Back to home
Frequently asked questions
What people ask us before we start
The questions raised in a first meeting, with the answers we actually give.
- Do we need to be "ready" for AI first?
- No. A ready organisation is one that has already done the work — mapped its processes, tidied its data, identified its bottlenecks. That work is exactly what we do first. What you need at the start is not technical maturity: it is one specific task that costs someone time, every week.
- How much does an artificial intelligence project cost?
- We publish no price list, because a process audit, a team training session and the implementation of a connected agent share neither duration nor deliverable. Cost depends on three things: how many processes are involved, the state of the data to connect, and the level of autonomy you are aiming for. A one-hour scoping call is enough to establish an order of magnitude before any commitment.
- How long before we see a result?
- Team training has an effect within the following week: people stop using AI as a search engine. An assistant connected to the organisation's documents usually takes a few weeks, most of the delay coming from data access rather than development. End-to-end automation is deployed in stages, starting with the most expensive process.
- Will our data leave the organisation?
- That is an architecture decision, not a given. Depending on how sensitive the data is, a solution can run on hosted models, on models operated in a closed environment, or by separating what leaves from what stays. We document that choice before building: you must know which data goes where, and be able to justify it to your board or your donor.
- Will AI replace our team?
- Our work is to remove tasks, not people. The processes we automate are the ones nobody claims: re-entry, sorting, chasing, formatting, consolidating reports. An organisation that deploys AI to cut headcount loses the only people able to supervise the systems it has just installed.
- What is the difference between using ChatGPT and what you install?
- A public assistant answers from what it knows in general. It knows neither your procedures, nor your contracts, nor your customer history, and it cannot execute anything in your tools. We distinguish three levels: chatting, connecting — the AI reaches your data and your systems — and delegating, where agents carry out complete tasks under human supervision. Value appears from the second level onwards.
- Does this work with limited internet connectivity?
- Yes, provided it is designed for from the start. The solutions we build are sized for irregular bandwidth: batched rather than continuous processing, lightweight interfaces, tolerance to outages, and lossless resumption when the connection returns. In the DRC this is a structural constraint, not an edge case.
- Which tools and models do you use?
- We are agnostic. Depending on the problem, a solution may rely on models from OpenAI, Anthropic, Google, on hosted open models, or on a combination of all three. The choice is justified by the need — confidentiality, cost per request, language, latency — and documented. A provider who proposes the same tool to every client is selling their partnership, not your outcome.
- Do you train teams too, or only executives?
- Both, and separately. An executive needs to know how to decide: where to invest, which risks to accept, how to measure. An operational team needs to know how to do it, on their own files. The same material served to both audiences works for neither. RAPIA Academy covers both formats.
- Do you work outside Goma?
- Yes. We are based in Goma, Kinshasa and Lubumbashi, and we work remotely with organisations elsewhere in the DRC and across the continent. Scoping and training phases benefit from being held on site; the rest is delivered remotely with no loss of quality.
Your question isn't here?
Ask it directly. We answer precise questions with a precise answer, not with a brochure.