AI e KPIs

Na semana passada estava numa reunião atolado numa discussão sobre um indicador de desempenho para um processo. Abstraí-me da reunião por uns momentos, copiei o texto da finalidade do processo e perguntei ao ChatGPT que indicadores me sugeria. Ofereceu-me doze indicadores. Escolhi três e apresentei-os aos participantes na reunião. 

A discussão voltou a engrenar e dois dos três indicadores foram revistos e aprovados.

Entretanto, ontem de manhã cedo encontrei este tweet:

According to new @mitsmr research, organizations that use AI to improve existing KPIs or create new ones realize more business benefits than organizations that adjust their KPIs without AI.

The research identified three ways to enhance strategic metrics with AI:
1) Improve… pic.twitter.com/L7ToP88jJ9

— Brian Solis (@briansolis) March 18, 2024

Eis as minhas notas de The Future of Strategic Measurement: Enhancing KPIs With Al:

"Executives who recognize the power and potential of Al-enabled KPIs acknowledge that their current KPI design and review practices are anachronistic. More and more companies are rising to this challenge by embedding governance processes and functions to oversee KPI development. KPI governance mechanisms that maintain

...

A common corporate objective is to improve KPIs over time. Our research shows that using Al is one of the most effective ways to enhance KPIs.

...

Six out of 10 survey respondents agree that improving their KPIs (in terms of what gets measured and how it gets measured), not just improving performance, is critical for effective decision-making. Every executive we interviewed explicitly declared that improving KPIs with Al is an enterprise interest, and it is an immediate priority for several.

...

We see three ways that Al-enriched KPIs improve on legacy metrics that simply track performance. Smart KPIs better describe ongoing and past performance, more effectively anticipate future performance, and make more useful recommendations to promote outcomes. These three types of smart KPIs (descriptive, predictive, and prescriptive) map to a well-known distinction between descriptive, predictive, and prescriptive analytics.

...

Identifying strategic and operational KPIs is a common function of human leadership teams. Yet our research confirms that Al - not only humans - can propose and even design new and novel KPIs to guide executive decisionmaking. We heard story after story of leaders turning to algorithms to source key performance metrics that are not only strategically and operationally valuable but, in some cases, derived from data patterns too complicated for humans to discern."

"humans with Al will replace humans without Al"

Há cerca de um mês e meio em E na sua empresa, como está a ser aproveitada a AI? recordava uma frase de Kasparov:

"Algures li uma afirmação de Kasparov que dizia que um grande mestre perdia sempre contra um supercomputador, mas que um jogador mediano com um computador mediano, juntos batiam sempre o supercomputador."

Agora em "Training with Al: Evidence from chess computers" encontro:

"Using the case study of chess computers, we examine the linkage between Al-backed simulation and learning strategic interactions. It is plausible that Al will democratize opportunities to learn strategic-and potentially more broadly social-interactions. If so, Al may become a game changer well beyond the game of chess. While it has often been asked whether Al will replace humans, our study suggests that a more imminent change may be that humans with Al will replace humans without Al."

Again, como é que na sua empresa está a ser testada, usada, aproveitada a AI? 

E na sua empresa, como está a ser aproveitada a AI? (parte III)

Parte II.

"Our case studies, based on our growing global community of over 3,000 GenAI practitioners, point to a new category of work, more precise and actionable than “knowledge work.” We call it WINS Work: the places where tasks, functions, possibly your entire company or industry are dependent on the manipulation and interpretation of Words, Images, Numbers, and Sounds (WINS). Heart surgeons and chefs are knowledge workers but not WINS workers. Software programmers, accountants, and marketing professionals are WINS workers.

GenAI has the potential to be power tools for WINS work. It can generate new prose, computer code, images, narration, music, and videos as well as ingest and summarize, critique, improve, and reformat almost any manner of document or analysis. Every WINS task, subprocess, and end-to-end process within your enterprise (and in many cases the entire enterprise) should be evaluated for potential leverage with GenAI."

Na semana passada estava numa reunião com alguém de uma empresa metalomecânica e discutiam-se objectivos para o sistema de gestão ambiental. 

 - Como se pode reduzir o consumo de óleo por peça produzida? 

Eu, um nabo em metalomecânica, perante o plissar da reunião, fui ao chatGPT e fiz uma pergunta sobre que variáveis influenciam o consumo de óleo numa "stamping operation". E logo recebi uma lista de 10 variáveis que faziam mesmo sentido. E não sou um WINS worker.

Recomendo a leitura de "Where Should Your Company Start with GenAl?" Segundo os autores, as empresas fortemente dependentes do trabalho do WINS devem agir agora para evitar uma concorrência mais acirrada e concorrentes disruptivos dentro de 36 a 60 meses, evitando custos elevados, processos desactualizados, desvantagem de dados, perda de talentos e capital mais caro.

