improving the model will not resolve their disagreement. First establish whether the missing contribution concerns data pipelines, business rules, analysis or integration into an application.

An internal assistant answers from an outdated procedure. Should you change the model, remove the document, revise retrieval or clarify who maintains the content? The first task is to isolate the cause. The same symptom can call for very different skills.

A data engineer can work on pipelines and their checks; an analyst focuses on measures and their interpretation. When developing AI systems using personal data, CNIL guidance calls for clarity on purpose, responsibilities and the data needed.

The same symptom can call for very different skills.

Once you have identified the work, use our guide to writing a data and AI project brief. It covers access, deliverables and the evidence to discuss with prospective specialists.

Cybersecurity: what contribution does the project need?

Security work should follow the access rights, data and applications actually involved. Opening a tool to new users may require an access review; adding an integration may change the risks. Describe the change and the expected outcome before asking for a general cybersecurity profile. Assessment, remediation and ongoing operations are different kinds of work.

For a document assistant, ask one simple question: could a user receive information they would not be allowed to access directly? Include external documents, service accounts, logging and incident reporting in the discussion. These points help define the assessment scope.

ANSSI addresses system security across its lifecycle, including data, components and connections. Its guidance does not replace analysis of your own environment or work on personal data protection. The two types of expertise can complement each other.

Our guide to choosing a cybersecurity expert for your project distinguishes the contributions to look for. Our Cybersecurity page sets out the relevant areas and qualification questions.

Cloud and AI costs: what are you actually measuring?

A total bill shows spending, not the value of a use case. Connect technical costs to an identifiable activity, then examine its volume and quality. For an assistant, cost per request handled gives a different view from cost per token. Compare situations with equivalent scopes before deciding what to optimise.

The FinOps Foundation describes this unit economics approach. You need a baseline and ongoing monitoring; selecting a metric does not guarantee savings.

Consider internal support: lower cost per query may be of little use if more answers need reworking. Bring together the people responsible for the service, infrastructure and budget. Before looking for a specialist, establish whether the gap concerns cost allocation, architecture or product decisions.

Selecting a metric does not guarantee savings.

Which contribution should you look for first?

Start with the skills needed to remove the next obstacle. An uncertain project may need an exploratory phase with a clear decision at the end; an existing service may need reliability and recovery work. Also define what your team retains: priorities, approvals and accountability for the use case do not disappear when a consultant joins.

Use the table as a discussion aid. Job titles and responsibilities vary across organisations. Ask what the professional personally delivered, which dependencies were involved and how they handed over the work.

If several teams need coordination, explore Product & IT projects. If the work is already defined, describe the deliverable and missing skills in your assignment brief. Already selected a consultant? Our contractor management service addresses the collaboration, contracts and invoicing.