_
RobertoLofaro.com - Knowledge Portal - human-generated content
Change, with and without technology
for updates on publications, follow @robertolofaro on Instagram or @changerulebook on Twitter, you can also support on Patreon or subscribe on YouTube


_

You are here: Home > Rethinking Organizations > Under the hood: best practices, transformation, and Weltanschauung in data-centric and AI times- part 1 conceptual

Viewed 4902 times | words: 5535
Published on 2024-03-10 23:30:00 | words: 5535



As you probably noticed, recently started to publish articles once a week (between Monday and Sunday- no fixed day).

The concept is simple: embed into my daily routine- also in preparation of further longer publications.

Let's consider this article a corollary on Linking heuristics, AI, legacy, and... demographic trends - the case of Italy".

Actually- in two parts, first the conceptual (this week), and then the structural side (next week).

The key concept is always to think systemically, which implies thinking beyond the now, and, when at a larger scale, also beyond your own time.

Sections in this short article:
_ Weltanschauung
_ model
_ integrating
_ threats & choices
_ be adaptive

The second part will follow the same structure (the reason will be explained within the last section of this part).



Weltanschauung



This article will actually start with an apparent digression, that will "frame" all the others: an example of thinking systemically.

It will be the longest section, but serves as an introduction to both articles.

I do not know if this news item has been reported outside Italy- as it has been almost a routine to read about "leaks" and even resell of confidential information to third parties, since I returned in 2012.

The old "loose lips sink ships" does not apply in Italy, apparently.

The latest scandal? Started the usual way, a while ago, when somebody noticed information that supposedly was not public being commented on- it does not matter if that was on media or just appeared to be in possession of those who should not, upon release from those who had access to.

Why it does not matter? Because, in Italy, at least since I remember reading about it first in the 1970s, it has been a routine.

Over the last three decades, only expanded the volume- courtesy of course of technology, but also of the volatility of political majority since the beginning of the Italian Second Republic (1990s).

I do not really care about the specifics but, as the outcry was from Ministers of the current centre-right government, initially seemed as if the oppositions did not care.

Until it was shown to cover not just few politicians or entertainers, but a cross-section of society, including industrialists etc.

Moving from a few reported earlier, to few hundreds reported few days ago, to a much higher (and increasing across the week- now it seems steady).

Again: I do not care about the specifics, albeit few days ago shared as soon as the first disclosure reached media on Facebook my perspective: "
In 2004-2006 accepted to work part-time for a partner as PM/BA on Government and Government agencies projects...

...at a reduced rate, and eventually even using proceedings from other activities and savings to first cover the difference in costs, then give also some "free time" (for activities that were outside the scope of my agreement), and finally even to work for free and paying my own costs

And the Government was held by political parties I had never voted for (not before, not then, not after)- also if those projects could have a political potential.

Why? It is walking the talk about being bipartisan, and...

...thinking systemically, i.e. considering that if things are improved at a structural level, whatever the tribe that temporarily scores a success, all will gain.

I discovered then how deep was the operational impact of the Italian spoils system, were not just managers, but also whatever position opens are filled by the winners- and become "ascending tenures" if the same side lingers long enough.

Personally, back then and years later found puzzling how was being pulled or criticized by both sides, for different reasons.

As looking at improving for all is still not considered acceptable.

Whenever I worked as a management consultant on change, often suppliers assumed that I was part of the customer's structure- in Italian was "aziendalista".

But the concept is the same: whatever you lead or support leading, it is a temporary stewardship- and should take care of phase-in, phase-out, and paving the way for succession, not secession..."


At you can read the original post, with a copy of a newspaper article.

Updated then- now thousands of potential targets, not hundreds.

So, as shared also online, stating that just one single operator was able to access repeatedly and continuously data focused on individuals massively is, in my view, more puzzling than the famous wikileaks, where the access was on documents and exchanges.

And then eventually also others piled up on newspapers along the same line, after an immediate instantaneous response that seemed to converge on seeing it as a marginal issue linked to a rogue or disloyal civil servant.

