Tag: Business

GaaS for CAS

For most of the industrial era, organisations could afford to treat their operating environment as a complicated system, which is to say, one that is difficult but ultimately knowable. Given sufficient expertise and sufficient data, the right answer could be found. Strategy was a matter of analysis followed by execution, and the gap between plan and reality was a problem of implementation, not of epistemology. That assumption no longer holds. The ecosystems in which modern organisations operate, whether commercial, regulatory, technological, or geopolitical, have become genuinely complex rather than merely complicated. In a Complex Adaptive System (CAS), cause and effect are separated in time and space; feedback loops are non-linear and often counterintuitive; actors adapt continuously to one another’s behaviour; and emergence, i.e., the appearance of fundamentally new system behaviour arising from interactions rather than components, cannot be predicted from any analysis of the parts. This is not a temporary condition of turbulence to be waited out. It is the permanent operating environment of our epoch.

The deeper problem is not that complexity exists, but that human cognition is structurally ill-equipped to navigate it. Our minds evolved for causal, local, sequential reasoning. We are skilled at identifying patterns across two or three variables, constructing linear narratives, and responding to threats that are proximate and concrete. We are far less skilled at processing non-linear correlations across hundreds of variables simultaneously; detecting the faint early signals of systemic change buried inside overwhelming noise; maintaining analytical consistency across months and thousands of documents without the accumulation of cognitive bias; or tracking the co-evolutionary dynamic in which every strategic action modifies the very landscape against which the next action must be taken. Ashby’s Law of Requisite Variety captures the diagnosis with precision: a system can only be governed effectively by an entity whose variety is at least equal to the variety of the system being governed. No human team, however talented, possesses requisite variety to govern a genuinely complex ecosystem. This is not a failure of intelligence or effort; it is a structural mismatch between the architecture of human cognition and the architecture of the environment.

It is against this background that aGentic AI as a Service (GaaS) represents something more significant than the latest iteration in a long line of enterprise software innovations. The evolution from software-as-product to Software as a Service (SaaS) was a revolution in distribution: the same work, performed by the same humans, accessed more conveniently. The evolution to GaaS is a revolution in the unit of economic value itself. Where SaaS delivered tools, GaaS delivers outcomes. Where SaaS automated tasks, GaaS automates the cognitive labour that surrounds tasks: planning, deciding, monitoring, adapting. An autonomous agent does not wait to be invoked; it pursues a defined objective by decomposing it into sub-goals, selecting and operating tools, evaluating results against intent, and recalibrating continuously as conditions evolve. The commercial corollary is equally significant: pricing shifts from per-seat subscription, which is payment for access, to per-outcome, which is payment for results, transferring value-alignment from the vendor’s interface team to the vendor’s delivery architecture.

The most important distinction in the GaaS landscape, however, is not between GaaS and SaaS. It is between two fundamentally different conceptions of what GaaS is for. The dominant conception in the current market is GaaS as efficiency infrastructure: autonomous agents that execute well-defined processes faster, more consistently, and at lower cost than human teams. The customer service agent that resolved two-thirds of Klarna’s support interactions, the procurement agent that compressed approval timelines by thirty per cent, the claims-processing agent that reduced manual review by forty per cent, all of these are genuine and substantial achievements, but they share a common assumption: that the problem to be solved is already known, the domain of relevance is already defined, and the criteria for success can be specified in advance. These assumptions are reasonable in stable, well-bounded operational contexts. They become progressively less defensible as operating environments become more complex, more adaptive, and more non-linear.

GaaS designed for genuinely complex environments requires a different architecture and a different epistemology. Rather than executing against known objectives in bounded domains, it must operate across three interlocking layers of capability.

  • Ecosystem sensing layer: a network of specialised agents that scans heterogeneous data streams, including structured and unstructured, quantitative and qualitative, internal and external, to surface weak signals, non-obvious correlations, and emerging patterns before they become legible to conventional analysis.
  • Sensemaking layer: agents that translate raw signal into strategic narrative without collapsing the complexity that makes the signal valuable, serving as the interface between machine-scale pattern recognition and human-scale judgment.
  • Adaptive execution layer: agents that implement strategy not as a fixed plan but as a persistent intent, continuously monitoring how each action reshapes the environment and recalibrating tactics in response, so that the gap between strategic intention and operational reality narrows in near real-time rather than surfacing as variance in a quarterly review.

Together, these three layers constitute something qualitatively different from process automation: they constitute a cognitive prosthetic for operating in conditions that exceed the native capacity of human cognition.

