Announcements about the implementation of Microsoft Copilot are everywhere, and many companies are considering activating the tool across the entire organization, possibly even purchasing the necessary licenses. However, the decisive factor for success is not the tool itself, but the quality of the data it will be directed towards.
When the data foundation is not properly prepared, problems arise: answers that cannot be fully trusted; low adoption rate of the tool; and lack of expected productivity. Thus, readiness to work with Copilot is a data problem, not just a matter of having licenses.
Copilot does not possess internal knowledge of the company's business; it analyzes the resources available to it—files, mailboxes, databases—and inherits existing access rights. If these permissions are vague, the system may display content that was not intended for the reader. Furthermore, outdated or duplicate data causes Copilot to answer based on any suitable version, with a confidence that the database itself does not warrant.
The vendor approach prioritizes readiness
Microsoft shares this view. In its Copilot implementation plan, the company prioritizes addressing excessive access issues before installing safeguards and launching. Independent studies on the implementation of generative AI in large enterprises reach a similar conclusion: if results are disappointing, the cause is usually found in the conditions of working with the model, in data governance, and in the tool's alignment with workflows, rather than in the model itself.
In South Africa, it is also a compliance issue
Under Popia legislation, the organization remains the responsible party for how personal data is used and accessed. This responsibility does not transfer to the service provider, much less to artificial intelligence. An assistant capable of extracting and disseminating personal information across all resources in seconds only increases the company's obligations, rather than easing them.
Readiness is a program, not just a switch
When facing this problem, two common instincts arise: either trying to quickly tidy everything up—fixing a few permissions, deleting old files, and moving on; or, conversely, freezing all processes until the data is flawless. Both approaches are ineffective.
Copilot begins to provide real value long before the entire infrastructure is perfect, so the goal must be narrower. Data must be managed for each use case before that scenario starts operating, and the readiness process must be synchronized with the launch process. This requires targeted work: understanding what data is stored and where it is located, modernizing the platform where it resides, defining access rights, and cleaning the data so that the information extracted by the AI is current, correct, and authorized.
Microsoft has released temporary solutions, the best known of which will soon disappear. Limited SharePoint search—a system toggle that held content away from Copilot while permissions were being fixed—stopped accepting new activations on July 31, 2026, and will be completely retired on January 31, 2027. Its replacement is Limited Content Discovery, which works by sites, not across the entire organization, and Microsoft will not migrate existing settings. Those who used the old control as a delay tactic now have a deadline, not a postponement.
The sequence of actions is as important as the steps themselves. The platform must be modernized so that data has a reliable home. Access must be optimized and controlled so that only the right people—and the right AI—can access it. The infrastructure must be secured so that Copilot reads current, authorized, and compliant information. When performed in the correct order, this work accumulates effect. When done piecemeal, the gaps that the AI will discover first remain.
How Ascent ensures readiness
Ascent's data platform modernization program makes an organization ready for AI. It begins with a structured assessment of the entire data array—including the platform, integration, governance, and access—transforming a chaotic process into a clear action plan. This assessment is the first part clients notice: it provides a clear picture of where the vulnerability lies, what needs fixing, and in what sequence, allowing work to be prioritized based on risk and value, not guesswork.
Next, the company modernizes the core itself, unifying disparate data, integrating governance and security directly into the platform instead of adding them later, and establishing checkpoints that define what the AI is allowed to read. This core is built on Microsoft Fabric. Thanks to Fabric's OneLake, the organization gains a single, governed data array—its own 'AI-ready data foundation' from Microsoft, where access, lineage, and quality are set once and applied everywhere the data is used, including Copilot. This is the difference between AI based on a reliable source and AI guessing from scattered, uncurated data.
Since Ascent is a Microsoft Direct cloud solutions provider, Copilot and Azure licensing can be handled by one partner who builds this foundation. This ensures a single point of accountability for both the license and the readiness, instead of buying the tool in one place and leaving the foundation unattended elsewhere. By doing this correctly, Copilot stops being a gamble. The business gains a trustworthy assistant, an infrastructure that can be defended to the board of directors and the Information Regulator, and an advantage that grows while competitors are still cleaning up. Modernize, optimize, secure, and then activate.