Knowledge Organization Systems (KOS) are a ‘looking glass’ by which we can render reality. Several KOS use modal verbs, adjectives and rules to indicate forward-looking assertiveness sentiment (opinions) within texts. However, no prior work has sought to assess the strengths and weaknesses of these KOS. Many existing KOS do not account for ‘intensity’ of forward-looking sentiment. This exploratory study aims to assess how a composite KOS could be an improvement on existing individual KOS when applied to company annual reports, in order to assess the extent to which cultural differences and rhetoric can be identified and future performance predicted. A range of existing KOS were critiqued leading to the creation of a composite KOS representing the ‘extreme edges’ of strong assertive and hesitant forward-looking opinions. The composite KOS was applied to the genre of company annual reports for four large multinational Oil and Gas (O&G) companies between 2008 and 2015. Word frequency and biologically inspired diversity ratios for the categories were generated and examined. A strong association was found between a decline in the use of assertive language for some companies over time, potentially indicating increasing business uncertainty. Sharp increases in mentions of ‘future’ and ‘learnings’ appeared to be linked to industrial disasters and crisis management in two of the companies studied. This may evidence aspects of organizational rhetoric and renewal. There was evidence to support the ‘Pollyanna effect’ in one company: the social phenomena of over-positive business language. In the same company, a moderate association was found between increasing diversity of assertive language and declining financial performance the following year. This marker has not been reported before and provides an area for further research. It is possible that while some companies render changing attitudes towards the future state of affairs through word usage in company reports, others may not and some may even deploy rhetoric. This may point to a more complex model than ‘universal laws’ when generalizing and interpreting organizational word usage in company reports.Sentiment is typically applied with an a priori hypothesis in mind. Embedding these sentiment algorithms in standard enterprise search and discovery technology deployments may help generate new insights and knowledge in the most unexpected of places.