There is a silver lining.

When you shift toward a systemic perspective, and accept that what matters is not quantity, but of the sheer possibility of an encore, it is not anymore a question of "rogue agent", but a question of structural changes, checks and balances, and, yes, "quis custodiet ipsos custodes"- who oversees the overseers.

The key element is shifting from what is a personal ("does affect me/my tribe?") to a systemic perspective ("is it structurally sustainable?").

Friend or foe, if we share the same social-business ecosystem, a systemic approach should be a default- anything else is at best shortsighted.

Yes, what two Italians centuries ago discussed as "particulare" (a short summary, in Italian, is within this 2014 article: Universale e "particulare": le concezioni opposte di Machiavelli e Guicciardini), is still the first instinctive answer whenever something happens in Italy.

If you read the list of sections contained within this article, you saw what is going to be within the next sections- at least, conceptually.

In reality, that you accept it or not, it all starts with your own perception of reality- acknowledging it so that you can remove biases.

If you are interested on exploring the concept of "bias", beside the latest few articles, since 2016 posted as of today 55 articles referencing the concept.

The idea is, anyway, to contextualize your own Weltanschauung within a wider context.

It is now trendy to talk about "commons", but generally we still have to learn to start asking from "universale" to "particulare"- and not the other way around- something that might be easier for some than others.

Fifteen years ago I was in Brussels and bored, so my morning walk to work, and the evening walk from work, were focused on listening to language courses recordings.

Specifically- Dutch, plus the basics of Mandarin and Russian: understanding, reading, listening in languages has been always more important to be than speaking or writing.

It was curious to learn how in Mandarin the specification is from general to specific, and think how this could affect thinking patterns.

Anyway, when living and working across multiple cultures (as I did), the impact of the most common tool (language at its basic, instinctive level, not its niceties that require formal education) on reactions to reality is sometimes visible (and can be used as a tool).

I will let others to consider and study (and I have no wonder that many already did- remember some studied on the number of ways to express "white" in some languages, or others such as Pirahá where counting is one-two-many).

There was actually an interesting book about the thousands of languages (many dying) across the world, but its focus was on linguistics not, e.g. how this could affect political behavioral patterns adopted by the community of speakers.

Still... as I wrote few days ago on Facebook, in a follow-up post (the first of a few), it seems that the two leading political parties, Government and opposition alike, are at last both discussing the issue as a structural element that should concern everybody- not just those directly affected.

Or: extending their own Weltanschauung (at least temporarily) to cover not just "who benefits", but also "who can get potentially affected".

And the obvious answer is: anyone.

The first bureaucratic reaction was to ask avoiding impacts on tools useful to fight crime: Italy is country where law allowed integration between databanks for access by State basically without any prior authorization on the specific individual.

It is a matter of efficiency and efficacy, of course, part of any digital transformation and e-government initiative since the 1990s, when OECD launched the concept to the front stage.

Anyway, within a tribal society, where tribe-family-State is the usual order of priorities that observed as a "locally born foreigner" since 2012, this could generate parallel lines of communication.

All this implies also that many layers of traditional privacy that relied on segregation of data de facto do not exist anymore.

I remember the protests in UK when the Government there tried to implement a similar measure, assuming that any civil servant would be able to access, removing the traditional "compartmentalization", and even, more recently (early 2020s) the complaints about potential privacy impacts of a more limited system called LEDS.

In Italy, we got used to it since decades: step by step, we might be behind on digital transformation and digital divide, but the level of digitally integrated exchanges between authorities, born out of a need to make less cumbersome the continuously overlapping and expanding bureaucracy demands (it was common for an office to ask you information that they had already), including to focus on organized crime fight, opened a Pandora's box within a tribal society- paycheck source does not imply tribal allegiance.

Hence, in this case it is not the quantity but the quality (accessing repeatedly information for unauthorized uses) that makes sensible to invoke the "quis custodiet ipsos custodes"- from a bipartisan perspective.

