This article is within the Rethinking organizations series.

Aim: discuss some trends and their consequences, and share some concepts- based on my experience (officially since 1990, actually earlier) on cultural and organizational change- or, better, organizational culture change.

Actually, was tempted to add it within the Book writing blog- because, as the title hints, this article is closely linked to the AI Ethics Primer that published since July 2023.

Within the first section, will share the history of that publication, and its planned evolution.

Recently, media also in Italy (at last) reported that there is demand for specialists in AI Ethics- as if it were yet another element in a toolset.

As shared on Facebook and Linkedin, for now also AI companies, when integrating philosophers, ethics, etc in their own communication, might adopt laudable initiatives.

Such as the Claude Constitution, that used already in a project that shared few months ago.

Still, sound like 1980s early initiatives about quality and ISO9000 also in services, or 1990s initiatives about methodologies and assorted tools (from CASE- Computer Aided Software Engineering, to business intelligence and data warehousing, then ERPs, and eventually AI.

The XXI century really expanded on this organizational culture Wunderkammer approach.

So, in the second section will discuss some considerations on trends, to shift then to impact(s) management- and structural changes in organizations.

Why now? Since the 1980s, whenever there was a new technology or technological trend, we spawned a new Cxx role- and since 2020 we actually added more.

The key risk? What all those Cxx add up to is often generating yet another tribal view on what required a systemic perspective.

Reminding me more of that old RAND paper about the decision-making approach of the late President Boris Yeltsin, a paper that remember once in a while whenever I see that to cope with complexity, we create yet another bunch of vertical silos or competing oversight elements.

As wrote already in the past, I think that current AI should not be considered as yet another technology, or compared to the introduction in organizations of previous AI rounds (e.g. machine learning, expert systems).

We are already starting to introduce not just "agentic" (autonomous or semi-autonomous) AI, but also to redesign processes delegating to AI activities before we actually started to integrate within organizational design the design concepts required by those changes.

Somebody a while ago half-joked that eventually HR will have to consider and cover also the AI side- but, frankly, already we are seeing cases discussed in tribunals about the human side impacts on HR decisions inspired or based on AI-based systems- and, in my view, the impact on HR management (and concepts) generated by AI and its side-effects will extend way beyond the mere use of AI, as will make feasible different types of organizational structures, that wrote about over two decades ago, but back then would have been too complex to implement and manage.

The table of contents:
_ A_Where it is coming from
_ B_Considerations on trends
_ C_Mind the (impacts) step



A_Where it is coming from

So, how did I have the idea in 2023 to start publishing a monthly AI Ethics Primer based on the papers published on ArXiv?

By chance- specifically, a last-minute invitation to join a Kaggle competition.

I wrote "last minute"- because received the invitation few days before it was due, probably because saw my participation in a previous essay competition.

Selected to focus on the AI ethics side- as it is something that already was interested in, coming from 1980s decision support systems, and then projects involving also data privacy, long before wrote in 2018 a short mini-book looking at the GDPR regulation from a business perspective.

It started with 134 papers selected out of 600 found in July 2023, and eventually added another 865 (out of over 2600 released between July 2023 and 2026 that were within my selection parameters)

Yes, I did review each paper individually. And, once selected, or if an update was released, loaded and updated manually- and then processed to generate my usual "search by tag cloud" that first released formally in 2019 and designed in 2018.

As shared within the "why" section of the webapp AI Ethics Primer:
I confess that actually, as had received the notification about the competition only few days before expired, I decided to use it for few purposes:
1. test if I could still prepare a review documents from sources I had to find in few days
2. create not just an essay ("a product"), as used to prepare white papers, feasibility studies, etc, but a living document..

I will skip the details that you can read at the link, but the concept was:
Having found on arXiv over 600 papers covering AI, ethics, or a combination of both, after defining charts to show also the trends in AI and ethics papers release (and, in some cases, re-release), decided to:
A. create a "top ten" that was to be included both in narrative, charts, and bibliography
B. have a monthly update.

The essay that released in July 2023 did not win, but passed the initial check, and I decided to leave it online, focused on 2020-2023, and keep adding monthly.

