Software Media and Entertainment
MultiChoice
One HR portal and one login for staff across MultiChoice, DStv, M-Net and SuperSport.
Build AI into your products and operations.
We build AI features and integrations into software, from assistants and search to automating repetitive work, and test them against your data before they reach users.
AI features can take on repetitive work and help people find answers faster.
Teams spend hours answering the same questions, searching for documents and reading forms. Customers wait for answers that already exist somewhere in the business.
AI built into your software can handle much of that work, using your own information. The value depends on doing it carefully: grounded in your data, tested before launch, and checked by a person where a wrong answer would matter.
Where AI features tend to earn their place
- Answering routine questions from your own documents, for customers or staff.
- Search that understands what people mean, not only the words they type.
- Reading forms, invoices and contracts, and checking what they contain.
- Handling less structured steps in an automated workflow.
- Spotting patterns in data that point to demand or risk.
- Summarising long records so people can act on them faster.
Making it dependable is most of the work.
A demonstration of an AI feature takes days. Making it dependable takes longer. Answers have to come from your own information, not from whatever the model happens to know. Access has to follow the same rules as the rest of your systems, so people only see what they are allowed to see.
We test features against real questions and real data before launch, measure accuracy and running cost, and keep a person in the loop wherever a wrong answer would be expensive. After launch we keep measuring, because both the data and the models change.
Our software work
Software Finance
Ai SPARK
Built the web application that puts Ai SPARK's bond performance prediction technology in front of its users.
Software Real Estate
Property24
A real-time Facebook app where every like added R5 towards a house, tracked by a growing fan meter.
Software Media and Entertainment
Daily Maverick
Redesigned the news site for over a million readers and supported the Maverick Insider launch, now 17,500+ contributors.
AI and machine learning FAQs
How do you stop AI giving wrong answers?
We ground answers in your own data, add checks and human review where the stakes are high, and test accuracy before launch.
What does it cost to run?
Running costs depend on usage. We estimate them during the prototype and design the feature to keep them predictable.
Can AI work with our existing systems?
Yes. Most AI features are useful only when they are connected to the systems and documents you already have.
Do we need an AI strategy first?
If you haven't decided where AI fits, start there. If the use is clear, we can go straight to a prototype.
Latest insights
Tell us what you want AI to do.
Give us a sense of the task, the data and the systems involved.