AI: Managing the Goldilocks problem
If you are a member of a Board, or an executive management team, the chances are quite high that you have already encountered the AI ‘Goldilocks’ problem at a management awayday or Board strategy session.
Author: Jon Rowden
Somebody with knowledge of the multi-faceted impact of AI provides a lively and entertaining presentation that is relevant to your organisation, statistics are quoted and a brave new world is strongly suggested, one full of opportunities and threats for organisations and their people.
The presentation of course is just the hors d'oeuvres, as the main course is about to begin. The participants are split into table groups and invited to brainstorm the organisation’s action plan for AI. Twenty minutes with reporting back on the flipcharts provided.
It doesn’t always happen quite like this, Boards and management teams have other ways to identify someone responsible for assessing the effect of AI on their business and proposing a plan. However it's done, there’s a strong chance that sooner or later those looking at the challenge realise that AI presents a ‘Goldilocks’ problem.
In the well-known folk tale, Goldilocks was a character who wanted things just right. Porridge should be neither too hot nor too cold, chairs and beds neither too hard nor too soft. You get the idea – a suitable landing zone in the middle. A Goldilocks problem is where a state of balance is required between two extremes. For those in senior positions at listed companies, AI represents this because, of the ever-expanding array of things that can be done with AI tools, many of them are either:
- Helpful to people in their individual roles, but not clearly transformative to an organisation’s strategy, results, prospects or share price. Presenting these as an AI strategy begs the response “Is that it?” or “Aren’t we doing that already?”.
- So transformative that credulity is stretched beyond what is reasonable. Whilst exploring ambitious ideas can lead to good outcomes, anything that assumes AI technology will become super-charged in specific ways, that its flaws will be resolved or that human judgement will be replaced, involves a considerable leap of faith.
Finding new courses of action involving AI that land between these extremes is of course very possible. Corporate Reporting is a fertile area to consider. At Reportl we work with design agencies and companies that are doing it. Here’s how it works.
First, reflect on the following paradox. The most trusted and comprehensive source of information about a listed company is its annual report. Directors make declarations, controls and digital tags are applied, auditors conclude on the financial statements and read every page for inconsistencies. All that cost and effort because investors need the information and regulators created rules to govern its provision.
But today’s annual reports remain amongst the least discoverable information in the AI world. Because typically its information is buried in a PDF document that AI struggles to crack open and extract accurate meaning from. Type a search request for information you expect to find in an annual report into one of the commonly used AI tools and it will often respond with no answer, a wrong answer or an answer that is sourced from a website that is unconnected to the company.
So the paradox is that your company information that has been most painstakingly assembled and often matters most, is the least accessible to AI.
This is not a new problem. It’s a fundamental limitation in the way PDFs are created. Web search engines have struggled with content in PDFs for 30 years, and AI’s struggles are just an extension of the same problem. Search engines have always ranked well-formed, accessible digital formats more highly than PDFs in search results.
However, with advances in digital software, there are now solutions to this problem. And for management teams or Boards of Directors it’s simple decision to publish the entire annual report directly on the web in AI-friendly digital format. Better still, it can be delivered by the design teams that already deliver annual reports. And it is recommended by regulators too.
An online (native HTML) format makes the annual report much more accessible for AI. There’s no need to take our word for it. This recent study has, in our view, proven the case beyond reasonable doubt. Incidentally, the acronyms involved can become a mouthful, so the shorthand term for creating an entire report natively in HTML format is “digital-first reporting”.
In digital-first reporting a full PDF and printed version remain. But crucially, they are produced at the same time from the same content as the digital version. This means that everything is geared for AI to find and understand the information better (without the need for any separate and costly bolt-on processes after sign off). It is truly digital-first reporting, and is optimised for all AI and digital search tools.
If you agree that AI-searchability of your corporate reporting is now essential, why not ask your reporting design agency about "digital-first reporting". They should know about it and may have been making plans for this question.
Most reporting design processes fall into two categories:
A digital-first process in line with best practice guidelines using digital-first, native HTML systems like Reportl.
A print-based design process using older tools to deliver report PDFs first, with very separate bolt-on processes and digital deliverables created as a by-product.
The first, native HTML, process is purpose built to meet AI’s needs, accessibility standards and deliver all formats end users now require, including the PDF and digital versions. This native HTML is the approach recommended by reporting regulators, data scientists and standards setters. It is more efficient and easy to implement.
The second, print based process, has been a workaround solution for many years, but doesn’t solve the inherent limitations of the PDF as a source format. It is also a more manual, fragmented and slow approach. The report versions typically lack the fully digitised report content which is what AI thrives on.
At Reportl we make the digital-first process simple to implement for design agencies and companies that wish to make reports AI-accessible. If you would like more details on this, please contact us - we would be delighted to talk it through and show you how it works in practice.