The current debate sometimes produced example of the same attitude that, frankly, I read often in articles by technologists or (other side of the coin) experts in philosophy and ethics that are stuck in a "silo" mentality, i.e. use their own perspective as a guiding light to analyze and lecture on what other "silos" should do.

This digression is actually useful to spot few key elements relevant to this article:
_ just because there is a potential shared benefit, this does not imply that, unless you get "commons-first", you will identify it
_ whenever there is a tribal mindset (also just "technologists" vs. "philosophers" or "politicians"), each tribe might project its own perspective on others
_ keeping in mind your own interest but then moving to a different, more general plane of analysis.

Just a closing point on the digression: do not worry, Italian politics is still tribal.

So, that silver lining I referred to, a glimpse of a potential systemic thinking focused on the "how" was possible, to avoid an encore...

...today again turned into the usual tribal differentiation: with oppositions apparently again trying to lower the tone, and those most affected (for now), i.e. centre-right and others from business instead focusing on "who" and "why".

Let's see what will happen next... but the digression anyway was useful as storytelling about how to contextualize your own Weltanschauung.

This can help both in defining a framework, and, more important obviously in "tribal" (or "silos", "experts", etc) contexts, implementation.

To make a long story short: any transformation assumes a model.



Model



I have been in the past in multidisciplinary groups, and eventually our "committee" was able to converge on something that was reasonable, and avoid group-thinking.

Not every committee having to select a horse comes out with a camel: it depends on both what members bring to the table, and how the coordination and facilitation is done.

Edward De Bono once described an approach to reduce water pollution from some manufacturers- have then a water intake that starts downstream from their own production, not upstream: i.e. consider, within your own "business as usual" a wider context.

Conceptually, this applies also to our digital and green transformation, the key initiatives that currently, not just within the European Union, are impacting our everyday reality.

And this brings about the "data-centric" side of the title, as neither digital nor green transformations within a sustainability framework (with all their associated paraphernalia of measurement, e.g. carbon impact across the supply chain, lifecycle product management, etc) would be feasible without massive amounts of data (yes, akin to that famous scene within the movie Matrix).

Ignoring the cost element, of course having massive amount of data potentially provided in real-time would not feasible even if you had legions of humans focused on the task of collating, filtering, processing them.

Hence, let's focusing just on the AI/data-centric side of both transformation (again within a sustainability framework): this implies designing models and concepts that actually "embed" a feed-back (and side-effect) on your own organizational choices.

To avoid time-delays due to harmonization and post-processing, to feed-back continuously on implementation adjustments, you need also "smart" technologies, both at the centre, and distributed across all the parties involved (technically, both "decentralized" and "Edge").

Just this week I read that, while nominally Italy and others adhere to all the sanctions, e.g. on exporting teak wood from Myanmar, it takes months after processing through customs to see the actual status, and that, anyway, the number of inspectors in many countries is between zero and a couple of dozens.

In our times, it does not make any sense, if you have sanction but then compliance has a time-delay that makes it an audit.

Anyway, I think that off-the-shelf AI is akin to what I saw in the 1990s with many off-the-shelf solutions for data warehousing and business intelligence or KPIs processing, or even ERP systems.

You are importing the culture "embedded" within the tools (or datasets, for AI) that you use.

Or: getting more data is not enough: you need to know how to integrate them and what they represent- a model.

Earlier this week, Yann LeCun, VP and Chief AI Scientist at Meta posted on Linkedin about the concept of "models": "
Lots of confusion about what a world model is. Here is my definition:

Given:
- an observation x(t)
- a previous estimate of the state of the world s(t)
- an action proposal a(t)
- a latent variable proposal z(t)

A world model computes:
- representation: h(t) = Enc(x(t))
- prediction: s(t+1) = Pred( h(t), s(t), z(t), a(t) )
Where
- Enc() is an encoder (a trainable deterministic function, e.g. a neural net)
- Pred() is a hidden state predictor (also a trainable deterministic function).
- the latent variable z(t) represents the unknown information that would allow us to predict exactly what happens. It must be sampled from a distribution or or varied over a set. It parameterizes the set (or distribution) of plausible predictions.

