We need MORE than an intelligent superhighway to unlock value from AI

At WEF2026, one theme is coming through clearly: enterprises that unlock AI’s full economic promise are those investing now in the intelligent superhighway required to support intelligence at scale.

While AI can quickly deliver pockets of value, material financial impact shows up on the P&L only after leaders strengthen the foundations that allow intelligence to move reliably across the enterprise. That work takes discipline, and it creates durable, compounding advantage. • Yes The organizations already pulling ahead share a distinct pattern: - They modernize the digital core, clearing long-standing technology debt and updating legacy data, processes and skills. • This sounds good, but like “cleaning our data” this is not realistic to do holistically, we have to heat map the capability dependencies for the use-cases that support our transformation and ensure the digital core is upgraded to support any legacy gaps in the way of the imperative. • Capability gaps might be seen in people, process, technology and information. • Here’s how it is done: • • My quote on this We used to say ‘if it ain’t broke don’t fix it’… Updating to today: ‘If it’s fixed and in the way of transformation, we need to break it’ - They codify how work truly moves, capturing decisions, handoffs and exceptions so intelligence can operate with clarity and consistency. • This is essential but it stops short. • To achieve agentic outcomes, we must also codify the contextual understanding, decision logic and rearchitect the decision-making authority so agents can understand, reason and act aligned with what humans do today. Failing to complete this step results in first generation AI insights vs second generation AI action. • This weekend’s newsletter I contributed to on this: (was in the queue to publish since late Dec) The Day the Navy Bought Visibility into their Supply Chain Article commentary - They redesign roles and talent systems, recognizing that technology doesn’t transform an enterprise, people do, and ensuring humans remain decisively “in the lead.” • Humans design what we are transforming, in execution, agents will take the lead. • Not making this leap was fine in 2024, but I fully disagree it is ok today, the human led approach stops us at first generation AI (insights not action) and it does not scale. • From the link above: Trying to slow the AI down or speed the humans up sums up the mistake perfectly. The industry isn't taking accountability that it is primarily OUR fear of action that is causing our agentic pilot's purgatory. Gen 2 AI creates agentic solutions to solve a problem with action. Defining the decision criteria and permission to act is the point of the entire solution. Instead of taking this leap we are satisfying our fear with inaction and blaming the technology for our data only and human bottlenecked Gen 1 solutions that cannot scale. • We must do the hard work to codify our thinking with guardrails and the action path (including determining when and if a human needs to be involved), and agents MUST execute these well-defined use cases in order for our decision speed and momentum to scale our Agentic AI solutions. Failing to do this is why our pilots end up in purgatory. Our human cognition and our organizational decision-making authority model do not scale. • An agent should become the mostly self-sufficient expert on its assigned task, then it can scale… laddering them up creates a “super intelligent” network of action that drives our business forward into competitively advantaged speed and agility. • An earlier newsletter I cowrote write up on this concept: https://www.linkedin.com/pulse/enough-intelligence-shaping-destiny-without-digital-gods-yrlac/?trackingId=bLPX441JTJOzknsuqKZqzQ%3D%3D - They rebuild operating models for intelligent execution, treating AI as a shared enterprise capability rather than a collection of function level pilots. • Yes, including the action steps • AI must be a federated capability across digital teams, executed with a common core and an innovation funnel to … • Innovate emerging AI capability pilots in controlled experiments (ex: A2A) • Co-Create developing AI capabilities with the AI center of excellence (ex: specialized AI Task Agents) • Federate mature AI capabilities leveraging off the shelf internal and external standardized AI solutions (ex: RAG) This shift, from Siloed AI to Structural AI and ultimately Systemic AI, is where value begins to compound continuously. Intelligence becomes part of the organization’s operating system, accelerating decisions, strengthening performance and opening new avenues for growth. The message for leaders is clear: AI rewards commitment and patience. Those building the intelligent infrastructure now will own the fast lane for the next decade. Those who hesitate won’t just wait longer for returns, they’ll be structurally outpaced as value, talent and customers gravitate toward organizations ready for scale. • True when we do all 5 … This article misses 3 and 4. • Focused on the use E2E processes and use cases that will drive our imperatives (value streams) • Mapped to the detailed interconnected capabilities required to achieve the outcomes and our gaps across people, process, information and technology • Drawn out to map codify the contextual understanding, decision logic and action paths • Rearchitecting trust and decision authority to empower agents to act • Responsibly federating our AI capabilities in partnership across the broader digital organization aligned with our maturity in each capability area. • At Ecolab we have started all except #4… • Your team is doing the E2E leadership work for our value streams • We are driving the value streams, and my team will help to map the interconnected capabilities required to achieve the outcomes of each • We have good progress advancing our thinking to codify understanding, decision logic and action in efforts like BiC. It is not yet an implemented enterprise capability and most of the industry hasn’t started this journey yet. • We have work to do on our culture to do the hard work to rearchitect trust and decision authority • My team is actively working to codify our maturity in partnership with Naveen so we can engage the broader digital organization in Agentic development.