Julgo que os autores não deviam limitar o conselho ao mundo WINS. Ainda na passada segunda-feira o meu parceiro das conversas oxigenadoras contou-me um caso da sua empresa. A empresa tinha uma dificuldade, contactou o fornecedor de há várias anos, e obteve uma resposta da treta. Alguém pesquisou o tema com o chatGPT, obteve uma solução para o problema e arranjou a morada de dois fornecedores alternativos. A empresa do meu parceiro chamou o fornecedor tradicional e explicou-lhe a situação, e disse-lhe qualquer coisa como:

- Se vocês não evoluem, põem-nos a nós clientes em desvantagem.

Acham que um vendedor de componentes assemblados para máquinas industriais é um trabalhador WINS?

O velho engº Matsumoto tinha horror ao que ele chamava "catalog engineers", os vendedores que só sabem recitar o que vem nos catálogos da empresa.

E na sua empresa, como está a ser aproveitada a AI? (parte II)

Parte I.

Consultei o 2023 Tech Trends Report Executive Summary e fiquei fascinado com esta tabela e as previsões sobre a relevância da AI já no curto-prazo:

"More and more leaders see AI as necessary for growth in the current macroeconomic environment, even as new developments make some job categories obsolete. In nearly every industry, AI will serve as a force multiplier for growth, bringing efficiencies, better tracking, business intelligence and assistance with decision-making. As training costs decline, more applications will be built. Spending on AI systems and hardware is likely to explode this decade, creating significant enterprise value.

...

HOW TO PREPARE: We recommend that chief strategy officers in every field develop a solid understanding of AI to engage more closely with the C-suite to develop a cohesive point of view on digitalization, augmentation and automation-and develop strategic plans. Importantly, businesses should keep abreast of emerging regulations that could restrict the use of consumer data, algorithms or generative AI systems. Risk models should be developed to determine plausible near-futures, so that leaders can adjust their strategies accordingly."

Trechos retirados de "The Tech Trends Every Leader Needs to Understand" de Amy Webb na revista Rotman deste Outono.

E na sua empresa, como está a ser aproveitada a AI?

Na semana passada, talvez na quinta-feira, já não tenho a certeza, o chatGPT esteve inoperacional. Por curiosidade fui ver se aparecia nas tendências do Twitter em português... e nada.

Sintomático...

Sempre que penso em AI penso nas palavras de Kasparov:

"Algures li uma afirmação de Kasparov que dizia que um grande mestre perdia sempre contra um supercomputador, mas que um jogador mediano com um computador mediano, juntos batiam sempre o supercomputador."

Na revista Rotman deste Outono em "ChatGPT: What Leaders Need to Know" encontro: 

"Curt Nickisch: In late 2022, people started talking about how ChatGPT was poised to change work as we know it. What is so special about this technology?

Ethan Mollick: ChatGPT-3, which is the base generation of this technology, was around for about a year, and it was okay. It worked at or near the level of a D-minus student. But we have since crossed a threshold of ability that has made ChatGPT incredibly useful. It can now achieve the output of a B-minus student. And if you use a more advanced model like the one on Microsoft's Bing search engine, you can achieve the level of an A-minus student. This represents radical progress in a very short period of time.

...

CN: What kind of opportunities does this open up for companies?

EM: There is no rulebook or manual to work from, so you have to explore for opportunities yourself. As indicated, this AI is really good at writing computer code and common documents like memos or reports. If you need to write a letter of recommendation or a performance review, it will do an amazing job. But when you start to incorporate the Bing version or some of the other new AIs coming our way, you'll be able to get it to do complex, multi-layered tasks. For example, you'll be able to say, 'Look up everything about the precision agriculture industry and give me a table showing the growth rates for each company along with their strengths and weaknesses, using the perspective of Michael Porter's Five Forces Model.' This type of request could save someone up to 12 hours of work and provide a document that they can build on - because as good as it is, you do still have to check Al's work."

A verdade é que entre Maio e Agosto últimos estive envolvido num projecto usando AI e foi impressionante o ganho em produtividade.

E na sua empresa, como está a ser aproveitada a AI?

Ainda na mesma Rotman deste Outono em "ChatGPT: What Leaders Need to Know":

"SP: Gillian, Do you think this technological moment is different from previous ones?

GH: I do. What is critical here is the speed with which AI transforms entire markets. For example, in the legal domain, I've seen how, in just a few minutes, tools like ChatGPT can do what it would usually take a week for a lawyer to do. That is going to be very disruptive, and the potential scale is massive. Because it's a general purpose technology, it can and will show up everywhere."

Ainda na mesma Rotman deste Outono em "ChatGPT: What Leaders Need to Know":

"You can't call up a consulting company and ask them how to implement ChatGPT across your company, like you would with Salesforce or some other system application. Over time, there will be more organizational applications, but so far, the main story is at the individual level. And people are using it. I guarantee you, there are secret users of ChatGPT in every organization who have cut their workload by two thirds and don't want to tell anyone about it. I'm hearing examples all the time. It's easy to get started; you just need to spend an hour or two learning how it works and what it's good and bad at.

I've talked to tons of people who are big fans. First of all, virtually everyone in programming is using it to help with coding. They've switched over from Stack Exchange to ChatGPT. But increasingly I'm talking to managers who have just automated tasks like performance reviews, grant applications or regulatory compliance. This is a general-purpose technology, as its name suggests."