The trick is to train the entire thing from observation triplets (x(t),a(t),x(t+1)) while preventing the Encoder from collapsing to a trivial solution on which it ignores the input.

Auto-regressive generative models (such as LLMs) are a simplified special case in which
1. the Encoder is the identity function: h(t) = x(t),
2. the state is a window of past inputs
3. there is no action variable a(t)
4. x(t) is discrete
5. the Predictor computes a distribution over outcomes for x(t+1) and uses the latent z(t) to select one value from that distribution.
The equations reduce to:
s(t) = [x(t),x(t-1),...x(t-k)]
x(t+1) = Pred( s(t), z(t) )
There is no collapse issue in that case. "


Too technical? Then let's just consider another side of the picture.

As I wrote above, any imported model comes with its own "representation" (what it learned from reality), and just ignoring that and overlapping what you need does not necessarily imply that you will get the results that you expect.

Requires integrating both the exogenous element and your own context- but working on something more than a mere conceptual integration.



Integrating



If you shift to human systems, the point 2. within the previous definition is the key element:
"2. the state is a window of past inputs".

Whenever I was asked to work on cultural and organizational change, or to introduce new technologies to support decision-making (from my first late 1980s projects, not just Decision Support System models, but also models that integrated data- be it in procurement or at the general ledger level), eventually somebody talked about "best practices".

Routinely I heard across the years the concept, notably when it was part of a "change" initiative linked to compliance.

Anyway, after few smaller initiatives, when, in late December 1992, accepted to start from January 1993 to be part of a multi-year initiative to "change the way our people think and work", I considered my duty (and a perk of my job) to keep looking around to concepts not just as tools to implement, but also as frameworks of analysis, before deciding to adopt specific tools.

A change initiative with major impacts often starts with a kind of USSR-style Intourist tour offered by the consultants or suppliers, to have sanitized visits of prior customers or other cases.

I shared routinely how, whenever I have to focus on a new subject or expand on something, after an initial assessment to identify boundaries, I rely of my penchant for studying and understanding cultures (or attempts thereof) since I was a kid.

Hence, in that early 1990s case, as it was to be a potential multi-year venture, "integrating" implied not just keeping eyes open when going around Europe.

Implied also investing a bit on research.

While the first year was in part a continuation of previous concepts, it was a wider and evolving range where I continuously integrated elements from other sources, eventually decided to spend the summer vacations not on one of my usual projects (also as part of my future initiatives, that then had anyway to sideline).

So, beside books, in 1994 attended in London a summer school at London School of Economics on International Political Economy (specifically "States and Firms in the International Economy", held by Professor M. Hodges), followed by a summer academy in Gothenburg on Intercultural Communication and Management.

Consider that a change initiative, whatever its scope and content, affects also behavioral patterns within the organization.

Sometimes it is intentional, sometimes is a side-effect, but as I was reminded by some customers in the past, just the continuous interaction with a different perspective can have a transformative effect.

Integration implies having a pre-emptive framework to contextualize that continuous interaction.

In project and program management methodologies books, often you can read about the triad "complicated-complex-chaotic".

Well, any continuous interaction such as that described makes things at least complicated even in the steadier environment.

Reason? If you have an existing "state", introducing new information that relates to a different conceptual framework generates some overlapping and some reconsideration.

If you started already with a "complicated state", where different parts were at least occasionally diverging, the new additional elements could generate...

...an aggregation of "complicated" interacting, turning into a complex.

But if you already started with multiple actors that were carriers of different perspectives, you can risk generating a chaotic situation.

What differentiates, in this context, a complex from a chaotic context? A complex context requires at least an initial injection of facilitation elements, to keep the "complex" from escalating into "chaotic".

Chaotic, left by its own, generating goes the "resonance" way- i.e. keep increasing and getting out of control.

Generally, when the introduction of exogeneous elements to integrate into a complex set of existing organizational relationship could turn that into a chaotic situation, super-partes facilitation is useful to at least "reduce" or "compartmentalize" chaos, allowing to progressively reduce.