We need an intelligent superhighway to unlock value from AI

Manish Sharma

Chief Strategy and Services Officer at Accenture | Board Member January 22, 2026 Every executive today is living through some version of artificial-intelligence whiplash. The technology is moving at highway speed. Inside most companies, however, progress remains slow and uneven. Many experience the business equivalent of driving on a congested highway that blunts the speed, comfort and safety that technologically advanced vehicles were designed to deliver. The limitation is not the technology but the roadbed. Decades of data debt, brittle systems, undocumented processes and outdated skills form a congested highway no model, no matter how powerful, can outrun. Until leaders confront this reality, the gap between AI’s promise and its payoff will continue to widen. To take full advantage of what AI has to offer, enterprises need an intelligent superhighway. AI is advancing faster than organizations can absorb it. To capture anything close to its economic promise, companies must build modern, connected and governed internal infrastructure designed for scale. Without it, AI remains stuck in enterprise traffic. 👆 This is first gen AI thinking, it stops short of codifying semantic understanding, logic and action with a culture that embraces rearchitected decision rights. The stakes are rising. The Accenture Pulse of Change finds nearly nine in ten (86%) organizations plan to increase AI investment in 2026, and most view AI as beneficial to revenue growth. Yet only 21% report redesigning end-to-end processes with AI at the core, our recent survey of 3,650 executives across 20 industries and 20 countries found. Systemic readiness, not ambition, has become the binding constraint. Our research and experience across some 6,000 AI engagements reveals five truths that separate companies accelerating ahead from those stuck on the shoulder. ✅ The end to end process is right (Step 1 for any transformation following BizBok, and companies like Danaher Business System (DBS) based on the Toyota Production System (TPS) 1. Material financial impact from AI is backloaded The first truth is that AI’s financial impact is backloaded. Meaningful value on the income statement follows the enterprise modernization required to support AI at scale, and that work takes time, at least two years or more. Early stages are dominated by sequencing efforts to clean data and fix processes so they reinforce rather than conflict. A major regional bank illustrates the point. After more than a year of pilots in enabling corporate functions with minimal returns, it adopted a practical roadmap linking eleven priority workflows through a unified intelligence layer that sits on top of the technology stack over an 18–36-month horizon. Results are now compounding, with a clear trajectory toward a materially positive return. An energy provider operating in a reliability-critical sector followed a similar logic. Rather than scaling early wins prematurely, it modernized its digital core and connected codified workflows end to end. The result was a 90% reduction in analysis time and a foundation capable of supporting intelligence rather than resisting it. ✅ End to end thinking is right, its important we don’t try to create a digital core that covers everything… it must be focused on reducing non-value steps/stages (waste) and the gaps in achieving E2E outcomes 2. Most organizations are not operationally ready Most organizations remain operationally unprepared for advanced AI because the way work actually moves through the enterprise is incompatible with intelligence at scale. We’ve found through our client work that about 70% of technology budgets still support legacy systems that slow the flow of information. While standard operating procedures exist, critical decisions, process handoffs and exceptions remain opaque and unstructured, embedded in emails, conversations and tacit judgment. This lack of codification limits reliability, governance and scale. Until decision logic and process flows are explicitly captured and integrated into systems, AI pilots perform well inside isolated tasks but falter when asked to traverse the enterprise. In practice, agentic operating procedures must increasingly mirror, and in some cases replace, standard operating procedures, often revealing where entire processes must be reimagined. Ecolab, for example, redesigned its lead-to-cash workflow with orchestrated utility, super and conductor agents that link sales, fulfillment and billing. ✅ The bridge between Structural and Systemic AI is the architecture. • Define the capability map intelligence will move through • Establish the semantic contracts agents rely on • Govern the operating model so agents, humans, and platforms