For many finance teams, the real work begins when the analysis is already "done." Before a report reaches a client or a management committee, there is an exhausting last mile to cover: reconciling the evidence, assembling and formatting the document, checking every number and linking every statement to its source. The final PowerPoint or Excel has to be editable and withstand scrutiny. AI applied to financial work promises to ease precisely that part, and a recent case helps us understand how far the technology reaches today and what is realistic to expect.
What the Model ML case teaches us
According to a case published by OpenAI, the company Model ML has built AI agents that accompany a finance professional from the initial request to the final deliverable: research, analysis and a deck or an Excel workbook ready for review. At the center is an "agent" that plans the work, chooses the right tools, reconciles the evidence and runs calculations, relying —according to the source— on the GPT‑5.6 Sol model for a good part of the tasks.
The underlying idea is interesting for any company: the agent doesn’t just answer questions, it completes a workflow from start to finish and generates native PowerPoint and Excel files with traceable sources. The professional stops reconstructing the analysis and focuses on what adds judgment: reviewing assumptions, refining the message and validating before sharing.
The "last mile" of finance, in numbers
The case provides striking data, always according to Model ML’s internal evaluation: at a large asset manager, a custom tearsheet that took an analyst around an hour to assemble came to be completed in about five minutes
In its presentation-creation tests, GPT‑5.6 Sol completed the PowerPoint workflow in 100% of cases (compared to 76% for a rival model) and cleared the "quality gate" of ready-for-review in 43.3% of cases (compared to 26.7%)
What’s relevant for the reader is not the scoreboard between models, but the trend: AI is no longer limited to drafting a paragraph, but chains research, calculation and formatting through to a deliverable that a human can review directly.
Why AI applied to financial work matters in your SME
No SME needs to replicate the case of a large asset manager to benefit from this idea. In almost any company there is that invisible "last mile" that consumes hours: closing the monthly reporting, preparing the dossier for the bank or the investor, consolidating data from several sheets, reconciling figures between systems. That is where AI applied to financial work offers a clear and measurable return.
Three lessons worth taking away:
1. The value is in the use case, not the model
The headline is the model of the moment; the business is in the concrete problem it solves. A good AI project starts from a repetitive, costly and measurable task —for example, generating a recurring report— and is evaluated by results: hours saved, errors avoided, cost per deliverable. Pursuing "AI just to have AI" leads nowhere.
2. The person still decides
In the case, the professional reviews assumptions, sources and message before sharing the work. That human verification is essential: models can "hallucinate," that is, generate incorrect information with the appearance of being true. That’s why source traceability and techniques like RAG matter, anchoring the answers to the company’s real information.
3. Without organized data, AI doesn’t perform
An agent that consolidates figures needs clean, accessible and governed data. If the information lives scattered across loose sheets and emails, no model will work magic. Data governance is the foundation, not a luxury.
From headline to project: how Tisa lands it
At Tisa we have been helping companies make their technology work for the business since 1987, and AI follows that same principle: applying it through concrete and measurable use cases. Our AI and data line accompanies each company according to its maturity: we identify the relevant data, define governance protocols aligned with regulations, evaluate the tools according to the investment capacity and, above all, develop and execute a use case adapted to your business.
In addition, as a Microsoft partner, we start from a solid base for many SMEs: Business Central integrates financial and accounting management, and the Microsoft ecosystem —Power BI for dashboards, Power Automate to automate processes and Copilot as an assistant— makes it possible to add AI on top of data that is already organized within the ERP. It’s the most direct path for a company that already works with Office and Dynamics, without setting up a parallel infrastructure.
The practical approach would be this: instead of buying "the latest AI," we identify together which report, which reconciliation or which process is eating up your hours every month, and we build a use case there with human verification and traceability. Start small, measure and scale.
Shall we talk about your use case?
AI applied to financial work is no longer science fiction, but it isn’t magic either: it works when it relies on organized data, measurable use cases and human judgment. If your company has a "last mile" that consumes hours every month, that’s a good starting point.
At Tisa we help you assess it with no obligation. Write to us at info@grupotisa.com, call us at (+34) 971 305 885 or visit grupotisa.com and let’s study together where to begin.
Source of the cited case: OpenAI article about Model ML and GPT‑5.6 Sol (provided by the editor). Figures according to Model ML’s own benchmark, not independently verified by Tisa.