Successful Use of AI Tools: "It’s Not Magic, It’s Management"
Interview with Rodrigo Helcer, founder of STILINGUE (acquired by Blip.ai)
In our latest report, Amrop’s Global Digital Practice examined the leadership competencies essential for successfully integrating AI into organizations. We invited CEO/GMs from midsize, PE-backed, family-owned and other companies, to share their real experience in leveraging AI strategies for their organization and customers.
Paulo Aziz Nader, a Amrop Partner at Amrop Brazil, leads Executive Search projects in digital transformation contexts, working for clients across Technology Services, Financial Services and Venture Capital / Private Equity invested companies. He is also an entrepreneur, angel investor and Board Member at several promising Brazilian start/scale-ups.
Paulo spoke to Rodrigo Helcer, the founder of STILINGUE, a Brazilian AI-powered platform for social media monitoring and customer service.
Paulo Aziz Nader: To start off, what do you envision for the future of AI tools, and do you see differences across industries? What effects do you think AI will have, both positive and negative?
Rodrigo Helcer: I have worked with AI since 2014 when we started our company here in Brazil. When you talk about AI it’s like talking about the field of medicine – you have many different types of expertise and specialization. We came from the area of Computer Linguistics and Natural Language Processing, which, coincidentally, are at the root of Generative AI and current technology, which is making all these conversations around the world take place. What changed significantly in a practical way is that the work and the effort to train a model, the effort to teach a machine to do something, changed massively. There’s “machine learning”, but I’d like to talk a bit also about “machine teaching” – because for the machine to learn you need to teach it. The effort it takes today boils down almost to merely writing a prompt – and that makes room for an explosion of use cases.
Paulo Aziz Nader: Could you share your observations from the past couple of years in this space and your overall perspective on AI in organizations?
Rodrigo Helcer: There were things we were able to do in the past but not as fast – and not as cheap as we have access to today. We used to supervise the learning and had to give several examples and information to the models we were training; nowadays we call it “zero shot” learning. Another large practical change relates to access to this technology. Five years you needed to have a team of scholars and software engineers to develop AI. Today you pay a few dollars and it’s done. There’s a trend that emerged side by side with AI - the low-code/no-code trend; most software developers realized that we should be making code for non-software people – so how can I make code to use or customize which you don’t need to know how to code? So, today we are seeing a boom in tools which enable people to create AI models, low-code/no-code. In simpler use cases you may not even need to have that team of scientists and engineers anymore – this team is now tackling much harder problems – ones in the foundation and the structure of software, and no longer at the software application level.
Paulo Aziz Nader: And what about the evolution of companies, which need to catch up too? What have they done to adapt and how do you see them evolving in the future?
Rodrigo Helcer: Today we are entering a new wave of Generative AI, while 2023 was the year of experimentation. Companies which grasped this opportunity started to experiment, to test & understand possible use cases of this technology. Now we are seeing some companies, which were lagging last year, starting to do it this year. However, this second half of 2023 marks a new stage of applications - of real, practical use cases in scale.
I think in this last half of 2024 and entering 2025 we can expect a real boom in use cases – new companies, new software, new features. A software product, even when it is done in a very, very fast and lean way, takes six months to a year to develop – so the new products and features derived from Gen AI are expected to tsunami the market from now on.
Paulo Aziz Nader: The question which gets asked a lot concerns AI’s impact on jobs. Do you see any developments regarding this too?
Rodrigo Helcer: The concern is definitely there but in my opinion it’s a misconception because the impact of AI is not on jobs, but on skills (and its derivation of tasks). Our worry or, better, opportunity should be around the automation of what I call "the mono challenge" – because monotony, skills and tasks that are monotonous, that don’t give us any joy, are the ones that will be given to machines, automated. Also, "mono" for jobs that are "mono skill", with a glass ceiling in one of few tasks. For example, if you work with translation, and your only job is to translate, you will lose out to the machine. But if you, on the other hand, work with diplomats, and you need to add sensibility to the translation, the tone of voice and so on, the technology will only make you stronger, it will become your co-pilot. It will automate the task that didn’t exactly have that main value that we maybe thought it had. And augment your job. So, to identify opportunities and risks, we should not look at jobs but at the skill level. Another “buzzword”, which comes up a lot when we talk about "skills" is “upskilling”. There is an urge to upskill, which all companies are talking about. We have seen that digital savvy has not permeated through the whole companies, and that’s not negotiable anymore. There’s a phrase that characterizes it: machines will not substitute men, but men with machines will substitute men. So, if you don’t, as a manager, invest in upskilling yourself and understanding this new toolbox you have, and how to use it, you will be obliterated. But if you approach it seriously, you will be stronger. The problem when it comes to technology is not with top or bottom level of talent – it’s with the mid-level. The top talent will gain much more power and resources to do what they already do best even better. The lowest-skilled talent will also benefit from more access and ease of use with technology – it will bring them closer to the average. And the people who possess average skill levels will now have a problem competing with them.
