Skip to Main Content

From idea to impact: What it takes to deliver
AI at scale

By Tania L. Binder and Ricardo Jorge
September 2026
| 6 min read

Authors

Key insights

  • Most companies are no longer struggling to build AI pilots; the harder task is deciding what AI is for, where it will create value and how to align governance, talent, incentives and operating models behind that ambition.
  • The organisations that succeed will balance quick wins with deeper business reinvention. Across these Portuguese case studies, AI’s greatest potential lies not only in efficiency gains, but in reshaping products, services, decision-making and sources of competitive advantage.
  • Clear board and CEO sponsorship, strong data and governance foundations, focused use-case prioritisation, central expertise combined with business ownership, and investment in workforce capability will all turn experimentation into repeatable enterprise-wide impact.

When it comes to companies adopting AI, the days of purely focusing on pilots and proofs of concept have been and gone. The priority is now to turn experimentation into actual and repeatable impact. The main constraint, though, is no longer purely technical.

Access to powerful models, cloud infrastructure and a growing ecosystem of tools are of course still important, but the more difficult challenge is organisational. Where should we place the biggest bets? What is the right level of governance? How can we align talent, incentives and business priorities? Such questions are ricocheting across borders, sectors and industries — and finding the right answers is by no means straightforward.

Portugal offers a useful prism through which to view these developments. Drawing on interviews with leaders from four Portuguese companies across banking, telecoms and satellite intelligence, this article explores how leaders are grappling with moving beyond experimentation, balancing speed with control and capturing lasting advantage. We would like to thank all of them for taking time out of their busy schedules to share their reflections and insights.

The sectors may differ, but the underlying issue is how to turn AI from a set of promising tools into a source of repeatable business impact.

  • AI success depends less on the technology itself and more on leadership clarity about where it will create value.
  • Impressive pilots are one thing, scaling AI through deliberate decisions on operating model, governance and talent is quite another.

It’s easy to be dazzled by the newest AI models but leaders must beware of being distracted by the siren call of the latest gleaming advance. Instead, they should take a step back and view AI as a set of leadership decisions rooted in where it will create value and how it could benefit their business.

The priority is not simply deploying AI, but rather making clearer choices about the operating model, governance and talent. This requires senior teams to align around a shared ambition, boards to balance innovation and risk, and organisations to be brimming with talent that can work across functions, cut through complexity and sustain execution.

Those businesses that don’t have these traits may still launch impressive pilots, but converting them into stronger performance over the long term will be far more challenging.

  • NOS is using a model-agnostic AI approach to avoid lock-in and preserve strategic flexibility.
  • Its leaders must decide whether AI will optimise the business or help reinvent it for long-term advantage.

Executives at the telecommunications and media company, NOS, are focusing on how to build at scale without becoming dependent on a single frontier model. To preserve the company’s strategic flexibility, NOS is investing in a model-agnostic approach that keeps its options open across providers and reduces the risk of being locked into one approach.

The core leadership challenge is that AI is forcing the executive team to lead a business transformation, not just a technology programme. They need to balance speed with safety, central control with innovation, and short-term gains with long-term strategic advantage — a set of challenges that must surely sound familiar to C-suite executives in businesses around the world.

For NOS, though, the executive team must decide whether AI will be used to optimise the current business or to fundamentally reposition the company in ways that preserve control, distinctiveness and long-term value. This is ultimately a board-level choice about what the company must build itself, what it can buy, where it is willing to accept dependency and how it will prevent value from being captured by external model providers.

  • Banco BPI sees AI as a leadership and change challenge, not just one rooted in technology.
  • The bank must choose between targeted efficiency gains or wider business transformation.

Banco BPI’s AI journey has been shaped by a sense of fatigue following years of digital transformation. But while large organisations are no stranger to such programmes, the bank’s executives are no longer debating whether AI matters, but how boldly to use it.

GenAI is different from past technological reboots because it expands access, changes ways of working and has visible productivity potential across coding, customer support and natural language interaction. The bank’s leaders frame AI not as a technology challenge, but as a people and leadership one, with change management, adoption, skills development and responsible risk management emerging as critical barriers.

In banking, digital platforms raise the bar for customer experiences and lead to rising expectations about what can be expected from financial services. Banco BPI is therefore considering whether to pursue a selective, efficiency-led approach focused on targeted use cases, or a more ambitious transformation that redesigns roles, workflows and decision-making, which would require greater investment in reskilling and leadership capability.

Against this backdrop, critical thinking, problem solving and adaptability become more valuable as routine tasks are automated, all of which increases the demand for leaders who can integrate technology, people and judgement.

  • CGD built early AI foundations through cloud, governance and strong board backing.
  • It scaled through central expertise, business accountability and tight use-case focus.

Caixa Geral de Depósitos (CGD) built its AI capability early, strengthened data governance, moved key infrastructure to the cloud and prioritised AI oversight.

The bank’s leaders treated AI as a long-term business priority rather than a side initiative and this approach started from the top. The executive board created a clear vision, and linked AI to the organisation’s sustainability and competitiveness over the medium to long term. Just as importantly, they balanced ambition with control, embedding responsible AI, model governance and data quality processes from the start.

A second leadership choice was to centralise scarce expertise while pushing accountability for adoption into the business. A central AI team built knowledge, standards and culture, while project managers were expected to deliver adoption and business KPIs. Leaders also prioritised training at scale, including top management education, hackathons and practical sessions to improve AI fluency across the organisation.

Finally, they were disciplined in choosing which projects merited time and investment: only those that could deliver meaningful impact with a rapid time to market were prioritised. The result was a more focused AI programme that was tied to customer journeys, measurable outcomes and strategic value.