behave predictably • Redesign roles, decision rights and flows so humans are correctly engaged, decisions have traceability and outcomes flow back in to create a self-learning and optimizing system 3. Success comes from strong foundations The third truth is that AI delivers meaningful impact only when built on strong foundations. Companies pulling ahead are not chasing the latest model. They invest in the conditions that allow any model to perform: a reinvention ready digital core characterized by clean data, modern architectures, disciplined governance, AI-enhanced cloud environments, semantic consistency, responsible-use guardrails, redesigned processes and a workforce equipped to partner with intelligent systems. As executives have discovered, confidence in AI outputs rises only when data provides consistent context - and better context drives better decisions. Leading UK bank NatWest Group demonstrates the effect. By replacing fragmented systems with a single, bank-wide data platform, the bank is creating a trusted data marketplace that feeds every part of the organization with governed, real-time data critical to better day-to-day decision-making and more personalized experiences for more than 20 million customers. 👆 Again this is first generation AI thinking, not focused on agentic action. Codification is the new enterprise architecture currency. This includes: • Semantic understanding • Decision logic and Policy constraints • Escalations (to a human), Exceptions (when to deviate) and handoffs (when to pass to another Agent or person) • Nonfunctional requirements • Outcome feedback loops and learning Once codified, these become machine navigable assets that power End to End intelligent workflows with: • Task level agents • Patterns • Knowledge graphs Ecolab’s L2C redesign is a good example of pushing past what the article highlights into ACTION – where outsourced solutions have aligned packaged solutions and partners that deliver on yesterday’s inefficient value stages (a part of the value stream that can be carved out) that were ripe for transformation. 4. AI value depends on reinventing talent and work The fourth truth is that unlocking AI value depends on reinventing talent and work. Technology does not transform enterprises; people do. In another recent Accenture survey, we found that only one-third of 1,320 executives say their talent strategy is fully integrated with their AI strategy. Most organizations still deploy AI into job structures never designed for human-machine collaboration, leaving roles ambiguous, incentives outdated and leadership behaviors misaligned. While more than 40% of executives report upskilling employees for AI-enhanced work, fewer than 10% are redesigning roles or responsibilities, according to our Pulse of Change survey. Leaders take a different approach. One financial services firm, for example, mapped work at the task level, revealing how shifting repetitive data processing to AI agents could unlock up to 30% more capacity for human creativity and insight. 👆 This ambition falls short…. Again referencing: https://www.linkedin.com/feed/update/urn:li:activity:7421011882062503938/ Trying to slow the AI down or speed the humans up sums up the mistake perfectly. The industry isn't taking accountability that it is primarily OUR fear of action that is causing our agentic pilot's purgatory. Gen 2 AI creates agentic solutions to solve a problem with action. Defining the decision criteria and permission to act is the point of the entire solution. Instead of taking this leap we are satisfying our fear with inaction and blaming the technology for our data only and human bottlenecked Gen This is one of the most consistent Chief Architect Network themes from last fall in Dallas and last spring in London. You cannot bolt intelligence onto an operating model built for a pre-AI era. • AI becomes a shared enterprise utility, not a function-run initiative • Architecture, Product, and Operations must converge • Governance shifts from approval based to signal based (where Ecolab has both technical and culture work to do, but we have a head start) This is where most enterprises stall, not on tech, but on operating model inertia and culture acceptance of our architecture of trust and decision authority. 