Specialist reporting and advisory firm Falcon Windsor and Insig AI, a technology firm delivering advanced data infrastructure and AI-powered ESG research tools have published Your Precocious Intern, a research paper and practical model for the responsible use of AI in corporate reporting.

This report, based on engagement with 40 FTSE companies and analysis of all FTSE 350 reports published from 2020 to 2024 reveals that generative AI use is increasing across UK companies, often without training, policy or oversight.

While investors see AI’s adoption as inevitable and look forward to the benefits and efficiencies it could bring, they are increasingly concerned about its implications for the truthfulness and authorship of corporate reporting.

Practitioners and investors agree that reporting must remain the direct expression of management’s opinion and that without guidance, AI use in reporting risks undermining the accuracy, authenticity and accountability that underpin trust in markets.

The growing momentum in the adoption of AI leaves only a short window of opportunity to upskill and prepare to mitigate the risks it represents to the financial system.

The report sets out clear recommendations for how companies can bring AI into the reporting process by design, not by accident. At its core is the guiding principle: treat generative AI like a precocious intern: useful, quick, capable, but inexperienced, prone to overconfidence and should never be left unsupervised.

Use of AI is Growing, But More Training is Needed
According to the research, most of the companies interviewed were investigating generative AI, from ad hoc use of chatbots and testing Microsoft Copilot to very tightly planned and controlled adoption.

Only a few, generally larger companies, had formal projects in place for its use, particularly in the finance teams. There was a distinct lack of formal training on how to use generative AI effectively.

The number of FTSE 350 companies mentioning AI in their annual report more than doubled between 2021 and 2024, while the average number of mentions increased fivefold over that period. In 2024, 68% of FTSE 350 companies made some mention of AI in their annual reports (including 76% of the FTSE 100).

Not a single report yet refers to generative AI in relation to the reporting process, which suggests we have a window of opportunity to develop a practical model for its use.

Investors Want Reassurance on Accuracy and Authorship
All focus groups in the research acknowledged the attractiveness of AI to reduce a growing workload driven by changing disclosure requirements. The average length of FTSE annual reports has been increasing over the past decade.

Incoming changes to UK and EU sustainability reporting requirements suggest the trend will continue and generative AI could make the
reporting process more productive and efficient.

However, they fear the risks. Even assuming that an ‘enterprise’ version of AI is used and data is securely ringfenced, the interviews uncovered the following concerns:

• All output will end up sounding the same – generative AI systems used by companies tend to be based on a limited number of foundational models.
• It will appear as though leadership ‘can’t be bothered’ with reporting.
• Generative AI will make it easier to ‘game the system’ by the inclusion of tickbox buzzwords and phrases, especially with the initial analysis of reporting often coming from machines.
• Reporting will include poorly sourced information given that generative AI is a ‘black box.’

While investors are open to the use of generative AI for handling large volumes of information, they are clear that the voice of the report must remain human. They expect the opinions, judgements, and forward-looking narratives to come from management and the Board and not from a machine.

The concern is that outsourcing these elements risks weakening the relationship of trust that reporting is meant to build. Investors are
also asking companies to state clearly how AI is used in their reporting processes.

There is a Narrow Window to get it Right
The research suggests that general usage in producing corporate reporting is still low and most companies are in early stages of AI adoption.

This gives companies a crucial opportunity: to introduce AI into reporting processes thoughtfully, with appropriate checks and balances.

Acting now allows organisations to reap the benefits of efficiency while ensuring that reporting continues to meet its fundamental purpose: building trust through accurate, clear, and authentic communication.

There is Scope for Guidance from the Regulator
The research participants do not expect or want more regulation but believe a clear signal from the Financial Reporting Council or the FCA would be welcome.

Even a gentle reminder that the use of generative AI does not change companies’ and directors’ existing duties, but could have a significant effect on how they discharge them.

As generative AI becomes more embedded, such reassurance could help companies strike the right balance between innovation and accountability.

Recommendations
The research suggests that ‘The precocious intern’ is bright, capable, and eager but inexperienced and prone to overconfidence.

Treating AI like such an intern means checking its work, giving it clear boundaries, and never allowing it to operate unsupervised.

Everything about how companies use generative AI in reporting should flow from this mindset, according to the report.

To make the best use of a new army of precocious interns, the report advocates:

• Introducing a formal training programme, with modules for reporting confidential information, requiring people involved in corporate reporting to take part and tracking their participation
• Learning, and practising, how to write good prompts, but never forgetting generative AI is trained to respond to a prompt, not to give a truthful answer
• Becoming a better reader, so you can properly judge what the output means and whether it will be understood by your audience.

The report almost recommends managing disclosure using:

• In the short term, a general statement either as a note up front saying generative AI has not been used or a discussion in the governance report saying it has
• In the longer term, specific disclosure explaining the policy the company has taken to using generative AI in reporting and including a negative statement of use in sections that cover forward-looking information and matters of opinion – neither of which are, in the authors’ views, appropriate uses of generative AI

Used well, generative AI can support administrative and process-heavy tasks: summarising meetings, cleaning up copy, and helping draft routine disclosures.

Recommended reading

• Gartner: CFOs Should “Reset” Expectations Around AI and Productivity
• Global AI Governance | Who’s Leading, Who’s Lagging?
• AI Anxieties Causing Business Leaders to Rethink Risks

It’s also useful for research, creating visuals and making early-stage edits. But it should not be used for writing opinion-led sections, strategic messaging, CEO letters or forward-looking commentary.

The report is adamant that it cannot be trusted with final sign-off, nor should it be allowed to work with sensitive data outside secure systems.

Companies should ensure training is in place, establish simple governance guidelines, and clearly disclose if and how AI was used in the
report to preserve both the integrity of reporting and the trust it is meant to build.