If you look at my CV, it is just a an outline and summary, a sampling.

And you might think that this is a case of "Cicero pro domo sua"- i.e. somebody advocating the use of its own services.

Nothing further from the truth.

Sometimes, I have been asked to work on getting from the right-hand side of the "complicated-complex-chaotic" and getting toward the left-hand side.

Anyway, as I said to my customers, the point was always for them to develop their own internal capabilities.

The wider the impact, the earlier I asked to have somebody involved internally who could take over eventually- sometimes even while negotiating the mission contract.



Threats&Choices



When I wrote above super-partes facilitation, I really was thinking to few characteristics:
_ should be familiar with the organizational culture they will intervene in- no "off-the-shelf, painting-by-numbers"
_ should be acceptable as super-partes facilitators by all those involved.

In my experience, organizational integration (or separation), even of temporary organizations (such as temporary associations built to deliver a single initiative) takes a phase-in and phase-out time, where potentially there will be increased demand on critical resources.

This is the area most critical to focus on support, as could be considered pivotal in delivering results.

Anyway, change initiatives with structural impacts, i.e. assuming to alter the organization(s) impacted, should look at that "complicated-complex-chaotic" element on a wider perspective.

Yes, it is again a function of having redefined your Weltanschauung- and that should be a continuous process as you get more knowledge through the integration part.

I received few days ago an invitation for an interesting two-days event in Trento, on foresight:

March 20th and 21st in Trento (Italy)

"20 e 21 MARZO 2024

Con l'obiettivo di stimolare il dialogo sul tema del foresight, Fondazione HIT, con Università di Trento e Fondazione Bruno Kessler, organizza l'evento dal titolo "Chasing Futures - Esplorazioni tra Foresight e Science Fiction: l'impatto dell'IA nel mondo che verrà", che si terrà a Trento i prossimi 20 e 21 marzo.

Futuri, fantascienza e intelligenza artificiale sono le parole chiave della conferenza, che prevede interventi di relatrici e relatori provenienti dalla ricerca, dall'innovazione e dalle istituzioni pubbliche, dal mondo letterario, cinematografico e del gaming, trattando tematiche chiave come salute, lavoro, sostenibilità, etica e rischi dell'intelligenza artificiale.

L'evento vede il patrocinio dell'Associazione Futuristi Italiani, dell'Italian Institute for the Future, di Trentino Film Commission e la partecipazione della Commissione Europea."


Two days (free event- will not be there, but others too might be interested - here on EventBrite).

You probably noticed that in the previous section on "integration" did not discuss AI at all, also if referenced again the previous section on "models".

The reason is simple: yes, also "integration" is data-intensive within a data-centric context, but I saw way too often how the way to recover a complex situation that started turning chaotic was not to expand on the data, but, as I said within the introductory section, to focus on the "how".

The "why" and "who" often in a complex or chaotic situation turns into shooting the messenger, i.e. exactly those who could help understand how to shift from "chaotic" to "complex", and even in few areas from "complex" to mere "complicated".

Note for my readers: if you work in a context where there are multiple parties involved, "complicated" should be normal.

Those two days probably could give you a multidisciplinary view of the issues, before you turn into what I discussed in previous articles, e.g. PEST or other frameworks of analysis that try to look beyond the boundaries of your own organization.

There might be existential threats and misleading choices, but 20/20 hindsight is the only way to avoid both of them 100% of the time.

In the real world, not just in business but also in society and politics, it is better to find the right balance between waiting too long and acting too early- without letting tools become the driver.



Be adaptive



Improving? Implies learning.

And learning implies... keeping track of positive and negative choices (as I kept saying at least since the early 1990s when I was teaching to customers on the subject of project management).

If I compare with the late 1980s, we have now not just data and tools, but also, more important, tools to interact and obtain feed-back.

We do not have just what I called in the 1990s "Infoglut", we have also a constant stream of new tools that you "must" use.

Whenever I look at my Linkedin stream, I do expect that those working on developing models, delivering data-centric services, etc would talk about the latest tool, model, etc.