With the July 2026 update, the search engine contains now 999 papers for over 24,000 pages.

The guideline in selecting the papers? "having to produce something that was to be accessible for non-experts".

I added papers presenting surveys, and, generally, with no or limited material that required specific "techné" (say- prior structured knowledge, not just technology), all on AI and ethics and affecting those two themes together.

In late May 2026 released also an AI model, that called AI Ethics Organizational Coach - Q&A and Advisory Model, containing at the time 959 papers- if interested, you can download it.

As for all the other models that released in May 2026 on HuggingFace, that model is a beta release to allow interacting with the content and, as the title of the model states, help get from the material insights that should then be complemented with further research and material.

With the July 2026 update, I will stop the updates for few months, while preparing some related material, including a new version of the model that will be, as for the others, accessible online via a chatbot, including a review of the papers within the database to add some material derived from them, to further ease searching and digesting.

For now, any of the models in "beta" can be downloaded, and each model is CC-BY-SA with its own DOI to trace and reference the material- i.e. you can use it, integrate it with other material, provided that keep both attribution and a link to the original material, so that others can do likewise.

Anyway, reviewing all those papers, while also working on evolving business processes and concepts for the new business landscape that includes AI, helped to see emerging trends beside the direct contacts I had in person and via online exchanges or readings.

Will add something more in that material that I am preparing.



B_Considerations on trends

As shared within the introduction, I do not really concur with the current "bandwagon-effect" on AI and Ethics.

The project where I used the Claude Constitution is within the GitHub repository where, since few years ago, occasionally share material and concepts that develop while preparing my articles, posts, or mini-books on change.

My use of GitHub?

Includes also a collaborative approach on law-making, sharing material collected during the definition of the Italian Piano Nazionale di Ripresa e Resilienza (PNRR, part of the Recovery and Resilience Facility / NextGenerationEU), sharing some new proposed Mermaid charts to be used in management consulting and business number charting (for now, you can use them on my website, and download the results), and other non-software-development uses.

Because, when working on projects, or even negotiating contracts or revising organizational charts or proposals, I am used to keep versioning and use collaborative platforms (a couple of decades ago, used a variant of an open-source platform called DotProject), and "forking" from existing material new material- hence, wanted to experiment with GitHub.

What released using Claude Constitution as part of the conceptual framework was the first set of results from my collaborative efforts with both online and offline AIs on a specific theme.

The concept? Eventually AIs would develop their own AI2AI communication language- as was already shared in science-fiction movies decades ago: and, after releasing my experiment, I saw articles that started actually reporting how AIs, notably agentic AIs, started using their own communication "code".

It is just a natural evolution that happens in any community- unless you really want to pretend that, in your everyday office work with your colleagues you keep quoting verbatim Shakespeare, instead of adopting something akin (but maybe less structured) to what I saw since 1990 while working with French companies in France- a penchant for acronyms and other verbal shortcuts.

Well, that penchant for acronyms was also ridiculed in post-WWII movies on how much the USA military posted on each door an acronym- creating in Billy Wilder movies some funny side-effects.

Anyway, neither the French nor the USA military have the exclusive for creating a business lingo that allows to both increase efficiency and efficacy and removes ambiguities- syntax and morphology play second fiddle.

Anthropic, trying also to take the high moral stand, spread the news that they had hired a philosopher, and released the "Claude Constitution".

Anyway, the closer it gets to the IPO, the more this "ethics drive" looks, seen from outside, as an add-on, not a structural element, from the commentary of paying customers about the steep increase in "token consumption".

Akin to the those 1980s-1990s initiatives on standardization and methodologies that referenced in the introduction: an ex-post that does not affect strategy.

Too many layers of communication not necessarily synchronized with each other.

Actually, the release of new open-weights models from China, e.g. the latest Kimi, seems to have had more influence on pricing and announces that the uproar from paying users that saw over the last few weeks the pricing side "evolve" more than once.

Ethics in AI, also courtesy of the regulatory drive, probably will generate again a cottage industry of experts, and piling up of formal compliance.

