AI director, consultant or agency: who should you bring in?
Three ways to bring AI into a mid-sized company, compared on lead time, price certainty, who runs it and knowledge of the CEO's job.
Thibault Pons de Vincent
Founder of Oliverlist and VivaSync
You have decided to move on AI this year. The next question is who with. Three options keep coming up: hire an AI director, bring in a consultant, or hand a project to an agency. Each has its own logic. Each can also fail, for a different reason.
I have run Oliverlist since 2016. I watched automation, then AI, enter the product and become a leadership topic rather than a technical one. Here is how I compare the three options, using the criteria that matter to a CEO.
Three options, three different jobs
The AI director. Sometimes called head of AI or chief AI officer, this is a person whose job is AI: choosing the use cases, building or commissioning the tools, running them, measuring them. The role can be a full-time hire or a fractional one, shared with other companies. Either way, the person stays after the launch.
The consultant. A consultant studies your organisation, spots the possible uses, sizes them and ranks them. The deliverable is usually an assessment and a roadmap. Some stay on for the implementation; others stop at the plan. The large firms tend to work for big groups and government entities, so a company of a few hundred people will often deal with independents and boutique firms.
The agency. An agency builds: a chatbot, an automation, a connection between your systems. It works from a specification, delivers, then moves on to the next client. Maintenance is a separate contract.
These three jobs answer different questions. The consultant answers "what should we do?". The agency answers "how do we build it?". The AI director answers "who looks after it, week after week?".
The comparison, criterion by criterion
| Criterion | AI director | Consultant | Agency |
|---|---|---|---|
| Time to a first result | Long for a full-time hire: search, notice period, visa, onboarding. Shorter for a fractional role. | A few weeks for the plan. A working tool depends on who builds it next. | A few weeks for a well-scoped tool, longer if the brief keeps moving. |
| Price known upfront | Yes: a salary package or a monthly retainer. The outcome is not. | Often billed by the day: the total depends on how many days. | Often a fixed price for the project. Changes and maintenance come on top. |
| Who runs it day to day | The AI director: that is the core of the role. | Your team, once the consultant has left. | You, or the agency under a maintenance contract. |
| Knowledge of the CEO's job | Varies: a technical profile and a former operator do not do the same job. | Good on strategy, rarer on day-to-day operations. | Usually limited: the agency knows the tools, not your trade-offs. |
These cells describe tendencies, not rules. A consultant who has run a company is worth more than an AI director who has never read a P&L. An agency that stays with you after delivery is worth more than a polished roadmap left in a drawer.
What each option does well, and where it stalls
The consultant helps when you do not know where to start. An outside view sees what habit hides. The risk is a well-presented roadmap that nobody executes, for lack of time or skills in-house. A plan does not run itself.
The agency works well when the need is precise and stable: connecting your CRM to invoicing, sorting a shared inbox, producing quotations from a template. Agencies selling AI agents and chatbots are easy to find in the region; ask to see a tool that has been running at a client for six months, not a demo. The risk appears after delivery. Your data changes, your processes move, and the tool degrades without anyone noticing. The agency knows how to build; deciding what to build is your job.
The AI director is the right answer when AI becomes a permanent topic, with several projects a year. The risk: a rare and expensive profile that produces prototypes if left isolated, without access to data or weight with the leadership team. A fractional role costs less, but you have to frame availability and priorities.
One more point: whoever you choose must handle every language your business works in, in your documents and in your customers' messages. Test it on your own material, not on a demo.
What the three have in common: the hard part is rarely technical. It lies in knowing your business, in the state of your data and in keeping someone in charge over time.
Choosing for your situation
Before you choose who, be clear on what. I recommend you automate the predictable first: part of the work does not need AI at all, and turning it into an AI project means paying a lot for what a simple rule does for almost nothing.
Then start from where you are.
- You do not know where to start. A short outside review, provided it ends with a plan that has names, dates and a first task you can launch straight away.
- You know exactly what you want. An agency, with a precise brief and, from day one, the name of the person who will run the tool after delivery.
- AI has to become a function of the company. An AI director, full-time if your size justifies it, fractional otherwise. Seat them close to the leadership team, not at the far end of IT.
Whatever you choose, keep three things in-house: the accounts and subscriptions in your company's name, access to your data, and documentation of what has been built. If the provider leaves, the tool must keep running.
My conviction after nearly ten years at the head of Oliverlist: a tool without an owner eventually stops. The right choice is the one that leaves you with someone to run it.
Key takeaways
- The consultant delivers a plan, the agency delivers a tool, the AI director keeps it running: three different jobs.
- The deciding question is often forgotten: who will run the tool on Monday morning, once the provider has gone?
- A price known upfront does not guarantee an outcome. Agree from the start on what will be measured.
- With equal technical skills, choose the one who understands the CEO's job.
- Whatever the choice, accounts, data and documentation stay in your name.