The data architecture that underpins this kind of GaaS is equally distinctive. Conventional business intelligence demands clean, structured, well-labelled data to answer questions that have already been formulated. CAS-aware GaaS demands something epistemologically different: data that is wide enough, heterogeneous enough, and temporally granular enough to reveal questions that have not yet been asked.

  • This means network data capturing who interacts with whom and how those interaction patterns shift over time, since changes in ecosystem structure often precede changes in ecosystem behaviour by months.
  • It means high-frequency temporal data, because the leading indicators of bifurcation, i.e., the moments when a system tips from one attractor state to another, typically manifest as changes in variance and volatility rather than changes in mean values, and these are invisible in periodic snapshots.
  • It means cross-domain data that combines sources not conventionally analysed together, since weak signals of the highest strategic value often emerge at the intersection of domains that no single analyst would routinely hold in view simultaneously.
  • And it means systematic attention to absence: what is not happening, which actors are conspicuously quiet, which volumes have declined outside seasonal patterns.

In a complex system, the dog that does not bark is frequently the most informative signal of all. Two constraints must be acknowledged honestly: complex systems are never fully observable, so the architecture must treat incomplete information as a normal operating condition rather than a problem to be solved; and observing a system changes it, as actors who know they are being monitored adapt their behaviour, which means the data model must itself be adaptive, continuously updating its assumptions about what the signals mean as the ecosystem responds to the act of observation.

The market opportunity here is substantial and, for now, largely unoccupied. The market for GaaS as efficiency infrastructure is already competitive and will become more so as the underlying models commoditise. The market for GaaS as a navigation instrument for complex adaptive environments is, by contrast, almost entirely open. The reasons are structural: building it requires not only AI engineering competence but genuine fluency in complexity science, which is to say, in the theoretical frameworks of emergence, edge-of-chaos dynamics, co-evolution, and requisite variety that give the approach its intellectual coherence. Organisations that are positioned at the intersection of these two bodies of knowledge, and that can translate complexity science from an analytical framework into an operational one, hold a differentiating capability that is difficult to replicate by assembling AI engineers alone. The offering that emerges from this intersection is not, ultimately, a technology product. It is a new kind of strategic capability: the ability to help organisations sense the shape of complexity, maintain the productive tension between order and adaptive capacity that complexity theorists call the edge of chaos, and act with strategic intent in environments where the terrain is changing faster than any static plan can accommodate. That is the genuine promise of GaaS, and it remains, for the moment, largely unredeemed.

Padi UMKM — A Complexity Case

When I first designed what later became Padi UMKM, I did not do it in a boardroom. I did it at home, during long months of WFH in the middle of the Covid-19 pandemic. I drew the system on papers spread on the floor. At that time, my head was full of ideas about ecosystems, complexity theory, and complexity economics. I was not thinking about building another digital platform. I was thinking about how economic coordination itself breaks down under systemic shock, and how new coordination patterns might emerge when old ones collapse. In that sense, Padi UMKM was born less from a product mindset than from an ecosystem mindset, with complexity theory consciously in the background.

When the pandemic hit, what collapsed was not only the economy. What collapsed was the coordination logic of the economy. Supply chains broke, demand evaporated, SMEs lost access to markets, and institutions discovered that their standard operating procedures were designed for stability, not for systemic disruption. Many organisations reacted by accelerating digital projects, launching platforms, and optimising internal processes. That helped, but it did not address the deeper problem. The economic ecosystem itself had lost its organising structure. Actors that were rational in isolation could no longer produce coherent outcomes collectively. This is how complex systems behave under stress: when established coordination patterns fail, local rationality no longer aggregates into systemic order.

Padi UMKM did not start as a brilliant digital product idea. It started as a response to a coordination failure across a fragmented system of SOEs, SMEs, banks, regulators, ministries, and development agencies. All were acting with good intentions, yet through incompatible logics, timelines, and mandates. The system was not short of initiatives; it was short of coherence. In complexity terms, the economy had been pushed far from equilibrium, and the challenge was not optimisation but reorganisation. What was needed was not another tool, but a new pattern of interaction among heterogeneous agents.

The real innovation of Padi UMKM was therefore not the platform. The platform was the easy part. The digital workforce of Telkom Group can design platforms; that is an operational capability. The platform was necessary, and it became the core infrastructure of the ecosystem, but it was not the breakthrough. The breakthrough was the deliberate redefinition of roles within the economic system. SOEs must reposition their procurement operation into a capability of creating new market, i.e. an SME-based market structure. SMEs were not framed as beneficiaries of aid, but as economic agents that could be structurally integrated into formal procurement and value creation. Banks and financial institutions were not treated merely as lenders, but as part of an enabling architecture that combined financing with capability development and pathways to export. What changed was not a feature set. What changed was the pattern of interaction between economic actors.