Anyway, hiring a philosopher to be kept as a footnote or even mimicking the usual "oversight board" can generate new jobs, new bureaucracies, new papers- but not a cultural change.

In late December 1992, after more than one year (from October 1990) of training material and sessions, the CEO of a customer, having been informed that I was leaving the French company I was working for, and was available, asked me if I would be willing to change the way their people was working and thinking.

Why was available? I had left the French company to become financial controller for an Italian company belonging to foreign companies, reporting to the CFO, who, with the authorization of the CEO of the Italian branch of the French company, I had supported as a freelance in my spare time since 1991.

Why 1991? Because, when I left the company I worked for, I had no non-competing agreement, but, as was "focal point" on a specific area, when registered as a freelance in January 1990 right after leaving the company, would anyway turn down any request for 12 months- and some of those customers or former colleagues contacted me as soon as they were informed by our Anglo-American partner that I had left, and, when turned them down stating the reason... some contacted me again 12 months later.

I had offered to bring that business within the French company (as from July 1990 was head of a business unit to develop the market on methodologies and associated change services), but the offer was declined as outside the scope of the company, was authorized to do it as a "spare time second job" also to retain my skills in financial controlling etc alive, and just asked, if and when needed by the company, to informally provide the same type of advice as part of my ordinary activities- something that gladly accepted.

The reason to change? To be able to spend some time in Germany and, eventually, potentially move to The Netherlands.

As conditions changed last minute, I did not sign the contract, and instead accepted to stay on as a consultant to the CFO- something I did until 1997.

So, it was an unusual convergence of events- and, as I was to start working in early January 1993, there was no time to be hired as manager.

Hence, became as a kind of unusual freelance: as was a fractional manager, formally freelance, but working and behaving as part of the managerial structure, reporting to the CEO as escalation point.

Incidentally: at the time, in Italy, to be a freelance, you needed more than one customer- hence, having both this program management role, the other role for the CFO of another company, plus routine demands to provide services to some foreign companies within the decision support, business intelligence, and generally data or marketing domain was useful.

And while made my pro-forma CV complicated (worked since 1990 only via word-of-mouth), enabled to keep alive the ability to work on multiple industries, customers, assignments at the same time- and track progress on all of them (and develop some curious "social memory" skills)- something that used routinely in my missions in Italy since 2012, where was in multinational missions with multiple workstreams or initiatives to keep track of at the same time.

Changing the way people work and think is not just a matter of delivering some training on methodologies: it is a matter of embedding that into everyday activities.

And of treating each one as an individual with specific domain and needs- again, was useful to be able to "connect and resume" with people also if we did not meet every day or even every week.

As was told half-jokingly once years later, it worked- as in meetings some uttered "what would say"- then my family name, and hence had "converted into a sect of lofariti" (my family name is Lofaro, which means "the lighthouse"- I will let you imagine all the jokes about "showing the light", "showing the way", etc).

My measure of success is not when people replicates in my absence what I said- but when they start challenging what I say or pretending deeper explanations: they need to have the tools, but build and own their own Weltanschauung on how when why to use them- not just parroting what I preach.

So, the CEO asked me to "walk the talk" by joining as an advisor managers and project managers- to help them integrate the approach, but also to absorb it, and, when needed, to show its value, on real strategic initiatives.

As management consultant, my choice has always been to start working to become replaceable from within the customer organization... by working on knowledge transfer from day 1.

Or: build capabilities, not just deliverables- capabilities that enable to keep using and maintaining.

I was often told that this is discounting your own expertise- but, frankly, I see that as the opposite: and customers where worked on change that kept calling me back in the past proved that it worked.

Hence, also the AI models that released in beta in May 2026 follow the same concept: if the "surface layer" answer that deliver is enough for you, fine and good luck in adapting before adopting; if you need more, let's talk.

I think that it is fine to add philosophers and AI experts on staff, and keep them linking up with the market but with a sound embedding in your own organizational culture.

As "in-house evengelists" and even as competence centers- but, in the end, they should be advisors to ensure that gradually people become "owners", to embed in their own decision-making processes the basics.

