Start with the job, not the model
Most people pick an AI assistant the way they pick a phone: by brand, by hype, or by whatever a friend uses. That works badly, because assistants are not interchangeable. The right question is not "which model is smartest" but "what do I actually need done, how often, and with what kind of information".
Write down your top five recurring tasks before you compare anything. For a lot of people in Mauritius that list looks something like: drafting emails in English and French, summarising long documents, preparing quotes or proposals, researching suppliers or regulations, and tidying up spreadsheets. Your list will differ, but the exercise matters. An assistant that is brilliant at coding but clumsy with French correspondence is the wrong tool for a bilingual office, no matter what the benchmarks say.
The four factors that actually separate assistants
Once you know the job, evaluate candidates on four things:
- Language and tone. Test it on your real writing. If you work across English, French, and the occasional Kreol phrase in a client message, paste in genuine examples and see whether the output sounds like you or like a press release.
- Context handling. Can you give it a long contract, a full email thread, or a year of meeting notes and get sensible answers? Assistants vary enormously in how much material they can hold and how well they use it.
- Privacy posture. Read the data policy, not the marketing page. Does the free tier train on your inputs? Is there a business plan that switches training off? For anything touching client data, this is not optional.
- Where it lives. An assistant inside your email, documents, and calendar removes friction. A separate chat window means copying and pasting all day. Integration often beats raw intelligence for daily work.
Notice that "which company makes it" is not on the list. The major assistants leapfrog each other every few months. Your workflow, once built, should survive a model change.
Run a two-week trial with real work
Do not judge an assistant on a demo prompt. Pick two candidates and run them side by side on real tasks for two weeks. Keep it honest:
- Use the same prompts on both, with your actual documents (redact anything sensitive first).
- Score outputs on how much editing they needed, not on how impressive they looked.
- Note failures. An assistant that confidently invents a clause in a contract summary is more dangerous than one that says it is unsure.
Two weeks is long enough to hit the boring Tuesday tasks, which is where assistants earn their keep. Anyone can shine on "write me a poem". Fewer tools survive "reconcile these two supplier lists and flag the mismatches".
Free tier, paid tier, or business plan?
The pricing decision is simpler than it looks. Free tiers are for evaluation and light personal use. If you rely on an assistant for work more than a few times a day, the paid tier usually pays for itself in the first week through better limits and stronger models. In rupee terms the subscription is real money, so treat it like any other tool purchase: if it saves you two hours a month, it is already cheap.
Business plans matter for a different reason: data controls, admin oversight, and the ability to say honestly to clients that their information is not training someone else's model. If you are choosing for a team rather than for yourself, start there, and consider getting rollout help from a local consultancy such as Nexus (nexus.mu) rather than letting every employee improvise with personal accounts.
Decide, commit, and revisit twice a year
Analysis paralysis is the real enemy here. The gap between the top assistants is smaller than the gap between using one well and not using one at all. So decide with imperfect information, commit for six months, and build habits: saved prompts for your recurring tasks, a routine for checking outputs, a rule about what data never goes in.
Then put a reminder in your calendar to revisit the choice twice a year. The market moves fast, switching costs are low if your prompts and habits are portable, and the assistant that was clearly best in January may be second best by July. Loyalty is for people; tools have to keep earning their place.
The short version: define the job, test with real work, check the privacy fine print, pick the one that fits your day, and stop reading comparison articles. Including, after today, this one.
A well chosen assistant gives every person on your team an extra pair of hands. Explore the wider Nexus health ecosystem.