Recently, quite often instead makes me think about the old joke about a famous economist having a ride in a taxi, and receiving investment advice from the taxi driver.

Then, running to do exactly the opposite, and escaping the 1929 crash.

Nothing against taxi drivers, but the point is that, if you keep following anything that it is trendy and neither you have use for it, neither access to resources or people to help you contextualize, the real question becomes...

... why do you have to waste time pretending to be trendy just for the trendy sake?

On Linkedin, it might make sense- again, here some famous meme comes to mind.

But in real life, maybe a better prioritization of your resources could be useful.

As with best practices, the key is adapting before adopting.

I shared in the past reports e.g. from McKinsey on the feed-back from companies about adoption of AI initiatives that were not producing the expect results.

Being adaptive implies, as I wrote under the "integration" section, having "antennas about reality outside always active.

In our data-centric future, AI implies also ubiquitous AI, i.e. close to where data is generated and consumed.

Pity that, despite all the "trendy" posts that I read daily on my Linkedin stream, really few cover the capabilities and potential of bringing AI where it can make a difference.

I wish Linkedin allowed to filter out by hashtag associated with other professional filters- would have to scroll past less and less useless posts about the latest trend... which, of course, since few months ago, is GenAI.

An interesting and readable overview of the possibility (obviously from a vendor perspective, but still informative) is a brochure from a supplier shared by Mouser, the German company where in 2018 purchased my "neural network on USB", Movidius from Intel.

The brochure (27 pages) is from the company Renesas, and has the title Bringing Intelligence to the Edge.

Anyway, whenever adding AI, I shared already in the past results from assessments on the usual parameters, as also the brochure states: ROI being the first obvious one.

Another article that I shared few days ago on Linkedin instead was focused just on this element, ROI, but applied to GenAI.

Worth sharing what I extracted from that article and shared on Linkedin:
"first, the bad news:
"by 2025, 90% of enterprise deployments of genAI will slow as costs exceed value, according to Gartner - and 30% of those projects will be abandoned after proof of concept (POC) due to poor data quality, inadequate risk controls, escalating costs, or unclear business value.

By 2028, more than 50% of enterprises that have built large language models (LLMs) from scratch will abandon their efforts due to costs, complexity and technical debt in their deployments, according to Gartner research."

The good news? Read the article (no, I am not the author)

https://lnkd.in/eNUQkKFS"


But this chart based on a study of actual cases could actually help to identify if your own case... makes sense for your own business:



Anyway, being "adaptive" is not just about piling up technologies, tools, concepts: it is about integrating them within your own business (or organization) "organically".

If you have time, you can also watch on YouTube the recording of an interesting webinar that attended on March 1st at data.europa.academy, with the title New business models for data driven services.

Because ROI is useful, IRR is useful, NPV is useful- but if add AI because it makes sense, you probably are aiming for something more than a mere financial measure.

If you look at the current giants of technology online, none had a plan which fitted within a spreadsheet to showcase where they would be now- and, actually, even Amazon and others consumed resources long before started generating profits, but at the same time were creating value-generating assets.

Without having its own need for IT infrastructure to serve its own customers, would Amazon Web Services ever convinced a customer?

It was Amazon itself that was a showcase of the potential and value of cloud-based services provided by... Amazon.

Yes, the sections of this article would require probably few thousand words more each to discuss the theme- but consider this first part of the article an introduction not just of the second part, but of other publications that will follow.

For now, closing this last section, this is the cycle represented by the conceptual framework within this article:



If you read so far, by now probably each one of those keywords is "framed" within another set of mental representations from this article, associated also with something from your own experience.

It is curious how technology now is delivering a kind of "layering" that extends the time allotted to each activity that starts with a simple interaction.

Many call this the "age of distraction", but it is more "the age of serendipity".

Trouble is: if all your activity turns into serendipity, then you get a Wunderkammer... through "emergence".

Which is exactly the issue that will be discussed in the second part of this article, using a "structural" case.

Have a nice week.