South African companies whose own teams lead their AI projects are the most likely to say those projects beat expectations. In DayOne's State of AI Delivery ZA 26 report, 54% of practitioners at companies where their own team leads AI delivery said results beat expectations.
"Expertise can be bought. What cannot be outsourced is internal adoption, leadership ambition and trust," DayOne founder Liesel Bester writes in the report's introduction.
Insights on in-house AI
Most of the 282 practitioners surveyed said their own team leads AI work: 52.5% are in-house, 41.5% work in a hybrid model where their team works alongside outside partners, and 6% rely mainly on a vendor or consultant. Here is how results compare.
Who delivers AI | Beat expectations | Met expectations | Fell short |
|---|---|---|---|
Own team | 54% | 28% | 18% |
Own team with partners | 49% | 36% | 15% |
Mainly a vendor or consultant | 18% | 41% | 41% |
Teams that work with partners but stay involved do almost as well as fully in-house teams, and are slightly less likely to fall short.
The weak results come from the few companies that hand delivery over almost entirely. Only about 17 respondents fall into that group, so the finding is a signal rather than a firm rule, and the report notes that its comparisons show links, not causes.
Leadership matters too. 64% of respondents see a gap between what their leaders want from AI and the resources they provide, and 27% call that gap significant. Where ambition and resources line up, 61% say AI beat expectations.
Where the gap is significant, 36% say AI fell short, about 6 times the rate at companies where the two are aligned. The survey ran from 2 to 16 October, and 66% of respondents work at companies of 50 people or fewer.
What others are saying about in-house AI
DayOne, an AI build studio and network of AI professionals, summarises its wider findings as "adoption is solved, delivery isn't". Elsewhere in the report, only 2% of respondents said vendor or consultant claims shaped their expectations of AI, and 14% said their AI budget covers external partners or vendors.
You can buy expertise, not ownership
The lesson here is not "don't use outside help". Companies that bring in partners while keeping their own people closely involved get results almost as good as fully in-house teams.
What seems to go wrong is handing AI over and stepping back. That matches Bester's point: a partner can bring skills, but the people who will use the tools every day, and the leaders who decide what success looks like, cannot be replaced.
For a business planning its next AI project, the practical takeaway is to keep someone inside the company responsible for it, even if a partner does much of the building, and make sure the budget and time match what leaders expect.
As our pieces on AI training budgets and AI and job security found, the same report keeps pointing to people, not technology, as the deciding factor.
You might also like our piece on Claude vs ChatGPT in SA, AI agent swarms in Minecraft, and SA's AI policy.
Get more SA tech and business news and subscribe to The Open Letter.


