AI training for treasury: from sources to a concise brief

Build the program around the team's information

Syntalith proposes team training tailored to how treasury staff prepare information. When planning the program, we discuss the publications, tools, and internal briefing format the team uses. A piece of work that needed further explanation helps us choose exercises. The difficulty might be an overly broad question to AI, a comparison across different periods, or an abbreviation that a colleague in finance cannot follow.

The proposed scope starts with what the reader needs to know. Analysts practice selecting a primary source, finding its definitions and observation period, and comparing publications while preserving those distinctions. The writing work covers producing a short overview, retaining references, and separating source facts from the author's commentary. AI helps organize material and prepare drafts. The program concerns working with information; investment, hedging, and financing recommendations remain outside its scope.

Reviewing drafts together makes feedback relevant to the team's actual difficulties. A brief with many references may still fail to answer the treasury lead's question, calling for a different selection of information. If a shortened passage loses the observation period, that sentence needs revision. Authors return to their own drafts with feedback they can also apply to the next publication.

A new release can describe an earlier period

Suppose an analyst prepares a morning brief from a newly published release containing observations from the previous business week. The AI draft calls them “today's market conditions.” A reader could take this to mean the brief describes the current day, although the author only has observations from an earlier period. Linking to the new release does not explain that difference.

The revised brief states the publication date and observation period separately. It keeps the reference and makes clear that current observations are still needed to describe today. The treasury lead receives a summary of the available material and a focused question: “Which approved source can we use to add observations for today?”

The Federal Reserve's weekly H.10 release publishes daily observations for the previous business week. Its publication date and the dates of those observations therefore describe different things. This provides a concrete source for practicing how to preserve the relevant period when summarizing information.

Involve the analyst and the reader

The analyst knows the material behind the brief. The treasury lead helps define the questions it needs to answer and what belongs in the internal discussion. A finance colleague who reads the brief can point out an unclear abbreviation or a passage that sends them back to the sources. Their involvement in selected workshop activities helps the team write for an actual reader.

We consider financial experience and familiarity with AI separately. Someone new to the tool can work with one publication and a short summary of their own; a more confident user can compare several sources with different definitions or periods. A later independent task can involve a correction to an earlier release: the author updates the brief and explains the change while retaining the original observation period.

The extent of draft review, any work between sessions, and each person's involvement are agreed in the program. Exercises can use public releases or fictional material. Internal documents must be approved for use in the chosen tool.

What to discuss before commissioning training

If retrieving recurring data takes most of the time, start by checking the existing data service or spreadsheet. Training addresses how people use sources and explain information to their readers. Connecting systems to produce a recurring report is a separate scope, discussed in our article on AI for CFO financial reporting.

In a conversation with Syntalith, describe a brief that left its reader asking what the data meant. We can use that example to agree which tasks the program should cover, who should join the review, and which tools the team will use. Further service information is available on Syntalith's pricing page.