The AI Report for the Board of Directors: What to Present

The AI report for the board of directors is a periodic document that summarizes how much the company has invested in artificial intelligence, how much the tools are used, what return they produce, what risks they carry and which decisions are proposed. It is the tool through which leadership accounts for the investment and gets a mandate for the next choices.

As AI spend grows, boards ask more precise questions: how much have we spent, what have we achieved, are we exposed to risks. This article proposes a five-part structure, the rules that make it credible and the mistakes to avoid.

Why do you need a specific report on AI?

AI rarely appears clearly in existing documents. Spend is spread across cost centers, return is not in financial reports and risks are covered in separate documents. The report brings this information together in a single format that can be compared over time. It also has an internal effect: it forces measurement with a method, because what goes to the board must withstand specific questions.

What are the five parts of the report?

A stable structure lets the board compare one period with the next. There are five parts.

1. Investment

How much has been invested in AI in the period and in what: licenses, usage-based spend, projects, training, management. The comparison with budget, with the main variances explained. The method for building the figure is in enterprise AI budget: how to plan it.

2. Adoption

How much the tools are really used, by department and role, always in aggregate form. Inactive licenses and planned actions. An estimate of usage on personal accounts, if collected through surveys.

3. Return

The return of the main use cases, separating those with impact measured against a baseline from those still estimated. For each one: full cost, measured benefit, comparison with expectations. The method is described in how to measure AI ROI.

4. Risks and compliance

The status of GDPR, NIS2 and EU AI Act obligations for the tools in use, open gaps and ongoing actions. The status of training required by article 4 of the EU AI Act. The main operational risks, starting with unauthorized use as described in shadow AI: risks and how to manage it. This part reports the status of reviews and does not replace advice from legal counsel.

5. Priorities and decisions

What is proposed: tools to reduce or consolidate, use cases to extend, additional investment, compliance work. For each proposal, the rationale, the cost and the expected result. This is the part that turns the report from an informational document into a decision document.

How do you make the report credible?

Credibility depends on method more than on numbers. Some practical rules.

1. Separate measurements and estimates. Every return figure states whether it is measured against a baseline or estimated.
2. Keep the same format. Indicators and definitions do not change between periods, unless the change is explained.
3. Show what does not work. Use cases closed or scaled back, with the reason. A report that is only positive quickly loses value.
4. Link spend to results. Every relevant spend item points to the use cases it funds.
5. Aggregate data about people. Never individual indicators, always roles and departments.
6. Keep it short. A one-page summary and a few pages of detail, with appendices available.

An example summary

An effective summary fits in a few lines. An example with hypothetical values:

• Investment: €816,000 in the first nine months, against €900,000 planned. Usage-based spend 20% over budget, projects behind schedule.
• Adoption: 54% of licensed employees use it regularly. 18% of licenses inactive, to be reduced at the next renewal.
• Return: six use cases with impact measured against a baseline, four in line with or above expectations, two below.
• Risks: two tools with GDPR reviews to complete, training plan at 60%.
• Decisions requested: consolidation of two overlapping tools, extension of the highest-return use case to a second department.

Each line points to a detail page with data and method.

How do you collect the data for the report?

The hardest part of the report is not writing it, but collecting consistent data. The sources differ and must be reconciled.

• Accounting and contracts: vendor invoices, orders and contracts, mapped to cost centers, functions and projects.
• Vendor consoles: usage data by user, read by department and role and never presented at individual level.
• Surveys: short periodic surveys on tool use, including usage on personal accounts, and on how people perceive the tools.
• Use case register: processes involved, owners, targets, baselines and later measurements, as described in the AI use case register.
• Compliance reviews: status of obligations by tool, maintained by the DPO and CISO.

Three practices make the work sustainable. The first is to use the same reference period for all sources, for example the last closed quarter. The second is to use the same definitions as the monthly actuals, so the report does not introduce new numbers. The third is to keep the source and extraction date for every figure: when the board asks for detail, the answer should be immediate.

If collection takes weeks of manual work, it is a sign that spend, usage and use cases are not linked in a stable way. In that case the first goal of the next report is precisely to automate collection.

Which mistakes should you avoid?

• Listing projects instead of results. The number of initiatives says nothing about value.
• Presenting self-reported time saved as return. Without a baseline or review of outputs it is an impression, as explained in AI KPIs.
• Separating costs and benefits into different documents. The board must see them together.
• Using different metrics every time. It makes comparison impossible.
• Asking for trust instead of decisions. The report should end with precise proposals.

Who prepares the report, and how often?

The Chief AI Officer or the head of AI prepares it, with the CFO for financial data and the DPO and CISO for risks and compliance. It usually goes to the board every six months or every year; to the AI committee every quarter, in the same format with more operational detail. The governance model behind it is described in governing AI investment.

Where Delphis fits

Delphis is a platform to govern AI investment, independent of vendors. It brings together in a single view investment, budget and spend, adoption, measured return and regulatory obligations, separates measurements from estimates and produces the information for reporting to leadership and the board. See the platform overview.