From budget battles to portfolio intelligence
Service demand management (SDM) facilitates internal and external service provision within large companies by aligning business needs with functional units. When embedded into the annual planning process and closely linked to financial metrics, it is an efficient way to foster self-steering while reducing costs and boosting service quality.
In the absence of a structured SDM process, many organizations find themselves with internal and external service provision that is opaque, costly, and difficult to steer. This is especially true for large organizations and those in process-intensive industries. The result is an inability to optimize at the portfolio level and build a foundation for scalable self-steering.
In our work with clients, we frequently see broad misalignment related to service execution. Business units complain about the high costs of service provision, a lack of transparency about the level of required services, and their inability to influence the cost and quality of service execution. Services are charged to business units using a variety of mechanisms (e.g., direct costs, project-related costs, overhead costs) that are complicated and/or ambiguous. Additionally, many companies operate myriad autonomous business units that reduce the potential for synergy. This leads to a lack of coordinated demand for internal and external service orders.
Picture this: Your central IT department just committed to spending several million dollars on an ERP (enterprise resource planning) modernization project. Three months later, the R&D segment of a business unit requests a data analytics platform requiring similar infrastructure. Procurement simultaneously negotiates three separate cloud contracts, none of which leverage enterprise-wide volume. Each group made “rational” decisions, but collectively, they destroyed value.
Demand dysfunction stems from various factors, with companies experiencing one or more of the following:
Organizations sometimes see SDM and demand standardization as a bureaucratic burden: a way to ensure compliance, enable comparisons, and reduce transaction costs. In reality, SDM enables coordination previously impossible. When demands are expressed in a common language with comparable data, organizations can make portfolio-level decisions that optimize total value rather than local outcomes.
When everyone uses the same framework to articulate needs, several things happen:
Planning cycles exist to align resource allocation with strategic priorities. Finance owns the planning calendar, which cascades from strategy-setting (board level) through business planning (executives) to functional budgeting (departments). SDM fits within this construct as a way for functions to understand what business departments need.
Unfortunately, this sequence is backwards. To achieve self-steering capabilities, organizations must invert the relationship: demand becomes the organizing principle for planning, and budgets become an output of demand prioritization (rather than an input constraint). When planning revolves around demand rather than budget, several capabilities emerge:
The result is an organization that “steers” itself. The mechanisms for sensing changes in the environment (service demand pipeline), deciding on responses (governance forums), and executing adjustments (dynamic reallocation) operate continuously rather than in discrete annual episodes.
SDM systems capture, route, and track demands. They’re operational tools that replace email and spreadsheets with structured interactions and workflows, improving efficiency and visibility but not fundamentally changing decision-making.
However, when combined with AI, SDM systems become strategic intelligence platforms that surface insights humans can’t see, predict outcomes humans can’t anticipate, and greatly enhance decision-making. (Note that human-machine interaction is needed for best results. If demand management processes are simply handed over to technology, the results will be lackluster.)
The transformation occurs across the following three dimensions.
Traditional SDM is backward-looking, with organizations analyzing what happened to understand patterns. AI-enabled systems are also forward-looking, predicting what will happen to enable proactive responses. For example, an AI system ingests five years of demand data across all functions, external market signals, and internal operational data. It identifies that demand for specific IT services increases 40% in the six months preceding major capital projects, creating an early-warning system that enables IT to “pre-position” capacity.
This capability does more than improve response times. Instead of “We need this in three weeks,” the conversation becomes “Based on your strategic plans, you’ll likely need this in six months — let’s optimize the solution together.”
Humans struggle with complexity. We can’t simultaneously optimize across dozens of variables while considering hundreds of interdependencies — AI can. For example, R&D wants to accelerate a product development program. This demand means IT needs to provision test environments, procurement must expedite component sourcing, manufacturing must reconfigure production lines, and finance must adjust cash-flow forecasts. A human planner likely considers these sequentially, but AI considers them holistically, identifying optimal sequencing and resource allocation across all affected functions to minimize total time and cost.
This is where self-steering becomes powerful. The organization doesn’t need executives to orchestrate complex, multifunctional responses manually. Instead, the AI-enabled SDM system identifies optimal orchestration patterns and recommends them to human decision makers for approval.
The ultimate form of self-steering is autonomous decision-making for routine demands that follow predictable patterns. For example, an AI system learns that routine infrastructure requests, such as additional storage capacity, standard software licenses, and common maintenance services, follow predictable approval patterns. Once confidence thresholds are met, the system can auto-approve these demands and route them for immediate fulfillment, escalating exceptions that require human judgment.
This isn’t about removing humans from decisions. It’s about directing human attention to decisions where judgment matters (e.g., strategic trade-offs, novel demands, exceptional cases) and letting AI handle high-volume routine decisions where judgment adds minimal value.
Note that organizations often struggle to decide whether to build SDM capabilities and then add AI or implement them simultaneously. The answer depends on maturity. If your organization is still using ad hoc demand processes, build foundational SDM first. (AI requires structured data; without standardized processes generating quality data, AI will amplify chaos rather than reduce it.) If your organization is using manual SDM, then implement AI opportunistically in high-value areas (demand forecasting, duplicate detection, portfolio optimization) while continuing to mature foundational processes. If your organization has mature SDM, deploy AI aggressively to achieve autonomous capabilities in routine areas and advanced analytics in strategic areas.
The Arthur D. Little (ADL) SDM framework calls for creating a demand architecture and governance structure, outlining a process for integration, collecting data/metrics, and conducting a change management program (see Figure 1).
cloud technology axon
This pioneering initiative is designed to support both organic and ino...
