An AI consultant helps you work out where AI genuinely fits your business, then gets it working in production alongside the tools and teams you already run. In practice that means three things: starting from the decisions and tasks that actually cost you time or money, choosing the smallest technology that solves the problem, and being honest about the places AI does not belong. A good AI consultant is measured on what ships and gets used, not on the demo they showed you in the first meeting.
That is the plain answer. The honest version is more interesting, because the title "AI consultant" now covers a very wide range of behaviour, from people who will meaningfully change how your business runs to people who will sell you a strategy deck and disappear.
What a good AI consultant does
Good AI consulting starts with your business, not with the technology. Before anyone mentions a model, a competent consultant wants to understand how your people actually spend their week, which decisions are slow or expensive, and where the same information-heavy work gets done over and over. The AI comes second. It is the answer to a question about your business, and that question has to be asked first.
From there the work looks fairly unglamorous, which is a good sign:
- They scope from a real task or decision. Not "let's add AI", but "your team spends eight hours a week pulling quotes together from three systems, and here is what it would take to cut that in half."
- They favour the smallest thing that works. The goal is a working result in weeks, not a platform in a year. A narrow, boring solution that ships beats an ambitious one that stalls.
- They build for production, not for the pitch. A demo runs once, on clean data, in front of an audience. Production runs every day, on messy data, when you are not watching. Those are different engineering problems, and most of the value lives in the gap between them.
- They tell you where AI does not fit. This is the clearest signal of a consultant worth paying. If everything you mention is met with "yes, AI can do that", you are talking to a salesperson. Some of your problems are process problems, or data problems, or people problems, and no model will fix them.
- They leave you more capable, not more dependent. Good work hands over documentation, sensible defaults, and enough understanding that your team is not helpless the moment the consultant leaves.
If you want the longer version of this argument, we have written about why the useful work is starting from the decision, not the data, and what AI enablement actually means for a normal business.
What bad AI consulting looks like
The failure modes are just as recognisable once you know the shape of them.
- Demo theatre. An impressive live demonstration that quietly relies on hand-picked inputs and a great deal of preparation. It is designed to produce a feeling, not a result. The question that pops the bubble is simple: "can we run this on last month's real data, right now?"
- Strategy decks with no floor under them. Forty slides on your "AI transformation journey" and no one who can actually build the thing. Strategy that never touches production is expensive wallpaper.
- Vendor lock-in dressed as architecture. Solutions bolted so tightly to one platform, or to the consultant themselves, that leaving becomes painful by design. Independence should be a feature, not something you have to negotiate later.
- Solutions in search of a problem. Starting from a shiny capability and working backwards to find somewhere in your business to put it. This is how you end up with an AI feature that technically works and that nobody uses.
The common thread is that bad consulting optimises for the sale and the impression, while good consulting optimises for what still works in six months. One is measured in slides, the other in tasks that are genuinely faster, cheaper, or less error-prone than they were.
So what are you actually paying for?
Underneath the title, you are paying for judgement: the ability to look at your business and tell you which problems are worth solving with AI, which are not, and what the smallest sensible first step is. The building matters too, but the building is increasingly the cheap part. The expensive, valuable part is knowing where to point it and being straight with you about the rest.
If you are weighing up whether to bring someone in, it is worth reading how the engagement is usually priced (see what AI consulting costs in South Africa) and how to choose an AI consulting firm without getting sold to. You can also see how we frame the work on our AI consulting page.
FAQ
What does an AI consultant do? An AI consultant helps a business identify where AI genuinely fits, chooses the simplest technology that solves the problem, and gets it working in production alongside existing systems. Good consultants start from your tasks and decisions rather than the technology, and are honest about where AI does not belong.
What is the difference between a good and a bad AI consultant? A good AI consultant starts from your business problems, builds for everyday production use, and tells you where AI does not fit. A bad one relies on polished demos, strategy decks with no delivery behind them, and solutions that lock you into one vendor or into the consultant themselves.
Do I need an AI consultant, or can my own team do it? If your team already has the time and the production engineering experience, they can often make a start unaided. A consultant earns their fee mainly through judgement: knowing which problems are worth solving, which are not, and how to reach a working result quickly without over-building.
How is AI consulting different from AI integration? Consulting is largely about deciding what to do and why, whereas AI integration is the work of connecting AI to the systems, data, and workflows you already run. In smaller engagements the same team usually does both.
What should an AI consultant leave behind? A working solution in production, sensible documentation, and enough understanding for your team to operate and adjust it without being dependent on the consultant. Work that leaves you more capable is the goal; work that leaves you more dependent is a warning sign.
Written by JP, Sixees Labs. Last reviewed July 2026.