How to Build a Board Reporting Assistant for Executives

Board pack preparation consumes senior time out of all proportion to its value. Executives assemble figures that already exist in systems, write variance commentary that repeats last quarter's, and produce documents whose length correlates inversely with how carefully they are read. An assistant can take the assembly and leave the judgement. This guide covers building one, drawing on FISTA Solutions' AI agents work with executive teams. It complements the AI oversight for boards whitepaper and how to report ai progress to the board. This article is general guidance, not legal or financial advice.

Where must the numbers come from?

Governed source systems, through defined queries, with version and extraction timestamp recorded. Never from a model's recall, never from last quarter's document, never from a spreadsheet whose lineage nobody can state.

This sounds obvious and is violated constantly, because copying a figure forward is faster than re-deriving it. A board pack figure that cannot be traced to a source is a liability, and the assistant's discipline here is one of its main contributions.

Element	Source	Owner
Financial figures	Finance system, governed query	CFO
Operational metrics	Operational systems	Function head
Variance narrative	Drafted from evidence	Function head
Materiality calls	Executive judgement	CEO/CFO
Outlook and guidance	Executive judgement	CEO/CFO
Risk assessment	Executive and risk function	CRO

How far can variance narrative go?

As far as the evidence supports. Decomposing a revenue variance into volume, mix, price, and timing effects is arithmetic the assistant can do and explain. Identifying that the shortfall concentrates in one region and one product line is analysis it can perform from the same data.

What it must not do is speculate about cause. "Revenue fell because of softening demand" is a claim with no evidence behind it unless the evidence exists, and a board acting on a fabricated explanation is the worst outcome this system can produce. The rule is that every causal claim cites something.

Why do forward-looking statements stay with executives?

Because they are assertions the board will rely on, and in a listed context they carry legal weight. Outlook, guidance, confidence in a plan, and risk assessment are judgements informed by things not in any system — customer conversations, competitor signals, the executive's read on their own team.

An assistant drafting these produces confident text that no one has actually thought about, which is precisely the failure that makes boards distrust management reporting.

Why is cross-period consistency worth engineering?

Because boards read across periods. A metric that changes definition without explanation, a section that disappears, a chart whose axis changes — each costs credibility out of proportion to its significance. Packs assembled by rotating contributors drift constantly.

An assistant that holds the pack structure, metric definitions, and presentation conventions as governed configuration eliminates that drift, and flags a definition change as something requiring explicit note rather than something that happens silently.

What should a good pack look like?

Shorter. The dominant failure in board reporting is length: comprehensive packs that directors skim and that crowd out discussion. The right ambition is a pack that takes less executive time to produce and less board time to absorb, with detail available on request rather than included by default.

Automation that produces the same eighty pages faster has optimised the wrong variable. Building the assistant is a good moment to reconsider what the pack is for.

How should exceptions be surfaced?

Prominently and early. Directors need to know what changed, what is off plan, and what needs a decision. An assistant that leads with exceptions and relegates the steady-state to appendices produces a materially more useful document than one that presents everything at equal weight.

Deciding what counts as an exception is a threshold conversation with the board chair and the executive team, and it is worth having explicitly.

What about the questions boards ask?

The follow-up questions after a board meeting are a substantial hidden cost, often answered by analysts over the following week. An assistant with access to the same governed sources can answer many of them immediately during the meeting, which improves the discussion and removes the follow-up cycle.

That capability should be scoped carefully: answering from governed data is fine, and speculating in a board meeting is not.

How does it integrate?

With finance and operational systems for data, with the document toolchain the organisation already uses for output, and with whatever board portal distributes the pack. Executives should edit in the tools they know. A separate authoring environment gets used once.

How is it evaluated?

On executive hours spent preparing, pack length, time from period close to distribution, proportion of board meeting spent on discussion rather than clarification, and the number of figures queried as wrong. The last is the one that governs trust.

What does the build sequence look like?

Two weeks agreeing metric definitions and governed queries with finance, which surfaces disagreements worth surfacing. Two weeks on assembly and consistent presentation. Two weeks on evidence-grounded variance drafting with strict citation. One week on exception surfacing. Executive review workflow last, because it is the part that must feel natural.

What goes wrong?

Figures from documents rather than sources. Speculative variance narrative. Generated outlook. Automating an eighty-page pack instead of shortening it. Equal-weight presentation with no exception hierarchy. And measuring production time while board discussion quality is unchanged.

What does it cost to run?

Low, since the volume is a handful of documents per period. The investment is in the governed query layer and metric definitions, which most organisations need regardless and which pays back in every other reporting process that touches the same numbers.

Who owns the assistant?

The company secretary or the CFO's office, not IT. Board reporting is a governance process with named accountabilities, and the system that produces it needs an owner inside that process who can arbitrate metric definitions and decide what the pack contains. An unowned reporting tool drifts back toward whatever each contributor prefers within two cycles.

How FISTA Solutions helps

FISTA Solutions builds board reporting systems with governed source queries, traceable figures, evidence-grounded variance narrative, consistent cross-period structure, exception-led presentation, and executive ownership of materiality and outlook, through AI agents, AI enablement, and forward deployed engineers. The record behind the approach is 150+ projects for 50+ companies with 47% efficiency gains.

To shorten board pack preparation without weakening what the board sees, message FISTA on WhatsApp, or read the AI oversight for boards whitepaper.