know.2nth.aiCEO Briefing
AI & data strategy · The Microsoft stack

Your AI already has a home. It's called Dataverse.

Corporate South Africa is being pitched “AI data platforms” by the dozen — while the most governed, POPIA-friendly, AI-ready one in the country is already sitting inside the Microsoft 365 licences you pay for every month. It's called Dataverse, it's the store your Microsoft apps and Copilot agents already run inside, and most executives have never heard its name. Here's what it is, why it's a South African advantage, and where it bites.

Source · Microsoft Learn + the know.2nth.ai tech leaves 6-min read For decision-makers
01 · The platform under your stack

You're probably about to buy something you already own.

Strip the jargon and Dataverse is a single, governed filing system that sits underneath the Microsoft tools your business already runs — Microsoft 365, the Power Platform (Power Apps and Power Automate), Dynamics 365, and Copilot. It isn't a database you buy and bolt on. It's the place your Microsoft apps — and any AI agent you build with them — keep their data, with the security and audit built in.

What you're being sold

A new AI data platform

A separate product to license, stand up, secure, and govern from scratch — another vendor, another bill, another integration project before any value lands.

Cost and risk you may not need to take on.

What you already have

Dataverse, inside Microsoft 365

Included with the Power Platform and Dynamics licences most corporates already hold. The governance, identity, and audit are already wired in — you configure it, you don't build it.

Use what you pay for before you buy more.

Why it matters to you: the first question isn't “which AI platform should we buy?” It's “are we already paying for one?” For most Microsoft-shop corporates in South Africa, the answer is yes — and the cost of not knowing is a second platform bought to do a job the first one already does.

02 · Why it's a South African advantage

Two things you'd otherwise pay a consultant to build, included.

For a regulated SA business, Dataverse quietly answers two of the hardest questions on the compliance checklist — and it answers them the same way for every app and agent that sits on it.

1 · Keep the data in the country

You can pin your Dataverse environment to the South African region, so the data your apps and agents work on stays in country — the plain answer to the POPIA question a board keeps asking. It's the same in-country footing the rest of the Microsoft stack gives you; for the full residency picture see the Keep the Data Home briefing.

2 · Governance you inherit, not build

Who can see what (down to a single field), a complete audit trail of who changed what, and data-loss controls all come with the platform. For banking, insurance, healthcare, and public-sector work, that inherited governance is most of a compliance answer you'd otherwise pay a team to assemble — and because it's one platform, it's consistent across every app and agent instead of reinvented each time.

03 · How it sits under Copilot

Your Copilot agent isn't floating in the cloud. It lives here.

When you build a custom agent in Copilot Studio, it isn't a magic app somewhere on the internet — it's software that lives in Dataverse. That single fact drives the two things a leader most needs to understand about agent risk.

The upside — it inherits your controls

Because the agent lives in Dataverse, it acts as a user and sees only what that person's permissions allow, and every action it takes is audited. You don't build a separate security system for “the AI” — it inherits the one you already run. An agent is also only ever as good as the governed data it can reach: clean, owned data in Dataverse is what stops it guessing.

The catch — it needs to be treated like software

“It's software” cuts both ways. An agent shipped by clicking Publish in your live system has no version history and no undo. The discipline that fixes this is ALM — application lifecycle management: build and test the agent in a separate environment, promote it to production in a controlled way, and keep the ability to roll back. Plain version: don't edit your AI live, ship it like software. The full engineering treatment is the Copilot Studio leaf.

04 · The honest trade-offs

Where it's the wrong tool — say it before you build on it.

Dataverse buys inherited governance by giving up two things. Neither is a dealbreaker; both should be priced in before you make it the foundation of everything.

Lock-in is real.

Your data, its rules, and your agents become Microsoft-shaped. That's the cost of the convenience — name it and decide it deliberately, don't back into it.

It's priced by capacity.

Dataverse charges by storage capacity, not per record. For high-volume, plain data plumbing a normal database (Azure SQL, PostgreSQL) can be much cheaper — use the right tool for that job.

It's not for analytics.

Heavy reporting and number-crunching belong in a reporting/warehouse tool. Dataverse runs your operations; it doesn't replace your BI stack.

Not for a few simple lists.

If it's a handful of small lists, a spreadsheet or SharePoint is cheaper and simpler than a governed platform you won't use.

05 · The decision

Four questions for your exec and your IT lead.

Are we about to pay twice?

Before buying an AI or data platform, have we checked whether our Microsoft 365 / Power Platform licences already include Dataverse for the same job?

Is our environment pinned to South Africa?

Is the region set so customer data stays in country — or did we leave it on a default that quietly sends it offshore?

Do our agents inherit our permissions?

Does the Copilot agent see only what the user is allowed to see, with every action audited — or has someone given it a back door around our controls?

Do we ship agents like software?

Do we build and test in a separate environment and promote to production with a rollback — or are we editing a live agent and hoping?

The honest default

For a mid-to-large SA corporate already on Microsoft 365, the move isn't a new platform — it's to use the one you own, pin it to South Africa, ground your Copilot agents in it, and ship those agents with the same test-then-release discipline as any other software. Don't buy a second platform for a job M365 already does; do price the lock-in honestly before you make Dataverse the floor under your whole operation.

Keep reading

Related

The take

The AI-ready, POPIA-friendly data platform corporate South Africa keeps being sold is, for most Microsoft shops, already paid for. Use what you own, keep it in the country, and ship your agents like the software they are.