The Infrastructure Era Edition
For the past few years, most conversations about AI focused on the tools themselves.
This week reinforced a different reality.
AI is increasingly becoming infrastructure.
It's being embedded into research, software development, business systems, customer interactions, and everyday workflows. The questions is becoming less about whether organizations will use AI and more about how they will integrate it into the way work gets done.
For accounting professionals, that shift matters.
The future won't belong to firms that simply experiment with AI. It will belong to firms that learn how to use it intentionally.
- Brianne
What Happened This Week
Microsoft Redesigns Copilot Around End-to-End Work
- Why it matters: Microsoft introduced a redesigned Copilot with Home, Code, and Autopilot. Its Cowork capability is intended for work users delegate from beginning to end, and Microsoft specifically lists a financial close package as an example. The announcement also says Office in Copilot can create or update editable Word documents, Excel workbooks, and Powerpoint presentations. For accounting professionals, this is one of the clearest signs yet that mainstream AI tools are moving toward complete deliverables rather than isolated drafts and summaries.
- What to do: Choose one recurring deliverable, such as a monthly reporting package, and document its data sources, preparation steps, review points, and final approver. That will make it easier to evaluate where AI can safely participate.
- Source/Further Reading: Microsoft
OpenAI Releases Two Lower-Cost Reasoning Models
- Why it matters: Lower-cost reasoning models could make it more practical for firms and accounting-technology providers to test AI across higher-volume workflows. OpenAI released both Sol and Luna on September 22. From our standpoint, Luna may be relevant for routine, repeatable tasks, while Sol is positioned for more demanding work.
- What to do: Separate potential AI use cases into two categories: high-volume, lower complexit work & lower-volume work requiring deeper reasoning. Then evaluate these new models based on task accuracy and review requirements.
- Source/Further Reading: OpenAI
OpenAI Adds Controls for Connected Applications
- Why it matters: On September 24, OpenAI introduced external access controls for ChatGPT Business and Enterprise administrators. The controls allow administrators to manage whether ChatGPT Sites can use members' connected applications and wetherh applications can access ChatGPT ads. For firms connecting AI to email, files, or business applications, access governance is becoming more granular. An AI tool's risk changes when the tool can retrieve information from other systems.
- What to do: Review every application currently connected to your firm's AI tools. Confirm that access is necessary, limited to the right users, and consistent with your data-handling policies.
- Source/Further Reading: OpenAI
One Action This Week
Choose one recurring accounting deliverable, such as a monthly close package, client briefing, or management report.
Create a simple three-column outline:
- AI can prepare
- A professional must review
- A professional must decide
AI is no longer confined to a separate chat window.
It is moving into workbooks, documents, connected applications, and end-to-end business processes.
That creates an opportunity to rethink how work is completed. But it also creates a responsibility to decide which steps can be delegated, which conclusions must be verified, and which decisions should remain firmly human.
The value of the accountant will not come from competing with AI on speed.
It will come from knowing how to use that speed without compromising accuracy, context, or trust.
- Brianne Smith Hart
Responses