FROM INSIGHT TO ACTION:
THE AGE OF AGENTIC AI

The role of AI in business is evolving fast. What began as a tool for generating insights is moving toward execution, with systems now able to act on decisions, not just inform them. This shift toward agentic AI marks a new stage of operational maturity.

Gartner predicts that by 2028, 33% of enterprise software applications will incorporate agentic AI, up from less than 1% in 2024. This signals a rapid shift from assistive tools to autonomous systems embedded in everyday workflows.

In this edition, we explore how this transformation is taking shape with articles on:
• AI agents reshaping retail engagement and operations
• Data platforms enabling scalable autonomy
• Supply chain and finance deploying AI for real-time orchestration
• Long-run agents redefining governance and delegation

“With agentic, every function is impacted.”

How will AI agents impact shopping behaviors?
1. Consumers will use personal AI agents to search, compare, and buy products, creating new expectations for brand and website readiness.
2. Retailers are becoming embedded in AI ecosystems. Edouard cites the OpenAI and Walmart partnership, where conversational interfaces become new commerce channels powered by richer consumer insight than traditional transactional data provides.
3. Agentic commerce will transform internal retail operations, supporting sales teams, negotiations, promotions, and vendor management.

What are the key prerequisites for AI agent deployment?
1. Data readiness, with governed, standardized, semantically aligned datasets.
2. Process-readiness, with defined workflows where agents can operate effectively.
3. Tech readiness, from platform choices to build-versus-buy decisions.

Where will ROI emerge first in the TCG sector?
1. Customer service offers immediate returns through AI-driven self-care.
2. Marketing and content automation are accelerating campaign adaptation.
3. IT operations, from code migration to anomaly monitoring, represent major cost-saving opportunities.

Traditional reporting architectures struggle to support today’s AI demands, while modern models such as data lakehouses, data mesh, and data vault frameworks enable governed, business-ready ecosystems.

By unifying structured and unstructured data, these platforms power predictive analytics, GenAI applications, and decision automation. They are already transforming forecasting, customer engagement, and operational optimization. Yet technology alone is not sufficient. Success depends on overcoming legacy integration challenges, strengthening governance, and embedding strong data stewardship through phased implementation.

Agentic data platforms mark the next evolution, enabling AI systems to trigger decisions and workflows in near real time. Organizations investing now are building the intelligence layer required to compete in an AI-driven economy.

Supply chains have evolved from an operational backbone to a competitive battleground. Despite rising volatility, however, many organizations still underutilize AI in planning, logistics, and distribution. This practical guide is for leaders seeking to transform their supply chains from cost centers into strategic value drivers through targeted AI adoption.

The transformation starts with planning: AI enhances both Sales & Operations Planning (S&OP) and execution (S&OE), enabling companies to move from reactive forecasting to dynamic, data-driven orchestration.

In one home appliance case study, AI automated cross-functional demand, inventory, and bottleneck analyses, resulting in a 40% increase in operational efficiency, improved SKU-level visibility, and reduced inventory.

Beyond planning, AI unlocks optimization across the entire value chain:
• Intelligent inventory policies balance service levels with cost control.
• Logistics applications improve inbound material flows and outbound delivery performance.
• Digital twins and simulation models further enhance scenario planning and risk management.

Artefact shows how the convergence of generative and agentic AI is reshaping operations and value creation for banks, insurers, and fintechs.

High-ROI use cases include trading and portfolio optimization, customer engagement, document processing, and automated reporting. Early adopters report tangible gains: from 25% reductions in R&D cycle times to 40% productivity boosts in IT, as well as significant cost savings in AI-powered
customer service.

The real shift comes with agentic AI systems that can monitor transactions in real time, detect fraud proactively, optimize risk management, and automate workflows with minimal human intervention, moving AI from decision support to execution.

These new capabilities are accelerating innovation and responsiveness, but scaling them requires strong governance, data security, and compliance frameworks.

Victor Coimbra, Partner and Data Platform & IT Lead, Artefact LatAm, has been recognized in the Forbes Under 30 Brazil list for his outstanding contributions to AI innovation. He brings deep expertise in scaling AI solutions and building high-performance tech teams across international markets.

How long can AI sustain meaningful work before breaking down? Research shows task duration doubles every seven months. But production reality lags behind benchmarks. In enterprise testing, no AI outputs were usable without human cleanup. Documentation gaps, verification issues, and quality shortfalls were systematic. Cleanup time averaged one third of task duration.

This creates a productivity paradox. In controlled studies, AI users were 19% slower due to debugging, context switching, and quality remediation.
The root cause is architectural. As tasks lengthen, AI working memory saturates, causing context loss, contradictions, and compounding errors. Capability also varies widely by domain. Long-run AI is improving fast, but sustained autonomy remains limited.

Although AI degrades over long workflows, long-run agent operation can be extended via three architectural approaches:
1. Fresh-start cycling: Resets the AI when performance declines. Work is saved externally, sessions restart clean, and progress continues incrementally.
2. Selective memory: Preserves only essential context between sessions. 2. 2. Progress trackers, summaries, and change histories maintain continuity while reducing information overload.
3. Team coordination: Distributes work across multiple specialized agents managed by a central coordinator. By decomposing tasks and routing only relevant information, multi-agent systems significantly outperform single agents on complex work.

Each approach externalizes information that the AI cannot reliably retain internally. Cycling favors simplicity, memory systems favor continuity, and teams favor scale and specialization. Most production deployments combine all three.

As long-run AI becomes operational, AI assistants will shift to AI workers. Up to eight-hour workflows with full deliverables and implementations will emerge.

• Interfaces will evolve from conversational tools to delegation systems, with dashboards, checkpoints, audit trails, and recovery mechanisms replacing real-time prompting.
• Data quality and governance will become critical, as long-running AI can compound errors, overstep permissions, or operate on flawed assumptions.
• Bounded autonomy frameworks, including access controls, escalation triggers, and decision logging, will become mandatory.

Long-run AI is not yet fully mature, but it is operationally viable today for targeted use cases. Realistic opportunities lie in structured, attention-intensive work with measurable outcomes. Organizations that experiment deliberately, with strong guardrails in place, will gain an early advantage.