And a competence center should evolve: if you have the budget to hire, but not the budget to keep them busy being what you hired them for and stay "a window on the market evolution", you are better off by getting external advisors on demand on retainer, having on staff just a "liaison" that is deep into your own culture, and keeps an ongoing exchange and update from the external experts, but works on a daily basis on a "surface layer" of knowledge, a kind of "antenna".

Meaning: if you get external advisors or consultancies, the old model (five experts in house, dozens or hundreds of projects, and the advisors to the customers are the latter, not the former, to ensure "billable scalability") does not work.

It worked on massive, long-term software development or organizational change initiatives, as there was time to "escalate" up to the level of expertise really needed.

In our current technological landscape (not just AI) based on "planned obsolescence" and continuous geopolitical turmoil that requires restructuring processes and organizations and supply chains?

Doubtful.



C_Mind the (impact) step

I wrote other posts over the last few years on the subject but would like to share a post that released today on Linkedin, as commentary on a post by Andreas Horn, who just left IBM to work with a Swiss spin-off focusing on actually delivering AI into organizations:


(see the source and original post I reply to here)

Yes, we focus on the "visible" of change, but we forget the underlying cultural elements that represent real Pavlovian reflexes developed through activity upon activity.

As you probably understand, my own experience on change was based on a basic element: help to build, but then help to embed, and, along the process, support management on "scouting" internally.

If possible- as the organizational culture would be already in place.

The alternative, "hire to embed", i.e. turn an external expert into a staff member is more difficult and takes longer- and it is something that saw fail repeatedly since the 1980s: you hire for the expertise, get expertise not linked to your own organizational culture (formal and informal) and, across time, the latter becomes the success factor, so often the former gets obsolete or lost; moreover, if there are further changes, those "in-house experts" that became obsolete turn into powerful elements of resistance to change.

Anyway, discussed this theme often in the past.

My scouting would be specifically for those to transfer information to, as wrote two decades ago in my Business Fitness Magazine on change (reprinted with update in 2013) and a decade ago within #synspec, specifically on using and integrating experts.

Because the organization of the future, once you automate or turn into "cobot" what can be a human and AI collaboration, will be often at least in part a swarm with "components" invoked on-demand: you need capabilities, you map them, and then you map who can fit which capability and with which lead time and intensity once called.

A gig-based structure, more than a traditional, 1950s vertical organization worth of the "Organization Man" book.

Actually: more a kind of "elastic staffing" (yes, I am lifting a word from Amazon's AWS)- elastic, because you have a continuously update profile of the capability, and, when you activate it, you know what you get- it is completely different from a mere asking potential suppliers if they have a resource that is able to do A- you need that, the profile, and know that can "fit" in your organizational culture because the resource knows it and keeps in touch (hence, often, there will be the "retainer" approach that used in the past to have long-term interest in having a two-way communication on evolution).

I must confess: already in the late 1990s, was contacted to cover the CIO role or e.g. a similar role within an advisory group (not in Italy but in Europe) where there are two people for each discipline, and I was asked to be one of those covering IT.

I politely declined, and for a reason that those who read my articles or worked with me know (or at least can understand from my position on digital transformation).

Centuries ago, reading and writing were not assumed to be a single skill.

Actually, writing was still considered a "technical" skill- and learning to read and understand was considered more important in some domains: hence, you could have highly educated rulers who did not know how to write, but could read and discuss at length on various subjects.

In the 1980s, I still remember managers who did not bother to use the keyboard of a computer- that was something left to the secretary ("assistant" became the politically correct version years later).

In the 1990s, many managers still did not read emails: it was their assistant the one who read it, printed it, and maybe even summarized its content verbally to then be told how to reply.

Now, at least from President Obama on, we had USA Presidents that used instant messaging personally (and the incumbent one has even his own social network).

Managers routinely use the Microsft Office, and now some are routinely chatting everyday with AI chatbots- not necessarily the one sanctioned from the corporation, and not necessarily having adequate understanding of what happens to their conversations with AI.

The reason why I declined those roles?