  • Geosat is weighing how AI can accelerate the move into higher-value intelligence services.
  • That shift offers growth potential, but requires new capabilities, a different operating model and careful choices about trust and strategy.

The central question facing Geosat’s leadership team is how can AI reposition the business as an intelligence platform delivering faster, more actionable insight, while maintaining the reliability required to support operations in the most demanding scenarios.

The company operates in a mission-critical environment with zero margin for error, so it is understandable why AI has so far been deployed with caution. AI is being used to accelerate learning processes for feature identification and change detection algorithms, as well as to gain operational efficiencies. Yet the potential is far broader. That’s because AI opened the possibility for Geosat to deliver new analytical tools providing near-real-time interpretation, scenario analysis and decision support. As customers place greater value on speed, context and usable intelligence, the opportunity for growth is clear.

However, such a shift is no small feat. Adding AI-based intelligence services to the product portfolio requires new capabilities (hybrid agent-people teams), an enhanced operating model and closer engagement with existing and future customers. The latter is critical for Geosat to remain closely aligned with the geopolitical interests of the countries in which it operates, namely Portugal, Europe and NATO, which shapes both product and commercial decisions by helping the organisation determine what not to pursue — an essential discipline.

Their example helps illustrate that leadership decisions about AI often come down to how to redefine the business without compromising the trust their customers already place in it.

  • Leaders must navigate key trade-offs on speed, ownership, capability and ambition as they scale AI.
  • Long-term value stems from choices that balance quick wins with deeper business reinvention.

Sectors and industries may differ, but similar leadership tensions keep reappearing. Leaders are not making technical choices alone, but mulling strategic trade-offs that shape how quickly AI scales, where value resides and whether early momentum can turn into sustainable advantage.

  • Quick wins vs structural change. Early gains can build confidence, but too much focus on fast returns may leave the core operating model untouched and long-term value falling short of expectations.
  • Centralisation vs business ownership. Central teams provide standards and governance, but value is created only when business units have ownership of their adoption and outcomes.
  • Build vs buy. Few organisations should build every layer themselves. The key question is which capabilities are strategic enough to own.
  • Optimisation vs reinvention. Many use cases improve efficiency, but the bigger growth opportunity may be to rethink products, services and decision-making.
  • Model choice vs flexibility. Leaders must decide whether to stay model-agnostic or commit to a single frontier lab, and whether to rely on open-weight models or closed alternatives.

Lessons from the frontline

  • AI successfully scales when leaders treat it as a business design question, not just a technology deployment.
  • The winners will be those with a clear AI purpose and the discipline to turn it into enterprise-wide change.

What stands out from these examples from Portugal is not simply that companies are investing in AI, but the significance of treating it as a question of business design rather than digital capability alone. Across banking, telecoms, aerospace and satellite intelligence, AI scales when leaders make explicit choices about where value will come from, how risk will be governed, which capabilities must be built internally, and how the operating model needs to change.

AI is no longer simply yet another wave of digital transformation. It is increasingly a test of whether leaders can move beyond experimentation, confront the trade-offs early and decide how AI should reshape value creation, workforce capability and competitive advantage.

The organisations most likely to pull ahead will be those with the clearest understanding of what AI is actually for, and the discipline and capability to translate that vision into cross-enterprise change.

Ten leadership decisions that shape whether AI scales

  1. Use AI to redesign the operating model, not simply to improve what is currently used. The biggest potential gains tend to occur when leaders use AI to redesign operating models, customer experiences and sources of value, rather than limiting it to standalone productivity tools.
  2. Keep AI firmly on the board and CEO agenda. Scaling AI requires visible sponsorship from the top, clear prioritisation, and executive alignment on trade-offs involving risk, investment, talent and business-model change.
  3. Prioritise the use cases that can change the business — inside and out. Leaders should distinguish between quick wins and strategic opportunities, and focus on the use cases most closely linked to revenue growth, cost savings, customer outcomes or competitive advantage.
  4. Build the foundations for scale before trying to accelerate. Data quality, cloud infrastructure, model governance, cyber security and responsible AI frameworks underpin what makes enterprise-wide deployment possible in practice.
  5. Design the operating model deliberately, with clear ownership at the centre and in the business. Central teams can provide standards, architecture and scarce expertise, but business units must retain ownership of adoption, outcomes and frontline relevance.
  6. Treat governance as a strategic enabler from the outset. In regulated or high-consequence environments, explainability, human oversight, model inventory and clear accountability are not compliance afterthoughts; they are part of what makes AI scalable.
  7. Invest in leadership and workforce capability as seriously as in the technology itself. Boards, executives and employees all need greater AI fluency, while organisations must prepare for changes in roles, workflows, apprenticeship models and future skills.
  8. Own the capabilities that will remain distinctive, and buy the rest with discipline. Leaders need to decide where proprietary capability matters, where external platforms are sufficient, and how to avoid unnecessary dependence on any single vendor or model provider.
  9. Measure AI in business terms, not activity metrics. AI programmes should be tied to concrete outcomes such as time to market, customer satisfaction, digital sales, cost savings, resilience or process performance.
  10. Look beyond optimisation to the opportunities for reinvention. The organisations most likely to lead in the next phase of AI adoption are those willing to use it not only to improve current work, but to rethink products, services, decision-making and what makes the business distinctive.

Contributors

Our thanks to:

Jorge Graça: executive board member and CTIO, NOS Comunicações

Afonso Fuzeta Eça: executive board member and COO, Banco BPI

Francisco Vilhena da Cunha: CEO, GEOSAT

Madalena Talone: executive board member and COO, Caixa Geral de Depósitos

Authors