So, what remains as one of the most important skills in this new configuration of man and machine, is one’s capacity of directing – giving direction. You need to see the desired end-product and know how to get there using the right tools – the abilities of both man and machine. Without the skill of directing, you’re only executing.
Paulo Aziz Nader: What mistakes do people/companies tend to make when it comes to introducing technology and using AI?
Rodrigo Helcer: There’s a classic mistake working with AI generally. There’s a case from some years back where some operators automated stock exchange trade operations using data that came from Twitter. A hacker hacked the associated press profile and said that there was a bomb in the White House. Because of that, the algorithmic trader lost billions in minutes, creating a huge mess. Who is responsible for that? It’s not the algorithm because it’s not a person – it’s code. The person responsible is the one who delegated the job to it and didn’t supervise it. That’s the kind of thing that happens when you over-trust, lack responsible supervision and don’t put the right guardrails to avoid mistakes. You don’t delegate a make-or-break decision to your student intern. We sometimes seem to be confused – to think that magic exists. But it’s not magic, it’s management. We’re sometimes too enthusiastic about giving the power to decide on a group of codes, which is something we shouldn’t do. There’s a school of thought when it comes to AI where the belief is that the future is not one of robots – but instead a future of augmented intelligence. Kasparov didn’t lose a chess match to a machine but to a machine with dozens of scientists behind it training it. He then went on to create a type of freestyle chess where a human and a machine plays against another human and a machine. I believe in a future of men and machines as co-pilots rather than autopilots. And all autopilots should come from co-pilots which have been tested so extensively that they can then be trusted. We should be really careful and responsible, skipping this step and jumping right to autopilot.
Paulo Aziz Nader: With this co-pilot augmentation, do you think that there would be changes in organizational structure, or will it just be an enhanced version of what we have now?
Rodrigo Helcer: I’ve not thought about it in terms of organizational structure yet, but I’ve thought about it in terms of size. The first thing that comes to mind when talking about automation is always about cutting costs, reducing the size of the team. But it might not be true and depends on the market you’re addressing. If you have a market with a large total addressable market that can take the growth of your output which you now generate with the same size of the team, you, of course, do it, rather than reduce. Today product roadmaps have lots and lots of parked ideas that are waiting to be developed – when you bring AI to coding, you can gain between 30% to three times the productivity of a software engineer. If you have an amazing product roadmap backlog you will keep your talent. If you’re in a saturated market and in competition for margins, you’ll have job cuts. It depends on the industry and market size.
Paulo Aziz Nader: We often talk to boards and CEOs about hiring leaders. There’s now lots of talk among board members about who should be the next leaders in the age of AI. How would you advise them? What will make a good CEO and also good C-level leaders for the AI age?
Rodrigo Helcer: The first step, which is very simple but not obvious, is for boards and CEOs to learn and understand this new toolbox – it’s very important to be tuned in and updated with regards to the practical use cases of these new tools. The challenge is that it’s all changing constantly, but we need to create discipline around it, and maybe dedicate a team or expert from outside who can help you keep up with these developments. To be able to take risks, you need entrepreneurial skills in your group of executives and board. Maybe this technology will create your next competitor, so you should be doing it internally too. So, there’s a set of entrepreneurial skills that should be considered in the board composition and C-level recruiting because companies will need that. Usually, tech entrepreneurs are skilled product developers, they have the product vision, and they know how to get there, build a team, put all the pieces and people together. And what companies perhaps need right now is more freedom of thought of how to get there, without so much constraint from bureaucracy which one usually faces within a large organization. Boards in general are tending more for compliance, governance and liabilities than entrepreneurial opportunities with this news technology. Indeed, this is an important agenda to address in the strategic and governance discussions.