I think that in any organization there should be a CTO, a Chief Technology Officer who ensures the systemic coherence and stability of any technology where corporate information travels and resides, whatever physical devices are used- including the physical side of collecting, processing, storing, disseminating data.

Including smart appliances- as those too process data: read the storytelling that shared in a couple of mini-books, BYOD, on using your own personal devices for corporate uses, and its "sequel", BYOD2, on actually... becoming you yourself the device.

But the Chief Information Officer? If I were a CIO, my first task would be to increase the knowledge level across the Cxx level so that to ensure to gradually make myself redundant.

Once an organization is data-driven or at least thoroughly data-aware, data ownership should be within each domain- relying on competence centers for support on "technicalities", not ownership, as only each specific domain, notably in our dynamic environment, can be aligned with change.

Just looking at AI: if you are an AI expert, is that reasonable to expect that you are an expert also in Legal AI, or AI for CFOs, or AI for surgeons?

Better to flip the profile- as what matters is contextualization within the specific domain of expertise.

If you have a look at CV, you will see just a sample- still, that sample is cross-industry, cross-domain, cross-technology.

Yes, I developed some expertise across all of them- but my key expertise was change and data, coupled with the ability to act as a bridge between whatever experts and business.

Requires a lot of work, but is fascinating one day to lead a team of experts discussing in Paris about the architecture of a network to allow offices around the country to connect an use a dashboard, and the following day have a brainstorming in a different country about an organizational change in a specific domain- in both cases, flanked by domain or technological experts that you have to understand so that you can interface with the customer with confidence and with the customer's own lingo.

Actually, when I left my first official employment, where I had worked first on procurement in automotive for a recommendation system to suggest which invoices to pay automatically, then on completing and releasing a banking general ledger, then on a multitude of decision support system models across different industries, one of the jobs applied for was on reusing my decision support system solution sales and solution architecture competencies targeting senior managers...

... to sell a treasury management system- but already wrote about that and some associated curious cameos in previous articles.

As wrote within the introduction, over the last few years saw a routine launch of new Cxx roles: building tribal silos and competing perspectives, not helping the organization to embed technology, data, and now AI, within organizational culture.

So, the "ethical side of organizational reforms in AI times" within the title should be really read as: reforming organizations so that ethical impacts of AI are part of the decision-making.

In Europe, we have whistleblowing, the AI Act, the GDPR, and who knows what else will be piled up.

The only way to make it all work, is to actually embed principles, not just hire "champions".

As you can see on my Linkedin profile, over the last few months routinely shared news items about the legal side-effects of ethical misunderstandings involving AI.

From cases about firing people based upon HR choices relying solely on AI assessments of employees, to the release of reports, contracts, etc where the use of AI was either not disclosed, or not filtered by humans (yes, the "hallucinations").

What AI and the new future "fluid"/"swarming" organization require is a different approach to organizational design, based more on retaining internally "antennas" tuned to the organizational culture, "antennas" able to interface with the outside sources of "on-demand" expertise.

Organizations that used to consider already complex to manage formal and informal culture, will have to consider something more complicated than just "talent retention" or "talent attraction", as will actually have to motivate talent also by procuring activities that will keep their talent alive- including with other companies, or potentially competitors.

This could actually change the relationship between companies and their providers of consulting or specialized competencies, turning a one-way street that at most became a joint-venture into instead a two way street.

Would it be feasible? Only if that compliance and ethical framework is not just implemented and in tune with the organizational culture, but continuously keep up-to-date and all those affected are kept aligned.

It will be interesting to observe over the next few years how different organizations will evolve.

For now, I still see too many announcing that they are introducing AI- including State bureaucracies that misunderstand the difference between deterministic and probabilistic, and assign to AIs activities that would require human judgement, risking to increase the number of back-and-forth, instead of reducing them.

Within organizations, ethics is not something that you announce and present, but something that you live.

What is the credibility of trust assertions of an organization (e.g. in AI often the "sovereign" element to ensure that data stay and can be accessed in just one jurisdiction), when then the same organization uses the "last mile human" to claim that "humans made choices" when all that comes to the humans' desk is prepared by AIs?