5. A “future-ready” AI operating model is vital The fifth truth is that AI cannot scale inside an operating model built for a pre-AI era. Governance, decision rights, architecture and the relationship between business and technology must be redesigned. BBVA offers a blueprint. After unifying data, redesigning workflows, strengthening governance and restructuring roles, loan approvals fell from days to hours, personalization improved and predictive digital channels attracted millions of new customers. The breakthrough did not come from better algorithms, but from an operating model capable of absorbing intelligence at scale. Taken together, these five truths reveal the underlying requirements for intelligence to move through an enterprise, and why so many efforts stall despite heavy investment. The returns described in the first truth are realized only after organizations complete the other four truths: becoming operationally ready, building strong foundations, reinventing work and adopting a future-ready operating model. As with any major infrastructure project, accommodating the resulting higher volumes of intelligence traveling at greater speed with reliability should follow a predictable path. 👆💡A glimmer of “2nd generation AI” hope in this section about decision rights, but I don’t think he is taking it far enough. Rearchitecting decision rights for a smoother E2E process is right, but we must go further to redesign the architecture of trust for AI agentic execution specifically. This requires a depth of analysis to teach agents how to think on our behalf by codifying our human understanding, reasoning, guardrails, actions and the way to learn from feedback. AI requires a wholesale rewrite of enterprise roles, not just upskilling. • AI-fluent business roles • Human-machine collaboration patterns • Decision intelligence roles • Agent lifecycle and ownership structures • Product centric governance redesigned for AI velocity The organizations that treat this as a culture shift, not a training program, are the ones emerging into the Systemic AI stage. The three-phase path to systemic AI Progress from pilots to scale begins with siloed AI, where pilots sit in pockets. Many attain structural AI, the critical bridge where data, platforms, workflows and governance are rebuilt to carry intelligence across the enterprise. Only then can organizations reach systemic AI, where composite agents orchestrate work end to end, decisions accelerate and value compounds continuously. In effect, AI becomes embedded in the organization’s operating system. Few organizations reach this final stage. Most remain between on-ramps and unfinished roads, uncertain why progress has paused. The intelligent superhighway is the strategic infrastructure of the next decade. Companies that build it by codifying processes, modernizing data, workflows, governance, talent systems and operating models will widen their lead quarter after quarter. Those that hesitate will find the cost of delay is not temporary but structural, as value, talent and customers migrate to the organizations that rebuild their roads first. And that is the point. The technology is ready. The question is whether enterprises are. Leaders who act now will own the fast lane, gaining advantages that strengthen with every mile of progress. Those who wait will find themselves stuck in enterprise traffic, watching others define the next era of economic performance. Organizations that persist through the early flat stretch will build the systems required for returns to compound. AI rewards commitment, not impatience. Nobody wants a racecar in a traffic jam. 👆 Manish frames the intelligent superhighway as infrastructure. Extending that metaphor: • Data is the pavement • Workflows are the lanes • Governance is the speed limit and signage • Organizational talent passes their driver’s exam… • Transformation teams envision, design and build autonomous vehicles • Operating model is the highway system, modified for our new vehicles • Agentic systems (not human assistants) are new vehicles on our roads Most enterprises aren’t building these highways: they’re resurfacing county roads and expecting Formula 1 performance. Systemic AI demands more than modernization. It demands intentional, architectural design and a reformation of our E2E processes for a new type of independent worker that humans in new enterprise roles design. Manish was right about the concept of human-led, but he stopped short: we are human led in design and build with empowered Agents, not human led in execution with AI-Assistants. The Bottom Line Manish’s article aligns closely with what Chief Architects across the CAN community have been saying, but we expand it in bold: • AI value comes from E2E focused modernization, including carve-out value stages that can be performed by packaged solutions or vendors • Codification is the unlock • Platforms, processes, governance, and roles all need reinvention • Our architecture becomes our strategic lever for scale • Systemic AI is the future operating model, not a distant aspiration • Culture must support our re-architecture of trust and decisions Enterprises need an intelligent superhighway – but adapted for agents with an intelligent architecture. Without it, AI cannot travel anywhere at speed.