But, in my factual optimism - participating with a couple of teams building AI and their success so far - there are enormous opportunities sectors wide, functions wide, that suggest that discussions of "acceleration" should overcome the "caution" discussions. Tech Bubble? This is a matter of lenses and creates fear where it should not exist. Are we with the lenses of stock investors or value creators? If the latter, yes, we might be in a bubble, it is expected (as happened with the internet) to happen as part of every shift of tech paradigm where ventures invest much more than necessary to win the race. But with the lenses of value creators, we shall strike while the iron is hot. While the internet investment bubble burst, its crescendo of value creation never stopped.
Paulo Aziz Nader: Do you think that companies perhaps need a specifically assigned CAIO – Chief AI Officer to drive the initiatives and provide others with the direction when it comes to AI?
Rodrigo Helcer: I think a team needs to have three core abilities: how to architect it, how to engineer it, and how to value it. Today in one of the companies I advise we have the Head of AI and he’s an expert engineer. But that’s not enough – there also has to be an architect who understands the work to be done for the client, the large enterprise, and the processes that, if you tweak with new technology, you create value. He’s not an engineer, he's a product guy.
Paulo Aziz Nader: So, it probably cannot be one person, right?
Rodrigo Helcer: Most likely not. I find that it’s very, very rare that one person encompasses these three pillars, maybe two. In one company we have the engineer, the business versus the market development person, and the person who knows how to go from the product/marketing part to the market and sales management, distribution to make the value really come to the company. That’s the structure we’ve lived with for the last 10 years. I myself can easily navigate the product part and the sales part, but don’t ask me to code. But I can speak to the engineers, and I can build a team of engineers. If I think about our current Head of AI who is an engineer – me as a mentor and our whole team, we help him architect the roadmap in order to build.
There are examples of companies which have top-skilled AI engineers, but the companies haven’t grown – they became service companies which were not able to build a product. What happened? They didn’t have the product and the going to market components – AI scientists are usually not product and businesspeople. If I could choose, I would have an AI person specializing in the product, because a good product director knows how to make money and engage the team specialists to get there! But the talent pool today, I believe, is more biased towards engineers than to product people and businesspeople, meaning that people are trying to catch up with the new toolbox through the world of science.
Paulo Aziz Nader: So, you think it might be a misconception that it’s the engineers that will bring AI to life?
Rodrigo Helcer: It depends on the arena you’re competing in. There are the hardware companies which are the arena for the engineers and there’s another arena which is around data and the right type of data you need to create what’s in the intersection of those two – and that’s the foundational models, which is an edge-science play, engineers’ play. But there are two more arenas – one that is making platforms and software for builders and developers; that helps you code and train models. This area is concerned mainly with product, though engineering skills are, of course, still necessary. And finally, you have the application arena – the vertical SaaS, which deals with different kinds of automation, and I see it as a product and business play. The foundational models don’t need to be built in-house; you just need to understand how to use them, there's a growing plethora of low or no code solutions and the race is for enduring value creation.
Paulo Aziz Nader: And that’s why the Board and leadership team’s ability to understand AI is so relevant, right? They need to be able to know which arena they’re playing in…
Rodrigo Helcer: Yes, and it shows clearly that AI is a very broad concept, and we need to be careful with wanting to have black and white answers here. It needs to be contextualized. It is a sophisticated discussion of structure and impact. Mostly the answer to all these questions is – it depends.
Paulo Aziz Nader: Do you have any other comments?
Rodrigo Helcer: There’s just one insight, which I think is relevant in the people directors of today. We talk a lot about carbon footprint, our responsibility around balance and compensation.
And how about AI and the transitory "deforestation" of jobs? We must think and act similarly for compensation for AI in the context of jobs that the AI will substitute. We are always going after the opportunity and the cashflow that this opportunity can potentially generate, but we need to also be responsible with regard to the impact that these decisions will have when it comes to our employees. How can we compensate? It can be education, upskilling… This might not yet be on many agendas, but the problem will become more and more visible, as more gets automated with the use of AI.
Paulo Aziz Nader: Leaders should be looking at that too.
Rodrigo Helcer: Yes, I see it as a corporate responsibility – we will need to manage our AI job-replacement footprint.
Key Takeaways
Find out about Amrop’s technology and digital expertise, methods and tools, by contacting Paulo Aziz Nader or the Amrop Digital Practice members in your country!
Our full report "Digitization on Boards 7th Edition" is available for download here.