The formal launching of Padi UMKM itself was not initiated by Telkom or by the Ministry of SOEs. It was planned within the nationwide BBI (Bangga Buatan Indonesia) program, because the central government needed a real, executable instrument to accelerate domestic economic circulation under crisis. Telkom showed a commitment to develop the platform, even though it was still imperfect at that time. The urgency was national, not corporate. This matters, because it positioned Padi UMKM from the beginning not as a corporate product launch, but as a systemic intervention embedded in a national recovery narrative. The early external promotion of Padi UMKM, beyond the internal SOE environment, was also driven by the BBI program. Over time, almost by systemic selection rather than by design, Padi UMKM became the de facto e-commerce infrastructure for BBI, as other platforms could not fit the specific institutional and ecosystemic roles required by the program.

From the beginning, we made a counterintuitive choice in the way the system was governed. Telkom deliberately limited its role to being the product and platform owner. The ecosystem itself was not branded as Telkom’s program. The community was symbolically owned by the Ministry of SOEs and by SOEs collectively. Even the name Padi UMKM did not originate from Telkom. This was not a political compromise; it was a strategic design choice grounded in complexity thinking. In complex systems, ecosystems tend to collapse when one actor over-claims ownership. When the platform owner also claims to own the ecosystem, other actors reduce their commitment, hedge their participation, or quietly resist. By stepping back from symbolic ownership, Telkom created space for other institutions to step forward. The platform provided the infrastructure, but the legitimacy of the ecosystem was deliberately distributed across actors.

At some point, something structurally interesting happened. The initiative crossed a threshold where no single actor could kill it anymore. The CEO of Telkom could not simply shut it down because the ecosystem had become institutionally embedded beyond Telkom. The Minister of SOEs could not dismantle it easily because it had become part of the official narrative of national economic recovery. The President could not disown it because it had been publicly positioned as a success story through BBI, PEN, and related programs. This was not political theatre. This was the moment when the system acquired path dependence. Once an initiative becomes embedded across multiple layers of institutional narrative and governance, it ceases to be a project and becomes part of the system itself. At that point, you are no longer managing a prograe. You are dealing with a living economic structure.

Value in Padi UMKM did not come from transactions alone. It emerged from the coupling of multiple layers of interaction. Transactions between SOEs and SMEs were reinforced by access to credit, by certification mechanisms that enabled formal participation, by development programmes that upgraded SME capabilities, and by pathways to export markets. None of these elements, on their own, would have been transformative. The transformation emerged from their interaction. This is how complex economies create value: not through linear pipelines, but through ecosystems in which different forms of capital, i.e. financial, institutional, social, and operational, reinforce one another over time.

Internally in Telkom, there was a structural separation of roles that proved critical. The Digital Business Directorate (DDB) operated at the product and business level. Its logic was operational: build, run, scale, monetise, and maintain the platform. Even as the platform owner and economic keystone, it remained only one agent within the broader ecosystem. In parallel, the Synergy Subdirectorate under the Strategic Portfolio Directorate worked at the ecosystem level. This role was not about features, roadmaps, or KPIs. It was about sensing emergent patterns of collaboration, mediating conflicts between institutions, and navigating collisions between policy signals and organisational incentives. In the early phase, the Synergy team also played a foundational role in organising cross-SOE agreements, preparing the multi-actor launch, embedding Padi UMKM within the BBI program, and connecting it with multiple SME build-up initiatives involving the Ministry of SMEs, the Ministry of Trade, and other institutions. This work was not linear project management; it was ecosystem orchestration under uncertainty.

In Indonesia’s context, the interaction between SOEs, SMEs, banks, and regulators is not merely complex; it is quasi-chaotic. Mandates overlap, incentives conflict, and policies evolve at different speeds and under different political pressures. In such an environment, precise prediction is an illusion. What becomes possible instead is navigation: sensing where constructive patterns of emergence are forming, dampening destructive feedback loops before they escalate, and shaping the boundaries within which the ecosystem evolves. This is not classical management. This is leadership under complexity.

As a result of its early success, there was a moment when the government, again through the BBI programme, asked to expand Padi UMKM to cover all government agencies (K/L/PD). On paper, this looked like success, with an enormous projected GMV. In reality, it carried a systemic risk. Full integration into the broader government procurement apparatus would have imposed rigid compliance structures and administrative constraints that could have frozen the adaptive dynamics that made the ecosystem work. The decision to return that expansion to LKPP, while positioning Telkom only as a platform provider for LKPP, was a deliberate choice to preserve modularity and flexibility over symbolic scale. In complex systems, scale without adaptability is not growth; it is fragility disguised as success.

