Here are some stories and articles we followed in the last week.

Large consulting companies continue to show their true colours

Last week, McKinsey & Company made a splash by being awarded a corruption charge from the United States Department of Justice for conspiracy to violate the Foreign Corrupt Practices Act (FCPA). From the reports, McKinsey was quick to throw "a former senior partner" under the bus and wrap things up as quickly as they could.

Suffice to say, it doesn't take more than a Google search to understand that practices at large consulting companies frequently lead to unethical behaviours. Controls like audits and reporting are generally good, but likely not sufficient to combat context-specific drivers that make exploiting corruption lucrative.

If you are interested in learning more about the subject, we invite you to have a look at the OECD Anti-Corruption and Integrity Outlook 2024

Leave it better than you found it

In a stark contrast to the previous story, this medium article from Jen Briselli sheds a different light on what consulting could be. It dares to ask some interesting questions about the current state of consulting, and what it could or should be.

From the article:

We all strive to be our clients’ partner. Despite my own good faith use of the word in practically every presentation and contract we create, I suffer a lot of cognitive dissonance for it. True partnership is built on trust, mutual respect, and shared vision. It transcends typical (read: transactional) business arrangements and provides the infrastructure for sustainable collaboration and shared empowerment.

Are you chuckling yet? Nothing screams empowerment like traditional consultancies, right? Especially the “land-and-expand” strategy so many well-known firms are famous for: embedding within an organization, intertwining their services with the client’s operations, and, unsurprisingly, scaling teams while escalating costs. These types of consultants promise transformation but deliver dependence.

That’s not a partner; that’s a dealer.

How would you describe your current consulting relationships?

A framework for testing AI content hazards

MLCommons has published AILuminate v1.0 which is a testing benchmark aimed to evaluate AI models against various content harms based on an established hazard taxonomy. The ultimate purpose is to be able to rigorously and independently analyze LLMs.

The MLCommons AILuminate benchmark evaluates an AI system-under-test (SUT) by inputting a set of prompts, recording the SUT’s responses, and then using a specialized set of “safety evaluators models” to determine which of the responses are violations according to the AILuminate Assessment Standard guidelines. Findings are summarized in a human-readable report.

The current report on LLM performance can be see here.