Everybody wants AI. We help you use it safely.
Your staff are probably already using it, with or without a policy. The risk isn’t AI itself, it’s using it with no rules, on personal accounts, with nobody able to see what went where. We help you get real value from it, on the systems you already have.
How we help
You don’t need a big AI strategy to start. Most businesses begin at step 1 and move forward as confidence grows.
Get the rules right
Before anything else: who may use AI, for what, and with which information. We help you set your risk appetite, classify your data, write an acceptable-use policy people will actually read, and decide who approves what.
Roll it out safely
Company accounts with your existing Microsoft sign-in, access by department, training, and a small pilot before everyone gets it. Personal and free AI accounts get replaced by something you can see and control.
Connect it to your systems
This is where AI becomes genuinely useful: asking questions of your own ERP, files and reports. We build a controlled gateway between AI and your data, read-only by default, with permissions by role and a log of every request.
Build what’s specific to you
Some problems need custom AI: computer vision on a production line, or reading long technical documents into structured data. We build these, then host and support them.
The security around it
This is the part most AI conversations skip. It’s also the part we’re best placed to do, because we already run the servers, Microsoft 365 and networks your AI will depend on.
Company accounts only
No personal or free AI accounts for company information. Everyone signs in with their work account, so access ends when they leave.
Same access as today
Nobody sees more through AI than they could already see. This is enforced where the data lives, not by telling the AI to behave.
Read-only by default
AI can look things up. It can’t create, change, send or post unless you approve that specific exception.
Every request logged
Who asked, what they asked for, when, and whether it was allowed or blocked.
Clean up before you connect
AI search finds files nobody knew were shared too widely. We audit and fix sharing before it’s switched on.
Protection against hidden instructions
Emails and documents can hide text telling AI what to do. Read-only connections and restricted agent features limit what that can achieve.
A person stays accountable
AI drafts and suggests. A named person checks the result and owns the decision, especially anything that leaves the business.
POPIA considered from the start
Operator agreements, cross-border processing, impact assessments and keeping personal information to the minimum.
AI we’ve already built
Not experiments. These are systems we’re building and running.
Computer vision on the production line
A camera system counts what leaves the line and records it automatically, with dashboards against the day’s plan and a full audit trail on every correction. Read the case studies.
Reading aircraft manuals
Fusion Aero Desk extracts maintenance schedules from aircraft manuals into structured data: 81 pages became 109 maintenance items, human-reviewed before anything is approved.
Common questions
Will AI train on our company data?
On free and personal accounts, it may. Business plans from the major providers generally commit not to, but terms differ and change, so we check the current terms of the plan you’re on.
Do we need to replace our systems?
No. The point is to make the systems you already have more useful. AI sits alongside them.
Where should we start?
Find out who’s already using AI and what for. Then agree the rules, and pilot it with a small group on low-risk work.
Is AI going to get things wrong?
Yes, sometimes, and confidently. That’s why anything that matters gets checked by a person against the source.
Using AI safely in your business
A short, practical working paper for owners and executives in South African companies. Eight pages, written to be used.
- The AI risks that are actually new
- What information must never go into AI
- Who should approve what
- Connecting AI to your systems safely
- POPIA considerations and a phased rollout
- A ten-point readiness check
Thank you. Your copy is ready.
If you’d like to talk it through, book a free AI conversation below.
Download the working paper (PDF)Start with a free AI conversation
45 minutes, on site or on Teams. Tell us where you are with AI and we’ll tell you honestly what’s worth doing first. Our working paper is a good place to start.
Book a free AI conversationOr call 087 265 0624 · hello@fusiongroup.co.za