What this experience ultimately taught us is uncomfortable for traditional management thinking. In complex economic ecosystems, you cannot engineer outcomes. You can only design conditions: boundaries, incentives, roles, and narratives that make constructive emergence more likely than destructive collapse. The platform mattered. The technology mattered. But what mattered more was the humility to accept that once an ecosystem becomes alive, you are no longer the architect standing outside the system. You are one of the agents operating within it.

The strategic lesson for C-level leadership is this. In times of systemic disruption, competitive advantage no longer lies primarily in having the most sophisticated product or the fastest execution. It lies in the capability to shape interaction spaces across institutions, sectors, and policy domains. Leadership shifts from control to stewardship. Strategy shifts from optimisation to navigation. And success is no longer measured only by ownership, but by whether the system you helped catalyse can survive, adapt, and continue to create value even when you step back.

That, ultimately, is what Padi UMKM represents. Not a digital product success story, but a case of how leadership, strategy, and technology can be recomposed to operate effectively in a complex, adaptive economy under crisis. It is an ecosystem in motion. It is Synergy in action.

Navigating Business at the Edge of Chaos

This is a speech preparation for the CIMA & AICPA Strategic Leaders Breakfast Talk, to be held in mid-February 2026, under the theme ‘Leadership in the Age of Disruption — Strategic Leadership for Modern Finance Professionals’. I will deliver the presentation from the perspective of complexity science and complexity economics, before exploring the practical implementations for management accounting professionals.

In current economic landscape, business must be perceived as the development of an ecosystem that operates as a complex adaptive system (CAS). Within this framework, autonomous agents, both internal to the firm and across broader business networks, possess the capacity for independent decision-making and activity. From this complexity perspective, phenomena such as VUCA (Volatility, Uncertainty, Complexity, and Ambiguity) and disruption are no longer viewed as external threats to be mitigated or overcome. Instead, they are recognised as engines of evolution and qualitative opportunities to redesign business architecture. Strategy shifts from the mere optimisation of saturated, linear models toward the cultivation of dynamic ecosystems that generate new value through the process of emergence.

The optimal zone for such innovation is the Edge of Chaos, which is a critical transition state where a system balances order and stability with disorder and change. It is precisely in this zone, rather than in a state of total equilibrium, where optimal innovation occurs. For the modern enterprise, the Edge of Chaos is not a threat to be avoided, but a strategic space to be occupied and, if necessary, intentionally created. Competitive advantage in this regime is defined not by scale or static efficiency, but by architectural flexibility and the velocity of learning in response to constant internal and external feedback loops.

Leadership within this complex environment requires a fundamental shift in identity toward that of an ecologist. The leader’s primary duty is no longer the top-down control of outputs, but the creation of conditions and cultures that enable teams to self-organise. This involves managing the delicate tension at the Edge of Chaos, introducing enough healthy friction to trigger innovation without descending into systemic anarchy. Rigid & brittle SOPs are replaced by simple rules or heuristics that guide autonomous decision-making amidst ambiguity. Leaders must facilitate safe-to-fail probing, i.e. launching multiple, simultaneous, low-cost experiments to detect strategic signals and opportunities that traditional analytical models inevitably miss.

Strategic management in the exponential era demands ambidextrous design, balancing the exploitation of core operations with the continuous exploration of new ventures through modular structures. This necessitates the orchestration of resources far beyond traditional organisational boundaries, incorporating partners, start-ups, and regulators into platform-based strategies. Strategy is viewed as a process of co-evolution, where the organisation constantly reinvents itself to remain congruent with a shifting environment.

Finally, Management Accounting (MA) serves as the vital navigation instrument in this journey through the Strategic Planning for Exponential Era (SPX) framework. MA must evolve to support dynamic feasibility, utilising Real Options Analysis to value investments as strategic options—the right to expand, delay, or pivot—rather than rigid, one-way capital bets. This implementation includes Agile Capital Budgeting, where funds are allocated to strategic “buckets” rather than granular, unproven projects. By abandoning the stagnation of rigid annual budgets in favour of Rolling Forecasts and Throughput Accounting, MA ensures that resource allocation is driven by real-time feedback and the velocity of value conversion. Ultimately, the most profound business developments are market-creating innovations that not only ensure sustainability but actively uplift the economy and quality of life for society

© 2026 Complexity Center

Theme by Anders NorénUp ↑